Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post Book of Scouting Reports: The AI Agent Tech Stack appeared first on CB Insights Research.

To create the 2025 list, we analyzed 10 years of CB Insights’ Business Graph data, evaluating 12,000+ venture firms on portfolio outcomes (unicorns and exits), share of rounds led, portfolio quality via Mosaic Score, capital efficiency, and entry discipline. Smart Money VC portfolios offer a front-row view of where the sharpest investors are placing their bets. Use the list as an early indicator to spot emerging markets and promising founders.

Which VC firms are on the Smart Money list?

Firms are presented in alphabetical order.

1. Accel
2. Andreessen Horowitz
3. Bain Capital Ventures
4. Battery Ventures
5. Bessemer Venture Partners
6. Felicis
7. First Round Capital
8. Founders Fund
9. General Catalyst
10. Google Ventures
11. Greylock Partners
12. Index Ventures
13. Institutional Venture Partners
14. Kleiner Perkins
15. Lightspeed Venture Partners
16. Meritech Capital Partners
17. New Enterprise Associates
18. Norwest Venture Partners
19. Notable Capital
20. Redpoint Ventures
21. Salesforce Ventures
22. Sapphire Ventures
23. Sequoia Capital
24. Spark Capital
25. Thrive Capital

CBI customers: See Smart Money’s latest bets

How Smart Money VCs are outperforming the market

Our 2025 edition of Smart Money VCs:

• 6.5x more likely than the average VC to back a future unicorn
• 2.2x more exits per firm, either through M&A or IPO
• 2.3x higher share of rounds led, shaping pricing and syndicates

Smart Money syndicates amplify signal. The top pairs share dozens of portfolio companies — Sequoia & Andreessen Horowitz (43), General Catalyst & Andreessen Horowitz (42), and Sequoia & Lightspeed (36). Most widely backed across the cohort: Chainguard, Figma, and Wiz (each with 7 Smart Money backers).

Smart Money firms have also been the dominant backers of the AI wave — they backed 52% of new AI unicorns in 2023, 73% in 2024, and 77% in 2025 YTD — and that exposure is translating into outlier outcomes.

Since 2015, Smart Money VCs have backed 80 companies that exited at $10B+ — roughly 100x the $100M median exit. The largest Smart Money exits include Uber ($75.5B, 2019), Coinbase ($65.3B, 2021), and Coupang ($56.6B, 2021).

Mosaic shows where they’re headed next. Smart Money portfolios skew to higher Mosaic Scores — CB Insights’ 0–1,000 predictive rating of private-company health. The average portfolio Mosaic is 628 — about 2.6x the VC norm.

And the edge is most visible at the very top of the distribution: more than 65% of companies in the top 1% of Mosaic Scores are backed by a Smart Money VC. Top firms by average portfolio Mosaic include Meritech (759), IVP (741), and Thrive Capital (688). Standout companies in 2025 include Zepto, Bilt, Glean, Rippling, and Anthropic.

Where Smart Money is deploying now

Smart Money is still leaning into AI — especially agentic applications.

Over the last 18 months, agent-related categories led by deal count: coding agents and copilots (28 deals), agent development platforms (24), enterprise workflow agents and copilots (20), and legal agents and copilots (17). Infrastructure remained active as well, with 17 deals into LLM developers. Top recent AI deals by Mosaic include Glean (enterprise AI agents), Augment Code (coding AI agents), and ElevenLabs (voice AI).

Our M&A probability model points to cybersecurity as the most likely near‑term exit pool among Smart Money portfolios, with companies like Tenex.ai ranking highest. Activity is accelerating — highlighted by Google’s $32B acquisition of Smart Money–backed Wiz in March 2025. For acquirers, targeting Smart Money portfolio or syndicate companies can streamline diligence and post‑deal integration.

Outside the US, cybersecurity is also drawing Smart Money. Since Jan’24, Accel (84 deals), General Catalyst (64), and Lightspeed (55) are the most active by ex‑US deal count; their portfolios include companies like Tines, Cato Networks, and Torq.

Methodology

What is the CB Insights Smart Money list?

The Smart Money list is an unranked collection of the top 25 venture capital firms worldwide. We analyzed 12,000+ venture investors with 10+ unique portfolio companies using 10 years of CB Insights’ Business Graph data (2015–2025) to surface the highest performers via our Smart Money Index.

What makes a VC “smart”?

​​Comparable lists in other asset classes rank firms based on investment performance, but returns data is hard to come by in the VC world, and rates of return can be easily manipulated.

Our methodology factors:

• Portfolio outcomes — unicorn count/share and exit count/share
• Deal leadership — share of rounds led
• Portfolio quality — average CB Insights Mosaic Score
• Capital efficiency — portfolio value created per dollar raised
• Entry discipline — median stage at first check

Inputs were normalized and combined into the Smart Money Index. The top 25 became the 2025 Smart Money cohort.

What can I do with this collection?

Explore the Smart Money Expert Collection on the CB Insights platform to filter deals, build screens, and make faster decisions.

If you are a venture investor and want to submit data on your portfolio companies to allow us to better score you in the future, please reach out to researchanalyst@cbinsights.com.

RELATED RESOURCES FROM CB INSIGHTS:

• All 100 AI unicorns since ChatGPT launched
• State of Tech Exits H1’25
• The top angel investors in AI
• Are AI unicorns starting to move beyond hype?
• Book of Scouting Reports: Enterprise AI Agents

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post CB Insights Smart Money 2025: The top 25 VCs outperforming the market appeared first on CB Insights Research.

The market is already worth more than $2B, and appetite for these tools continues to accelerate. New players are flooding in: IDE startups are launching their own agents, 12 brand-new coding AI agent companies have been founded since 2024, and every major cloud provider and LLM developer has been rolling out offerings.

So who’s leading today, and who’s gaining ground fastest?

Using CB Insights’ revenue data, we measured the current size of the market and estimated market shares for players in the space. Download our book of scouting reports for an in-depth analysis of every private player with disclosed revenue in the market.

If you are active in the coding AI agent & copilots market and want to submit your company’s revenue data, please reach out to researchanalyst@cbinsights.com. 

See all coding AI agent & copilot companies ranked by revenue

Key takeaways

• Coding AI agent & copilot is a highly concentrated market, with the top 3 players currently holding just over 70% of the market. GitHub (owned by Microsoft) leads with an estimated $800M in ARR generated from its AI-powered coding offerings, demonstrating the power of superior distribution in the agentic AI space. With close to 40% of players showing low commercial maturity scores (emerging or validating), we expect leaders to be challenged and risk losing market share unless they turn to M&A to maintain their position — and technological edge.
• Explosive growth creates a dynamic leaderboard, with companies reaching and surpassing $100M in ARR at record pace. For example, Anysphere was generating $500M in ARR by June this year, up from $100M as of December 2024, a level it reached just 12 months after launching its product. Similarly, Anthropic scaled its AI coding solution (Claude Code) from 0 to $400M in ARR in just 5 months. This is adding more pressure on leaders as it highlights the low barriers to scale in this market, with new entrants able to win material share very quickly.
• The pie keeps getting bigger, with companies projecting top-line growth of 12x on average this year. Lovable recently said it expects to reach $250M in ARR by year-end, up from $10M at the start of 2025, and projects $1B by mid-2026 — a 100x increase in just 18 months. However, higher costs and reluctance from enterprises to adopt usage-based pricing could slow growth in the space or require significantly more funding to stay in the race.

Market overview

The coding AI agents & copilots market consists of AI-powered solutions that help software developers write, fix, test, and maintain code. These tools offer features like intelligent code completion, natural language code generation, automated testing, code review, debugging assistance, and technical debt management. Many solutions integrate directly with popular IDEs and development environments, while others operate as standalone agents or chat interfaces. The market includes both general-purpose coding assistants and specialized tools for specific programming languages, frameworks, or development workflows.

We count close to 100 players in this market, with a mix of early-commercial-maturity pure players (~40%), recently minted unicorns such as Anysphere and Lovable, leading LLM developers, and most big tech companies.

They have raised a combined $2.1B in equity funding so far this year, already surpassing the $2B raised last year. Traction in the market is also reflected in its average Mosaic score (a measure of company health) of 633, well above the 370 average across all private companies.

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post Coding AI agents are taking off — here are the companies gaining market share appeared first on CB Insights Research.

Companies that hit $100M+ ARR in months, like Anysphere (maker of Cursor) and Lovable, now face LLM inference costs growing up to 20x, forcing rate limits and price hikes, and putting reverse acqui-hires (hiring founders and licensing the tech) on the table as some founders seek exits.

Using CB Insights’ data on company momentum, exit probabilities, and customer sentiment, we analyzed how the coding AI market is adapting to this economic shock and what other AI agent companies (and their backers) can learn:

• Reasoning models spark vibe coding’s explosive growth
• Reasoning token shock pushes adoption of new pricing models
• Margin pressure drives consolidation of talent in the coding AI agents market
• Open models and usage-based pricing offer solutions to the market’s current challenges

Reasoning models spark vibe coding’s explosive growth

The coding AI agents and copilots market has been on a roll, generating an estimated $1.1B in revenue in 2024 and minting unicorns in as little as 6 months, which is 4x faster than the AI industry average.

Anthropic’s release of Claude 3.5 Sonnet in June 2024 has primarily driven this early momentum. This technology helped developers transition from autocomplete to partial delegation of coding tasks with a model that could reliably call tools and handle multi-file edits.

But it is the emergence of reasoning models, and specifically Anthropic’s Claude 3.7 Sonnet’s reasoning mode in February 2025, that made vibe coding possible — giving a high‑level goal and delegating multi‑step implementation to the AI. Developers could now set goals like “make this component responsive” or “add error handling throughout” and let the AI plan and execute the changes, sparking explosive growth in the space:

• Anysphere’s ARR grew 5x in 6 months, from $100M in December 2024 to $500M in June 2025.
• Replit’s ARR increased from $10M at the end of 2024 to $144M in July 2025.
• Lovable became one of the fastest-growing software startups, reaching $100M in ARR just 8 months after launching.

See all coding AI agent companies ranked by revenue

Reasoning token shock pushes adoption of new pricing models

As revenue surged on the back of reasoning, costs rose even faster.

Reasoning models inflate output‑token volume roughly 20x, according to Artificial Analysis. Because inference is billed per token — and output tokens are typically priced higher than input — that surge translates directly into higher compute cost. Anthropic’s May 2025 step‑ups on Sonnet 4 and Opus 4 (priced at roughly 5x prior models) added further pressure just as adoption was accelerating.

This is particularly impacting enterprise deals, which businesses often negotiate on an annual, per‑seat basis. That structure leaves vendors carrying the risk of uncapped compute costs while revenue stays fixed.

Using CB Insights Customer Sentiment data, we find most contracts fall between roughly $6K and $100K a year, with a median around $25K for a 50‑developer team. While margins once sat at 80%-90% on these contracts, compute costs from reasoning models can flip margins deeply negative.

The strain showed up quickly. Cursor tightened rate limits and introduced overage charges despite crossing $500M in ARR, prompting backlash and refunds. Anthropic throttled Claude Code after individual users exceeded $10K in monthly compute on $200 plans.

Vendors are shifting to pass‑through and usage‑based pricing to align revenue with compute cost. Companies employing usage‑based approaches show stronger momentum in our Mosaic data (median Momentum Mosaic of 683 vs. 671 for the broader market), but enterprise buyers are pushing back on variable bills and month‑to‑month swings.

Expect coding AI agent vendors to adapt pricing and GTM: moving to seat‑plus‑usage hybrids, stricter per‑seat compute guardrails, and model tiering that reserves reasoning for high‑impact work. ARR growth will moderate as flat‑fee expansion gives way to usage‑aligned pricing.

Margin pressure drives consolidation of talent in the coding AI agents market

Reasoning-driven margin compression is forcing consolidation in a category that has seen dozens of new entrants over the past 12 months.

Traditional acquisitions aren’t off the table, but acqui‑hires and reverse acqui‑hires have become the most active exit structures recently — albeit with trade‑offs.

OpenAI and Anthropic have logged 3 acqui‑hires since early 2025. Across AI, recent moves (e.g., Microsoft–Inflection AI, Amazon–Adept, and Meta–Scale) signal a tilt to talent‑plus‑license amid potential antitrust scrutiny. In coding AI agents, Windsurf’s failed sale and Google’s follow‑on reverse acqui-hire underscore the pattern of buyers taking teams and leaving products behind.

In these deals, acquirers hire the team and license the tech, leaving customer contracts and infrastructure — and the associated compute liabilities — outside the transaction. What they’re buying isn’t raw model IP; they’re buying proven operators with successful track records.

CB Insights’ exit probability analysis points to the next likely targets: companies with high Momentum Mosaic scores but lower probabilities of traditional exits.

The likely cause: private‑market valuations have outrun what strategics or public investors will pay given reasoning‑driven margin pressure, product overlap, and antitrust scrutiny — making full‑company M&A or near‑term IPOs harder to underwrite.

Seven stand out as potential targets: Sourcegraph, Augment Code, JetBrains, Qodo, Lovable, Cognition, and Harness.

Expect more reverse acqui-hire deals over the next few quarters as big tech continues to push for talent while coding AI agent companies struggle under margin pressures.

Open models and usage-based pricing offer solutions to the market’s current challenges

Against that backdrop, two levers dominate today: open models and usage‑aligned pricing. Here’s how each is playing out — and where it falls short.

Open models cut costs, but enterprise requirements slow adoption

Moonshot AI’s Kimi K2, Alibaba’s Qwen-Coder, and Z.ai’s GLM-4.5 approach Claude on coding tasks at a fraction of the cost, and OpenAI’s gpt‑oss goes a step further by offering a model that can run on consumer hardware.

Yet users need to access these models either through self-hosting or a third party. For enterprises, that means fresh security reviews, stringent uptime service level agreements (SLAs), multi-hour agent-run testing, and new infrastructure to manage.

The result is slower adoption, especially for six‑figure contracts that expect Claude‑level reliability.

Usage-based pricing fixes vendor margins, but most enterprises resist variable bills

Buyers tell us that token-metered pricing is difficult to budget, and expectations around costs for these tools are already set. CFOs want to anchor budgets and avoid month-to-month swings tied to release cycles, while usage-based pricing is the exact opposite.

In the near term, expect a shift from per‑message metering to effort‑based task pricing: agents quote a fixed rate for a defined outcome (e.g., “add error handling across this service” or “convert this component to TypeScript”), bundling planning, tool calls, and verification into a single charge with a visible pre‑estimate. Tasks are tiered (S/M/L) with caps on reasoning usage and admin‑approved overages, giving CFOs predictable bills while keeping compute under control.

This dynamic won’t be limited to coding

Other agent categories with surging usage are likely to rework pricing and contracts as reasoning costs mount.

Customer service is already operating on usage/outcome models. For example, in May 2025, Salesforce’s Agentforce shifted prices from $2 per conversation to a hybrid-usage Flex Credits system, tying credits to necessary actions for an outcome. Zendesk did a similar shift in pricing strategy in November 2024. Yet reasoning‑heavy workloads still create margin risk when the compute to achieve a resolution outstrips the value captured.

Beyond customer service, expect similar recalibrations across legal, healthcare, and sales agents. Outcome‑ or usage‑based models don’t fully eliminate compute risk. Explosive top‑line growth can mask deteriorating unit economics as reasoning workloads scale, and recent mega‑rounds may not be enough to foot the bill. Many players will reprice, add stricter usage guardrails, or raise additional capital to stay in the game.

If you are a coding AI agent startup and want to submit your company’s revenue data, please reach out to researchanalyst@cbinsights.com.

RELATED RESOURCES FROM CB INSIGHTS:

• All 100 AI unicorns since ChatGPT launched
• State of Tech Exits H1’25
• The top angel investors in AI
• Are AI unicorns starting to move beyond hype?
• Book of Scouting Reports: Enterprise AI Agents

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post The summer of vibe coding is over — How reasoning models broke the economics of AI code generation appeared first on CB Insights Research.

The endgame is fully orchestrated construction sites where AI coordinates everything from material delivery and site preparation to assembly and inspections. While implementation challenges persist — such as the need for sufficiently advanced AI systems, integration with legacy software, and inconsistent connectivity at remote sites — progress toward this vision is moving forward.

Recent surveys show 92% of construction professionals report improved decision-making capabilities with reality capture technology. Similarly, leading firms like Skanska have developed autonomous AI agents that deliver safety guidance in real-time, while Turner collaborated with Versatile to use its AI-powered crane attachment for improved crane utilization, material handling, and production rates.

Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post 280 AI companies automating the construction industry appeared first on CB Insights Research.

The tech’s potential has reshaped the innovation landscape  — with new AI unicorns now outnumbering non-AI unicorn births.

Below, we analyzed the 100 unicorns to understand how the cohort stacks up, top exits, leading investors, and more.

Download the full Book of Scouting Reports for in-depth analysis on the 100 AI unicorns. Combining CB Insights’ proprietary data and AI, scouting reports provide insight into each company’s:

• Key takeaways (including opportunities and threats)
• Valuation
• Headcount
• Outlook (Mosaic score and Commercial Maturity) 

Key takeaways

• Nvidia is building its AI ecosystem through strategic investments: Nvidia has backed 24 of the 100 AI unicorns (nearly 1 in 4), more than any other investor as it secures demand for its chips. Big tech overall has invested in over one-third of these unicorns, with Google (15), Microsoft (7), and Amazon (3) also placing strategic bets on their future customers and partners.
• LLM developers and AI agents dominate the unicorn landscape: Large language model development leads with 12 unicorns, while AI agent development platforms account for 5. These foundational infrastructure categories provide core building blocks, while coding AI agents & copilots (8 unicorns) represent one of the most successful application layers built on top of this foundation. Note: categories are not mutually exclusive. 
• Robotics emerges as an AI frontier: Physical AI is gaining momentum with 3 robot foundation model developers (Skild AI, Physical Intelligence, World Labs) and 3 humanoid robot developers (Figure, Unitree Robotics, Zhiyuan Robot) among the robotics companies achieving unicorn status. This signals strong investor confidence in AI’s transition from digital to physical applications.

The post All 100 AI unicorns since ChatGPT launched appeared first on CB Insights Research.

Among them, a few are set to win big, having placed early bets on Cursor, Cognition, and Sakana AI, and 200+ other AI companies.

Using CB Insights data, we analyzed which angel investors are building the most robust AI portfolios, and asked them to share their views on where AI is heading. Below is our ranking of the top 15 angel investors based on AI activity since January 2024, and key takeaways on the list.

Make sure we’re representing your full portfolio by reaching out to researchanalyst@cbinsights.com to set up a review of CB Insights’ coverage of your investments.

Track these investors’ deals in real time

Key takeaways

• Elad Gil tops the ranking with 36 AI deals since January 2024, ahead of Gokul Rajaram and Jeff Dean, each with 30 deals. Gil has scored heavy-hitters in the genAI space, such as AI search engine Perplexity, coding agent Cognition, and AI data platform Scale. 

• Nearly 90% of the top angels’ AI investments target the application layer. These companies build on top of foundation models to solve specific use cases, including browser agent Yutori, computer vision development tool Roboflow, and enterprise search platform Onyx AI, each backed by 3 or more top AI angel investors.

“Over the next 1-3 years, I expect the application layer to be very fruitful for AI startups. There are a tremendous number of spaces that were hitherto inaccessible for software but now are opened up thanks to AI.” — Gokul Rajaram

• 40% of companies backed by the top AI angels are founded by big tech veterans. This includes Meta’s 14-year product design leader Julie Zhou who founded the AI-powered analytics platform Sundial, and Nvidia’s 8-year engineering lead Ambuj Kumar, who launched AI security agent startup Simbian.

“AI is a relentless technology. Things are moving so fast and the models are getting better every day. Whenever you have a space that’s moving so quickly, the one constant you can bet on are the founders who are capable of navigating this change. My strategy is to simply find the founders building companies that are the right vessel to deliver the dramatic progress we’re seeing in model capabilities.” —Kulveer Taggar

• Angels are positioning early in multimodal AI’s next wave. With modalities like voice attracting $371M as of July 2025 funding and humanoid robotics projected to reach $2.3B by year-end, investors are betting on the technologies that will power the next phase of human-AI interaction. Our top angel investors have backed World Labs (world model developer), LiveKit (multimodal AI development platform), and Hamming (AI voice agent) since January 2024 as examples of this shift towards multimodal in action.

“The next generation of AI leaders will be cross-functional teams with deep vertical expertise. As foundational models continue to commoditize very fast, the edge will go to founders who work backwards from user pain points and harness emerging modalities like audio, robotics, world models.” —Mehdi Ghissassi

The post The top angel investors in AI appeared first on CB Insights Research.

Underpinning this momentum is an emerging tech stack. Infrastructure layers — from foundation models to oversight — are helping enterprises build, deploy, and manage AI agents more effectively.

Using the CB Insights Business Graph and proprietary signals, we mapped 135+ promising private companies building infrastructure for AI agents.  

Below the map, we outline the emerging markets and trends investors and strategy leaders should be watching.

Track 250+ companies building AI agent infrastructure

We selected companies for inclusion based on Mosaic health scores (500+) and funding recency (since 2023). Includes private companies only, organized according to their primary focus. Excludes general enterprise workflow automation platforms and non-pure-play LLM developers. This market map is not exhaustive of the space.

Please click to enlarge. 

Outlook & key takeaways

Private market momentum points to payments, voice, and security as key markets to watch

The AI agent tech stack is a high-momentum landscape, based on CB Insights Mosaic startup health scores. Private companies across the markets outlined below have an average Mosaic score of 768 — more than double the average of 370 for all private companies. They also have an average Commercial Maturity Score of 3, indicating widespread solution deployment.

A deeper dive into these scores, partnerships, and funding reveals 3 emerging markets to watch:

• Voice AI is the new battleground for the next wave of AI agents: With an average Mosaic score of 756 and nearly $400M in funding in 2025 so far, voice AI development platforms are building momentum. Big tech also recognizes voice as an essential AI building block — Meta’s first acquisitions since 2022 this year were PlayAI and WaveForms AI, both operating in audio and voice AI. 
• AI agent security startups see rapid momentum growth: AI agents create new attack surfaces and data breach risks, driving urgency for agent security startups. Companies in the market averaged a 56-point Mosaic score growth over 12 months, with Zenity, WitnessAI, and TrojAI each gaining 100+ points. The companies with the highest jumps in Mosaic score are partnering with larger tech firms and cybersecurity leaders. Public and established companies have also entered the conversation, with identity leader Okta and cybersecurity giant Palo Alto Networks both building agent security into their platforms.  
• AI agent payments startups get backing from incumbents: Agent payments infrastructure is one of the more nascent markets in this tech stack, with an average Commercial Maturity of 2.4 (validating) and an average Mosaic score of 697. The barrier to entry in payments is high, requiring complex technical and regulatory infrastructure. In an indication of the tech’s potential, established card and payment networks are investing and partnering with startups in the market: Coinbase backed Skyfire and Catena, Visa invested in Payman, and American Express participated in Nekuda’s recent seed round. Others like Crossmint and pre-funding PayOS have partnered with Visa and Mastercard.

Major LLM providers and tech incumbents all try to own a piece of the open standards pie

The growth of AI agents and development platforms has created a need to facilitate communication between agents and access to context. LLM developers and major tech companies are competing to own these standards. 

In less than a year:

• Anthropic launched Model Context Protocol (MCP), standardizing how AI agents connect to external tools and data sources 
• Google created the Agent-to-Agent (A2A) Protocol that allows agents to collaborate with each other, regardless of underlying framework 
• IBM introduced Agent Communication Protocol, which enables inter-agent communication across technologies and systems within a local environment 

These protocols have quickly become table stakes across the AI agent value chain. Professional services firms like Accenture, McKinsey, Deloitte, and KPMG contributed to Google’s A2A, and big tech companies like Microsoft and AWS support MCP. Meanwhile, startups in the tool libraries & integrations platform market like Speakeasy and Stainless are helping companies build MCP-compatible interfaces for their APIs (known as MCP servers), enabling AI agents to interact with their services.

MCP for the win: Make your AI smarter with our data and tools

Any MCP-compatible AI agent can tap into CB Insights’ datasets and tools – including ChatCBI – without a single line of code. Install our server into your environment to get started. Learn more here.

Big tech pushes deeper into AI agent development

While the above market map highlights the private landscape, tech giants and incumbents are also active across the AI agent infrastructure landscape. The top 3 global cloud providers — Amazon, Microsoft, and Google — are expanding their AI agent offerings across development tooling, hosting, orchestration, and more. 

Dive into the full report on how cloud leaders are shaping AI’s next frontier here. 

With many enterprises favoring established vendors, big tech companies have significant advantages in AI agent development. Similarly, enterprise software incumbents like Salesforce (Agentforce) and ServiceNow (AI Agent Marketplace) have launched agent platforms and marketplaces targeting their installed bases. 

Yet startups across the stack are carving out defensible positions by solving specific technical challenges and pushing the boundaries of what agents can do across areas like multi-agent orchestration (CrewAI) and enterprise data preparation (LlamaIndex). In the crowded AI agent development market, end-to-end platforms like WRITER and Dust are differentiating with vertical-specific implementations and promising speedy deployments. 

Autonomous agents drive the need for an oversight layer

AI agent reliability remains a major challenge in the landscape. Agents that fail, hallucinate, or behave unpredictably create immediate business risk. 

This is driving activity across observability, evaluation, and governance applications. The market has already seen 2 acquisitions in 2025 YTD. Early-stage activity highlights emerging technical needs, such as voice agent testing, with both Cekura ($2.4M seed) and Coval ($3.3M seed) focusing on evaluating and monitoring voice AI agents via simulated conversations. 

Securing agents is a growing priority across the stack. Based on one-year funding activity, the AI agent security & risk management market is the fastest-growing cybersecurity segment we track as agents proliferate across enterprise environments. 

White space opportunities for the AI agent ecosystem

As the AI agent tech stack matures, we predict the following areas will attract increasing innovation based on early-stage activity and recent product launches: 

• AI agent marketplaces: Distribution is a competitive advantage, with all major cloud providers launching dedicated AI agent marketplaces, including AWS in July 2025. Companies like Olas and Agent.ai are looking to differentiate through specialized agent discovery and customization. 
• AI agent monetization: Monetization emerges as an untapped opportunity, with companies like Paid giving visibility into AI agent costs and profit opportunities, and AGI Open Network tokenizes AI agents as tradable assets on blockchain networks.
• Cost management: At the end of the AI agent value chain, cost monitoring & productivity measurement will become more important as agents operate autonomously. For example, a16z-backed Larridin aims to give organizations visibility into AI spend and tool effectiveness. Other companies like coding AI agent Cline are building cost control solutions directly into their platforms to manage AI inference expenses. 

Source: CB Insights Deal Agent

Category overview

Click into each market to view the full description and market players on the CB Insights platform. 

Foundation models & infrastructure

Large language models (LLMs) form the cognitive core of AI agents. This layer also covers the compute, hosting, and inference systems required to serve models at scale. 

• LLM developers
• Model deployment & serving 

Agent frameworks & development platforms

Companies in this layer provide the software frameworks, SDKs, and low-code environments used to design, build, and deploy AI agents across different modalities and use cases.

• AI agent development platforms (including developer frameworks & platforms and low-code platforms)
• Voice
• Payments infrastructure  

Tool integration

AI agents leverage “tools” to interact with external systems and perform real-world actions, such as browsing the web. This includes Model Context Protocol (MCP) implementations that standardize how agents connect to data sources and tools.

• Browser infrastructure 
• AI web search APIs & scrapers 
• Tool libraries & integration platforms

Context 

This layer supplies agents with structured data, embeddings, and memory systems so they can retain, retrieve, and apply relevant information over time.

• LLM data preparation platforms
• Vector databases 
• Memory management

Orchestration 

This is the coordination layer that manages complex workflows involving multiple AI agents or models. 

• Multi-agent systems & orchestration 
• LLM benchmarking & model routing

Oversight

Companies here target authentication, security, monitoring, and governance functions that ensure agent actions remain safe, compliant, and aligned with intended outcomes.

• Authentication platforms
• Security & risk management platforms 
• Observability, evaluation, & governance 

The post The AI agent tech stack appeared first on CB Insights Research.

Combining CB Insights’ proprietary data and AI, scouting reports provide insight into each company’s:

Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post Book of Scouting Reports: Enterprise AI Agents appeared first on CB Insights Research.

Half of US federal agencies already report high levels of AI adoption, with these systems projected to handle most routine government functions within the next decade. Similar adoption patterns are emerging across municipal governments and international government bodies, particularly in Europe and the Asia-Pacific region.

Generative AI has already transformed procurement and fleet management through automated contract analysis and vehicle optimization, with major partnerships formed between government agencies and providers like Microsoft, Palantir, and specialized govtech firms.

Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post 310+ AI companies transforming government appeared first on CB Insights Research.

AI companies drove the surge, accounting for half of all mega-rounds. Many are building foundation models tailored to complex real-world use cases like robotics and healthcare.

Using CB Insights’ Business Graph, our monthly Book of Scouting Reports offers an in-depth analysis of every private tech company that has raised a funding round of $100M or more, to spotlight where capital is concentrating, which startups are gaining momentum, and who’s shaping the next wave of market disruption.

Download the book to see all 50 scouting reports.

Key takeaways from July’s mega-rounds include: 

• Clinical AI moves from development to scaling, with both Aidoc (a clinical AI foundation model developer) and Ambience (an AI medical scribe) having raised mega-rounds last month to build upon their early success and scale across more health systems. Last month also saw OpenEvidence and Tala Health raise $100M+ rounds to bring agentic AI solutions to clinicians, with the latter joining the fast-growing AI unicorn list. 
• Investors keep betting big on the next wave of the AI boom, physical AI. Recent commercial breakthroughs in the autonomous vehicle space and heightened interest in the humanoid space are driving capital toward physical AI infrastructure. This includes robotics foundation models (Genesis AI, TARS), and hardware platforms for embodied AI model training (Galaxea AI). China-based Meituan led both the $100M Series A extension in Galaxea AI and the $125M Seed round in TARS, as it doubles down on physical AI investments.
• AI newcomers are openly taking on tech giants. Half of last month’s mega-rounds went to AI companies, which accounted for 7 of the 13 new unicorns minted during that time. Some of these companies are directly targeting incumbents such as Reka AI which positions itself as a lower-cost alternative to OpenAI or Anthropic, and Perplexity which targets Google‘s core search business with its new browser product. 
• Fintech is minting a new class of financial services challengers.  Fintech companies accounted for more mega-round deals than any other vertical in July, including 2 of the top 4 largest rounds. Ramp’s valuation jumped from $16B to $22.5B in mere weeks, while Bilt more than tripled in value, from $3.3B to $10.8B. Beyond fundraising, fintech leaders are pursuing aggressive expansion strategies. iCapital raised $820M last month to accelerate its acquisition strategy focused on seizing the private markets opportunity. 

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post No summer break for AI: July 2025 hits 50 mega-rounds and 7 new unicorns appeared first on CB Insights Research.

Market leaders are transitioning from AI pilots to production systems, leveraging robots that autonomously navigate complex environments and digital twins that optimize all aspects of the industrial value chain.

Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post Industrial automation AI readiness: How Siemens, Hitachi, and ABB are building the industrial AI foundation appeared first on CB Insights Research.

Investors, merchants, and brands can seize this opportunity now, targeting the early movers for investment or partnership ahead of agentic shopping’s arrival.

We’ve been tracking these emerging solutions on our agentic commerce Watchlist. Within this list, we’ve identified 3 breakout markets, each accelerating a different piece of the agent-led shopping journey.

Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post 3 markets fueling the shift to agentic commerce appeared first on CB Insights Research.

Many of the world’s most influential financial services firms — like Allianz, J.P. Morgan, and Mastercard — have taken concrete action to adopt genAI technology.

The genAI adoption efforts have shaped 2 years’ worth of corporate strategy, unveiling key priorities — from the rise of agentic commerce to customer service copilots — across the competitive landscape.

Using CB Insights data, we identified and analyzed 100 real-world applications of genAI from 69 companies across banking, insurance, and payments.

Download the book to explore all 100 applications, and read on for 5 key takeaways and a breakdown of our methodology.

Key takeaways

1. Cross-functional platforms are now table stakes.

24% of applications center on deploying general-use genAI platforms to employees.

Prominent firms like BBVA have established enterprise-wide genAI capabilities across their organizations (typically via enterprise-wide deployments of platforms like Microsoft Copilot or ChatGPT). Early adopters — like Klarna, which shared in May 2024 that 87% of its employees are using OpenAI technology — now have over a year of genAI operational experience at scale, which can guide the development of more complex applications in the future.

Looking forward, financial services firms without a plan to provide genAI access to employees risk competitive disadvantage. Over the past 2 years, simply providing genAI capabilities to employees has shifted from cutting-edge innovation to standard operations.

2. Microsoft and OpenAI permeate the adoption landscape.

33% of applications analyzed disclose involvement from either Microsoft or OpenAI.

Microsoft and OpenAI (in which Microsoft has significantly invested) overwhelmingly permeate the landscape of genAI applications analyzed. Many of these applications anchor on foundational capabilities, from which organizations can build more complex applications and agents. Anthropic, Amazon Web Services, and Google Cloud follow a similar deployment pattern across multiple companies in the sector.

Looking forward, financial services firms should prepare for increasingly blurred “build, buy, or partner” decisions. The prevalence of genAI model developers (like OpenAI and Anthropic) and big tech partners (like Microsoft and Google) provide financial services executives with more flexibility to customize their tech solutions than what has traditionally been the case with many point-solution providers.

3. Emerging genAI vendors face a fierce competitive landscape.

Median Mosaic Scores among genAI startups analyzed are in the top 3% globally.

The 100 analyzed genAI applications include engagement from 25 startups as tech vendors, ranging from pre-seed companies like Twin — which offers an agent for invoice collection — to late-stage giants like Anthropic. These startups have a median CB Insights Mosaic Score — which measures the overall health and growth potential of private companies — of 732 out of 1,000, as of July 30, 2025.

Looking forward, financial services firms should prepare for increasingly capable tech vendors seeking to sell their genAI products. These vendors must exhibit a clear advantage over the alternative of building in-house solutions.

4. Customer-facing genAI will become increasingly prevalent.

16% of applications center on customer engagement & self-service capabilities.

Firms like ING, Wells Fargo, and Truist show that customer-facing genAI assistants are capable of powering millions of customer interactions. Customer-facing genAI deployment will accelerate as companies like Mastercard, Visa, and PayPal deploy applications centered on “agentic commerce,” where customers can autonomously shop and complete transactions with AI payments agents.

Looking forward, financial services firms need to develop a gameplan for how they will engage customers with agentic AI. The market opportunities for enterprise agents and copilots are growing, so customer-facing applications will quickly emerge.

5. Impact is now tangible, but success definitions remain elusive.

Only 30% of applications disclose quantitative tangible impact from deployment.

Most of the application sources analyzed lack disclosure of tangible impact (i.e., numbers, percentages, or figures to quantify effectiveness). Among the impact metrics that are available, the top-cited focus on operational considerations like call-handle times.

Looking forward, any financial services firm has the opportunity to define “what good genAI adoption looks like” across the sector. The lack of clear success definitions creates an opportunity for financial services firms to stand out among peers.

Methodology

We used CB Insights’ Business Graph — including data points like Dealmaking, Business Relationships, Earnings Transcripts, and Media Mentions — and third-party company releases to identify 100 real-world genAI applications across banking, insurance, and payments. These applications were disclosed between July 2023 and April 2025.

Then, using CB Insights’ Team of Agents, we analyzed these applications across 10 categories. Applications are detailed based on disclosure date, and are not exhaustive of a given company’s genAI initiatives. Applications and categorizations are not mutually exclusive or exhaustive of activity within their respective industries.

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post 100 real-world applications of genAI across financial services and insurance appeared first on CB Insights Research.

Based on CB Insights Commercial Maturity data, 42% of these companies are already deploying or commercializing their solutions (deploying, scaling, established), with a few leaders already crossing $100M in ARR. This includes Anysphere’s Cursor ($500M in ARR) as well as Windsurf and Moveworks, both of which reached $100M ARR shortly before being acquired.

This commercial traction signals the rapid adoption of specific types of AI agents by enterprises and who the early market winners are. The companies generating the most revenue often target workflow-heavy sectors where AI delivers immediate ROI — primarily coding and enterprise workflows. 

We expect these categories to continue driving adoption (and revenue), and predict the enterprise AI agents & copilots space will generate close to $13B in annual revenue by the end of 2025, up from $5B in 2024. 

Using CB Insights revenue data, we identified the top 20 private startups offering AI agents as their primary offering and analyzed how they are rewriting the VC funding playbook (see below graphic).

If you are an AI agent startup and want to submit your company’s revenue data, please reach out to analystbriefing@cbinsights.com. 

See the top AI agent startups ranked by revenue here

Key takeaways

• Top revenue-generating AI agent startups are just under 5 years old on average, with 50% of them having been founded in the last 3 years. This signals how quickly these AI-native companies are scaling and monetizing their products with recent breakouts including Cursor ($500M revenue, founded 2022), Mercor ($100M, founded 2023), and Lovable ($100M, founded 2023).
• Customer service AI agents command the highest valuation premiums, with an average revenue multiple of 127x compared to 52x on average across all top 20 AI agents by revenue. This valuation gap signals that investors are betting on aggressive revenue acceleration in customer service AI, driven by the sector’s universal market applicability and the expectation that businesses will rapidly replace human support teams with AI agents. 
• Some AI agent startups are already as capital-efficient as big tech companies. Mercor ($4.5M revenue per employee) and Cursor ($3.2M per employee) already surpass the likes of Microsoft ($1.8M per employee, FY 2024) and Meta ($2.2M per employee, FY 2024), and rivaling Nvidia‘s efficiency levels ($3.6M per employee, FY 2025). 

However, as new entrants enter the AI agent market at a record pace — both startups and tech giants pivoting into AI agents — the question becomes whether these early revenue wins can translate into defensible market positions. 

We expect competitive moats to emerge through proprietary data advantages, deep vertical specialization, and the creation of switching costs through deep integration into customers’ critical business workflows. 

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post AI agent startups are becoming revenue machines — here are the top 20 ranked appeared first on CB Insights Research.

Investors and big tech alike are betting that voice will be the dominant interface for interacting with AI, enabling a move away from traditional browser and mobile interfaces toward natural conversational interaction. 

Recent technological advancements have made this vision increasingly viable, with voice capabilities now delivering near-instantaneous responses with sub-300ms latency that matches human conversational flow. This speed breakthrough is critical to unlocking voice AI’s full potential, as Chris McCann at Race Capital, a backer of PlayAI, explains:

“Voice is how people naturally communicate – but most voice AI systems still sound robotic or have high latency in their responses. We believed fast, expressive voice tech would be critical to making AI feel human and useful in the enterprise, especially for IVR, customer support, and sales.”

With voice becoming an increasingly fundamental modality for the AI-powered future and big tech competing to win the AI device race, owning the building blocks that shape human-AI communication is becoming mission-critical. Expect a wave of acquisitions as companies scramble to secure voice AI capabilities.

Using CB Insights’ Mosaic score which measures company health, we identified the top M&A targets in the voice AI space and what makes them such compelling targets (see below graphic).

See the full list of voice AI startups ranked by Mosaic score

• Voice synthesis platform ElevenLabs tops the market with a Mosaic score of 955, making it an attractive acquisition target. Proprietary voice generation technology is becoming as valuable as foundational AI models, positioning the highest-quality voice synthesis as core infrastructure rather than a feature add-on.
• Enterprise-focused Cresta delivers immediate ROI, with some customers reporting 50% cost reductions in contact centers, and positioning it perfectly for companies looking to leverage voice AI to immediately impact enterprise productivity.
• Ultra-low latency startups like Cartesia have an edge, as their ability to deliver sub-100ms capabilities positions them as essential for truly conversational AI experiences that matches human conversation patterns. 

Investors also see companies owning the full-stack as a having key technological advantage compared to those relying on third-party components. This was part of the rationale for investing into PlayAI according to Chris McCann of Race Capital:

“Most voice AI startups rely on open source or other third-party components. PlayAI built the full stack in-house—their own TTS engine, real-time streaming, and sub-100ms latency. That gave them full control and a clear technical edge, which let them power real-time agents for support, sales, and IVR across several Fortune 500s.”

As the AI arms race continues, acquisitions will continue to be focused on talent, tech, and infrastructure rather than existing revenues. Companies that secure advanced voice AI capabilities now will dominate the next phase of AI adoption – whether they integrate into their existing offerings or cash-in on selling the tooling back to others.

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post Voice AI is having a moment: Here are the startups that could get acquired next appeared first on CB Insights Research.

AI is already fueling revenue growth for these cloud giants. 

First, AI workloads require more computing resources than traditional workloads, thus increasing per-customer spending. Second, AI companies (with their own computing needs) are proliferating rapidly, now capturing 20% of all venture deals globally. Together, these trends create both enormous revenue opportunities and unprecedented infrastructure challenges.

At the same time, new competitors are emerging to serve this insatiable demand. The OpenAI-led Stargate Project, with its planned $500B investment, threatens to reshuffle the cards in the cloud computing space that AWS has led for over a decade in terms of market share.

In response, cloud providers are spending tens of billions to capture their share of AI computing spend.

In the 19-page report, we cover 3 strategic pillars that emerged from our analysis:

• Cloud providers are investing heavily in compute infrastructure to meet explosive AI demand. Amazon, Alphabet, and Microsoft are planning a combined $250B+ in capex spend, primarily for AI data centers, in 2025 in addition to vertically integrating into energy production with 6 nuclear partnerships and creating custom AI chips to control costs and gain competitive advantages.
• Strategic partnerships and ecosystem development are key to cloud dominance, as providers lock in strategic partnerships with leading model developers (such as Microsoft’s $13B investment in OpenAI), develop proprietary foundation models, and build out accelerator programs to seed AI ecosystems. For example, Amazon expanded its genAI-focused accelerator from 21 to 80 startups between 2023 and 2024 while more than tripling the value of the cloud credits offered.
• Cloud providers are expanding their AI service portfolios into agentic AI and security to drive adoption and consumption. Alphabet recently made its largest acquisition ever to expand into the cloud security space, spending $32B to buy Wiz, while all 3 players are racing to expand their agentic AI offerings, including developer tools, dedicated marketplaces, and customizable agents.

Additional resources:

• Building the agent economy: How cloud leaders are shaping AI’s next frontier
• The AI data center value chain: 12 high-momentum technologies powering the future of AI
• AI is making big tech even bigger — here’s how the trillion-dollar tech giants are deepening their moat and fueling future growth
• Data centers are reshaping nuclear development, driving billions in new power infrastructure investment

The post Battle of the Cloud Titans: How Amazon, Microsoft, and Alphabet are preparing for an AI future appeared first on CB Insights Research.

Given this momentum, the ecosystem is larger and more challenging to navigate than ever.

Our annual AI 100 list is designed to cut through this noise and highlight the next wave of AI winners, with a focus on early-stage players that are showing strength in terms of market traction, investor quality, and talent.

Leveraging CB Insights datasets such as deal activity, industry partnerships, team strength, investor strength, patent activity, and our proprietary Mosaic Scores, we selected 100 winners out of a cohort of 17K+ companies. We also analyzed CB Insights’ exclusive interviews with software buyers and dug into Analyst Briefings submitted directly to us by startups.

Below, we map out the winners, categorizing them based on their core offering. Key trends and category definitions follow. Customers can track activity of all of these companies in this watchlist. 

Please click to enlarge. Data as of 4/23/25.

CBI customers: Get data on the winners

Key takeaways on the AI 100

1. AI agents dominate the conversation. These applications, which automate tasks and processes for human users, are the next wave of genAI. Having made their way into virtually every horizontal and enterprise function, AI agents are also coming for infrastructure and verticalized applications. AI agents and supporting infrastructure make up 21% of this year’s companies, and the investors we spoke with consistently cited this space as a priority. 
2. ML security has become table stakes. The need to secure AI applications has grown in lockstep with the proliferation of genAI and agentic AI. 46% of strategy team leaders point to security as the primary barrier to genAI adoption, according to a recent CB Insights survey. Machine learning security companies are hardening AI algorithms and foundational models like LLMs, while also defending against increasingly sophisticated AI-powered attacks. 
3. AI observability and governance are critical gaps. Widespread use of AI is exposing the technology’s cracks — hallucinations, lack of orchestration, and output inaccuracies. It’s clear that AI ubiquity can’t exist without robust monitoring. Companies are rising to meet this need. Startups in this year’s list cover areas like observability and governance, while a small cohort also monitors AI agents to ensure reliability and compliance.   
4. The future is physical. Looking ahead, AI will evolve beyond software AI agents to a physical state. Advances in disparate areas of AI development — including robotics, multimodal image and voice models, edge computing, synthetic data, and spatial intelligence — provide the scaffolding for physical AI, which pairs AI software with hardware to take action in physical environments. Industrial humanoids represent an early manifestation of this, while future permutations could include fully autonomous defense drones, home companion robots, and more.
5. Vertical applications are exploding. In 2024, the horizontal companies in this AI 100 cohort received more funding than their vertical and infrastructural counterparts — $1.6B compared to $1.2B each for infrastructure and vertical. But in 2025 so far, the funding picture looks very different: Vertical winners lead the way with $1.1B in funding raised.

Category breakdown

AI INFRASTRUCTURE 

On the foundation model front, infrastructure newcomers are rapidly releasing models that rival industry leaders, signaling a maturing market where technical excellence and novel approaches increasingly compete with raw computing power. We identified winners across large language, edge, reasoning, small language, and multimodal models. 

Meanwhile, as AI applications — particularly agents — become more autonomous and widespread, the need for robust monitoring, governance, and cybersecurity solutions has grown in lockstep. 

We’ve heard this in our conversations with AI investors, as well. Mozilla Ventures, a lead investor in Credo AI, views governance as a strategic imperative. Mohamed Nanabhay, Managing Partner, notes: 

 “…We think that the AI governance sector itself will take on a crucial role of creating value for enterprises, allowing companies that leverage governance to deploy AI faster through the reduction of risk with a greater competitive advantage as a result.”

Category definitions:

DATA

• Synthetic data: Artificially generated or altered information that mimics real-world data without privacy concerns. Aaru uses a multi-agent approach to create population simulations for predictive decision-making applications like consumer behavior and electoral modeling.  
• Data preparation & curation: Tools and platforms that clean, transform, label, and organize data to make it suitable for AI training and deployment, encompassing data cleaning and specialized data processing. Unstructured, for instance, helps organizations capture unstructured data from various documents and convert it into AI-friendly formats such as JSON to train LLMs.
• Vector databases: Solutions that provide enterprises with an easy way to store, search, and index unstructured data at a speed, scale, and efficiency that current relational (and non-relational) databases cannot offer. For example, Qdrant provides an open-source vector database that allows developers to build production-ready applications that use nearest neighbor search functionality.

DEVELOPMENT & TRAINING

• Foundation models: Pre-built AI algorithms and architectures that can be deployed, fine-tuned, or integrated into applications, spanning general-purpose foundation models and specialized domain-specific models. This category includes large language, edge, reasoning, small language, and multimodal AI models. For instance, Archetype AI‘s Newton model processes multimodal sensor data and natural language to provide insights and predictions about physical environments.
• Agent building & orchestration: This category covers AI agent development platforms for building, orchestrating, and monitoring agents. Companies like LangChain provide a framework for building context-aware reasoning applications with tools for debugging, testing, and monitoring app performance across the entire application lifecycle.
• Computer vision & spatial intelligence: Technology that enables AI systems to understand, interpret, and interact with physical spaces and 3D environments, including mapping, navigation, and spatial data processing capabilities. Notably, World Labs develops Large World Models (LWMs) that enable AI systems to perceive, generate, and interact with both virtual and real 3D environments using spatial intelligence.

OBSERVABILITY & EVALUATION

• AI observability platforms: These platforms monitor, measure, and assess AI model performance, reliability, and outputs, including tools for testing, benchmarking, and continuous improvement of AI systems. For instance, Arize’s platform allows teams to monitor, diagnose, and improve the performance of AI models and applications in production through tools based on open-source standards that integrate with existing AI infrastructure.
• Governance: Solutions that establish policies, processes, and controls for responsible AI development and deployment, covering risk management, compliance, ethical oversight, and transparency requirements. For example, Credo AI offers a platform that automates AI oversight, risk management, and regulatory compliance while providing AI auditing to ensure system integrity and fairness.
• Machine learning security (MLSec): Technologies that protect AI systems from vulnerabilities, attacks, and data breaches, including techniques for securing model training, inference, and data pipelines. Solutions developed by companies like Zama enable computation on encrypted data, allowing for privacy-preserving machine learning across industries that require data privacy and security.

ACCELERATED COMPUTING & HARDWARE

• Edge: Platforms that provide the infrastructure and models to operate AI on “edge” devices such as tablets, IoT, autonomous vehicles, or smartphones. For example, EdgeRunner AI constructs an ensemble of small, task-specific models that work together to solve complex problems locally on devices, ensuring data privacy and security for heavily regulated industries.
• Photonics: Solutions that use light (photons) instead of electrons for data processing, with the potential to significantly increase computing speeds. Companies in this category provide memory, interconnects, and system architecture. Xscape Photonics develops bandwidth-efficient photonics solutions to support AI/ML infrastructure. 
• Quantum: Companies providing novel techniques like model compression and hardware to support quantum commercialization. Multiverse Computing provides AI model compression technology to enable quantum AI workloads and processing.
• Chips: Traditional chips, in addition to chips to support new AI technologies. Etched develops chips designed specifically for transformer inference, capable of processing extensive data for applications such as real-time voice agents and content generation.

HORIZONTAL AI

This category includes industry-agnostic solutions across visual media, text, code, audio, and interfaces. These function-specific solutions address common business needs regardless of industry, offering specialized intelligence that complements both vertical applications and foundational infrastructure.

AI agents in particular are beginning to upend the way in which enterprises think about software. Decibel Partners, a lead investor in multi-agent platform Dropzone AI, sees a movement toward productizing agents as full systems. Jéssica Leão, a Partner at Decibel, articulates this vision further:

“…We’re going to see the software world change because, again, you’re selling agents almost as if you’re selling back-end software.”

Horizontal AI solutions are increasingly tailored to serve distinct business functions while remaining broadly deployable. Startups in this category are developing sophisticated AI systems that excel in capabilities like content generation, customer support, process automation, and software development — all of which can be applied across industries. 

Category definitions:

• Content generation: AI systems that create text, images, video, and other media forms — spanning automated content production and multimodal generation. For example, Moonvalley’s genAI video model helps filmmakers by enabling prompt adherence, motion generation, and physics simulation using cleaned, fully licensed data.
• Customer service: AI agents that autonomously handle customer service tasks or augment human agents. Sierra‘s platform, for instance, provides intelligent agents for customer support that engage in personalized interactions and integrate with existing call center technologies.
• Cybersecurity: AI-powered solutions that detect, prevent, and respond to digital threats, vulnerabilities, and attacks, covering network security, threat intelligence, and automated incident response. Companies like Binarly use AI to detect and remediate vulnerabilities in firmware and software supply chains.
• General-purpose humanoids: AI systems embedded in robotic bodies that mimic human capabilities, enabling physical interaction through perception and manipulation. For example, Figure develops autonomous humanoid robots that combine human-like dexterity with AI to perform a variety of tasks across industries like manufacturing, logistics, warehousing, and retail.
• Process automation: Intelligent systems that autonomously handle repetitive business workflows, increasing efficiency by eliminating manual tasks. Orby AI offers a platform that observes enterprise processes and generates executable automations — particularly for complex, data-heavy operations in industries like tech and finance. 
• Software development & coding: AI solutions that assist with software development, code generation, debugging, and programming tasks, including automated code completion tools. For instance, Poolside offers foundation models and APIs that can be fine-tuned using a company’s own codebase and documentation to support internal dev teams.
• Video security: Technologies that enable real-time analysis of video feeds, supporting faster detection and response to security threats. Coram AI develops cloud-based security camera systems with features like real-time AI alerts and natural language video search, allowing businesses to monitor properties remotely without extensive hardware replacements.

VERTICAL AI 

Vertical AI is on the rise, with this year’s vertical winners surpassing the other category winners to capture over $1B in combined funding in 2025 YTD. They span 10 industries that represent a convergence of high-value problems, rich data availability, and regulatory momentum.

Some of the VCs we spoke with see specialization as the way of the future. Lila Tretikov, Partner and Head of AI Strategy at New Enterprise Associates (a lead investor for Twelve Labs, World Labs, and Orby AI), told us:

“We believe that there is going to be specialization, even within the model layer. And there’s going to be innovation in this layer, especially as we look at verticalization for specific use cases.”

The most well-represented verticals on this year’s list are healthcare (8 companies) and life sciences (6 companies). The healthcare industry as a whole is seeing breakthrough applications across multiple AI modalities — from agentic AI systems that can augment clinical workflows, to advanced machine vision for medical imaging analysis, to AI-accelerated drug discovery platforms that dramatically reduce R&D timelines.

This year’s cohort also saw significant representation in gaming & virtual assets (5 companies), finance & insurance (4 winners), and aerospace & defense (4 winners). 

Category definitions:

• Aerospace & defense: AI solutions designed for aerospace engineering, aviation operations, military applications, and defense systems, including autonomous navigation and threat detection technologies. For instance, Quantum Systems creates eVTOL unmanned aerial systems that serve critical defense applications, most notably in Ukraine. 
• Auto & mobility: AI applications for autonomous vehicles, transportation optimization, fleet management, and mobility services. Companies like Wayve are developing AI systems that use LLMs to provide real-time natural language explanations of driving decisions, helping improve users’ confidence.
• Energy: Platforms that optimize energy production, distribution, and sustainability, including battery intelligence and AI assistance for electric grids. For example, Liminal leverages ultrasound and machine learning inspection solutions to improve battery cell quality, cost-effectiveness, and safety while enabling confident scaling of production. 
• Finance & insurance: AI solutions for financial services, banking, investment, and insurance sectors, covering payments, risk assessment, and portfolio monitoring. Skyfire’s financial stack enables AI agents to perform transactions without credit cards or bank accounts, allowing businesses to monetize their products, services, and data through AI agents.
• Gaming & virtual assets: AI technologies that enhance gaming experiences, virtual environments, digital asset management, and immersive entertainment, including content generation and NPC (non-player character) intelligence. Altera‘s platform creates digital human beings that can interact with users and perform tasks autonomously, bringing empathy and human-like traits to digital interactions.
• Healthcare: AI applications focused on clinical care delivery, medical operations, and patient management, including tools for clinical documentation automation, medical imaging analysis, decision support systems, remote patient monitoring, and healthcare supply chain optimization. In the dental field, Overjet provides an AI platform that enhances clinical care through radiographic analysis and optimizes claims processing for providers and payers.
• Life sciences: AI solutions for pharmaceutical research, drug discovery, protein engineering, biological data analysis, and therapeutic development, including platforms for multiomics analysis, antibody design, foundation models for biology, and scientific experiment automation. Lila Sciences has developed a platform that integrates AI with autonomous laboratories to design, conduct, observe, and redesign experiments for scientific discovery.
• Legal: AI tools for legal research, document analysis, contract management, compliance, and legal workflow automation, including case management, due diligence, and contract review. AI-powered tools like Eve help law firms streamline the full case lifecycle from intake to litigation by automating case intake, drafting legal documents, and managing discovery processes.
• Manufacturing: Technology that optimizes industrial processes like factory automation, using virtual development and simulation. PhysicsX applies machine learning to physics simulations that optimize design and engineering processes across industries including aerospace, medical devices, and electric vehicles.
• Supply chain: AI solutions that enhance logistics and supply chain operations, including warehouse management and route optimization & visibility. Dexory combines stock-scanning robots with a digital twin platform to provide real-time inventory and warehouse analytics for logistics and supply chain operations.

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post AI 100: The most promising artificial intelligence startups of 2025 appeared first on CB Insights Research.

While AI copilots have already made inroads across industries, the next evolution — autonomous agents with greater decision-making scope — is arriving quickly. AI agent startups raised $3.8B in 2024 (nearly tripling 2023’s total), and every big tech player is already developing AI agents or offering the tooling for them.

Implications for enterprises will be far-reaching, from altering workforce composition (with new hybrid teams of humans and AI agents) to maximizing operational efficiency through full automation of routine tasks. 

Below we identify 170+ promising startups developing AI agent infrastructure and applications. 

CBI customers: Track 250+ AI agent companies

We selected companies for inclusion based on Mosaic health scores (500+) and/or funding recency (since 2022). We included private companies only and organized them according to their primary focus. This market map is not exhaustive of the space.

Want to be considered for future AI agent research? Brief our analysts to ensure we have the most up-to-date data on your company. 

Outlook on AI agents

Fully autonomous agents remain limited due to issues pertaining to reliability, reasoning, and access. Most agent applications today operate with “guardrails” — within a constrained architecture where, for example, the LLM-based system follows a decision tree to complete tasks. 

Agents featured on this map include some combination of the following components: 

• Reasoning: Foundation models that enable complex reasoning, language understanding, and decision-making. These models evaluate information and form the cognitive core of the agent.
• Memory: Systems that store, organize, and retrieve both short-term contextual information and long-term knowledge.
• Tool use: Integration capabilities that allow agents to interact with external applications, APIs, databases, the internet, and other software. 
• Planning: The agent’s architecture for breaking down complex tasks into more manageable steps, reflecting on performance, and adapting as necessary.  

We expect more startups to move up the scale of autonomy as AI capabilities advance. Improvements in reasoning and memory will enable more sophisticated decision-making, adaptability, and task execution.

For example, in September 2024, legal AI startup Harvey announced that OpenAI’s o1 reasoning model, supplemented with domain-specific knowledge and data, was enabling it to build legal agents. The company, which raised $300M at a $3B valuation in February 2025, has doubled its sales force in the last 6 months, indicating rising market demand.

While the above market map highlights the private landscape (with a focus on enterprise applications), tech giants and incumbents are also launching agents. We predict big tech and leading LLM developers will own general-purpose AI agents, but there are many opportunities for smaller, specialized players. 

Looking ahead, watch for new form factors outside of the copilot/chatbot interface that will push the boundaries of what an “agent” is. Early indications of this include “AI-native” workspaces — tools and platforms built from the ground up around AI capabilities, rather than layering AI features on top of a traditional product. For instance:

• Eve’s legal platform aims to automate aspects of the whole case lifecycle (from case intake to drafting). 
• Hebbia’s Matrix product builds spreadsheets that mine information from files (in rows) and deliver answers to questions (in columns), proactively discovering, organizing, and surfacing data.
• With its Dia product, The Browser Company is exploring web browsing interfaces that can summarize content, automate repetitive web tasks, and even anticipate next actions.

Category overview

AI agent infrastructure

This segment covers companies building agent-specific infrastructure. (We excluded general genAI infrastructure markets like foundation models and vector databases from the map.)

Development tools

A diverse ecosystem of tools has emerged to support agents’ development. These range from memory frameworks like Letta that enable persistent, retrievable memory across interactions; to tools that allow agents to take action via integration (e.g., Composio), authentication (e.g., Anon), and browser automation (e.g., Browserbase).

Another set of companies is giving agents more utility across payments (which includes companies developing crypto wallets for agents as well as virtual cards) and voice (development platforms and tools for testing AI voice applications as well as speech models).

Meanwhile, demand for simplified, comprehensive deployment options is driving the rise of AI agent development platforms — the most crowded infrastructure market on our map. 

LLM developers including Cohere (with its North AI workspace) and Mistral have launched their own agent development frameworks, while Amazon, Microsoft, Google, and Nvidia all offer AI agent development tooling. With many enterprises favoring established vendors due to lower risk, big tech companies have significant advantages here.

Trust & performance

Concerns around reliability and security have helped establish a market for agent evaluation & observability tools. Early-stage companies are targeting applications such as automated testing (e.g., Haize Labs) and performance tracking (e.g., Langfuse). 

Multi-agent systems, where specialized sub-agents work together to complete tasks, also show promise in improving accuracy. Insight Partners-backed CrewAI’s multi-agent orchestration platform is reportedly already used by 40% of the Fortune 500. 

Vendors are also tackling reliability concerns directly. Based on our briefings with 20+ AI agent startups in Q1’25, companies are using 5 primary methods to build user trust: 

1. Transparency
2. Human oversight
3. Technical safeguards
4. Security & compliance
5. Continuous improvement 

Horizontal applications & job functions

Horizontal AI agent startups make up nearly half of the map and overall landscape. 

This segment primarily features startups targeting enterprises, with industry-agnostic applications across job functions like HR/recruiting, marketing, and security operations. Companies in the productivity & personal assistants market, including OpenAI with its Operator agent, are targeting consumers and employees directly.  

The AI agent markets with the most traction — based on companies’ median Mosaic health scores — are customer service and software development (which includes coding and code review & testing agents). These markets are also among the most crowded due to the value agents bring to well-defined workflows and testable environments. 

We see this reflected in adoption, particularly at the customer service layer: Among 64 organizations surveyed by CB Insights in December 2024, two-thirds indicated they are using or will be using AI agents in customer support in the next 12 months. 

Overall, horizontal AI agent applications are more commercially mature compared to the infrastructure and vertical segments, with over two-thirds of the market deploying or scaling their solutions based on CBI Commercial Maturity scores. 

Vertical (industry-specific) applications

We expect increasing verticalization as startups carve out niches by solving industry-specific customer problems, especially in areas with strict regulatory scrutiny and data sensitivity.

This category features companies catering to industries including: 

• Financial services & insurance: The most crowded vertical category on the map with 11 companies, startups here are targeting a variety of finserv workflows such as financial research (Boosted.ai and Wokelo), insurance sales & support (Alltius and Indemn), and wealth advisory prospecting & operations (Finny AI and Powder). 
• Healthcare: Solutions in this market aim to reduce the volume of manual tasks for healthcare professionals across use cases like clinical documentation, revenue cycle operations, call centers, and virtual triage. Solutions from companies like Thoughtful AI (revenue cycle operations) and Hippocratic AI (staffing marketplace) are targeting end-to-end healthcare workflows. 
• Industrials: These companies look to optimize processes and equipment — including control systems, robots, and other industrial machines — without relying on consistent human intervention. For example, Composabl launched an agent platform in May 2024 that uses LLMs to create skills and goals for agents that can control industrial equipment. Public companies like Palantir are also active in this space. Learn more in our industrial AI agents & copilots market map. 

RELATED RESEARCH

• What’s next for AI agents? 4 trends to watch in 2025
• The enterprise AI agents & copilots market map
• The Multi-Agent AI Outlook
• Future of the workforce: How AI agents will transform enterprise workflows

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post The AI agent market map appeared first on CB Insights Research.

These LLM-based systems mark an evolution beyond copilots: AI agents can accomplish complex tasks on a user’s behalf with minimal intervention, from sales prospecting to compliance decisioning. 

In the rapidly growing landscape for agent infrastructure and applications, over half of companies in the market have been founded since 2023. Meanwhile, funding to startups in the space nearly 3x’d in 2024.

Want to see more research? Start your free trial.

If you’re already a customer, log in here.

The post What’s next for AI agents? 4 trends to watch in 2025 appeared first on CB Insights Research.

Imagine: your personal AI agent notifies you that a hair dryer you’ve been eyeing is now on sale. The product page highlights benefits tailored to your curly hair, while the agent confirms it will arrive before your upcoming trip.

With your approval, the agent handles the purchase through your secure wallet. Later, it proactively suggests complementary hair care products for the summer season.

This world of autonomous commerce isn’t as far off as it seems. Tech and e-commerce leaders — including OpenAI, Nvidia, Amazon, Walmart, Google, and Apple — are already building AI systems that are steps away from conducting transactions. 

AI agents will impact each stage of the customer journey, streamlining the path to purchase and fundamentally transforming how businesses build relationships with consumers and drive loyalty.

We use CB Insights data on early-stage fundraising, public companies, and industry partnerships to analyze how generative AI — especially AI agents — is transforming the customer journey.

In the 11-page report, we cover 3 predictions that emerged from our analysis: 

1. First-party transaction data will shape the future of AI-driven personalization. As personalization becomes more sophisticated at the awareness and consideration stages, companies with direct access to first-party data will have an edge.
2. Direct-to-agent (D2A) commerce will kill traditional loyalty. With AI agents handling browsing and shopping, traditional loyalty programs will lose effectiveness as agents optimize shopping across a select group of merchants.
3. A few AI agents will own the customer relationship. Companies like Amazon, Google, and Apple — with critical distribution and financial services infrastructure — are well-positioned in commerce.

RELATED RESEARCH FROM CB INSIGHTS

• How generative AI could supercharge retailers’ ad networks
• AI is revitalizing the “virtual try-on” space — this is what’s coming next, including interactive Google ads
• The enterprise AI agents & copilots market map
• 15 tech trends to watch closely in 2025
• Generative AI — e-commerce search
• AI-optimized e-commerce UX
• Generative AI — customer support operations

For information on reprint rights or other inquiries, please contact reprints@cbinsights.com.

The post The future of the customer journey: AI agents take control of the buying process appeared first on CB Insights Research.

Prefer to listen in? Check out our discussion of the report here: 

An empowered digital workforce would reshape industries as we know them. The implications would be enormous, changing how companies hire and scale, as well as what they can achieve with a small headcount. 

That future isn’t too far off. 

The idea of autonomous AI agents — LLM-powered bots that can independently reason and execute tasks — caught on like wildfire in 2023, marking an important evolution beyond chatbots and copilots. 

OpenAI CEO Sam Altman has described agents as “AI’s killer function” as recently as May 2024.  

While much of the tech remains limited in its ability to execute tasks reliably, use cases are gaining traction in horizontal enterprise applications like customer support, sales, and engineering.

We mined CB Insights startup, financing, business model, and buyer interview data to map the evolving landscape and analyze its future. 

In the 28-page report, we cover: 

• The state of AI agents: Investment is surging to companies in the space, but limitations — most notably, agent reliability — remain. 
• Leading horizontal applications and impacts: The landscape of VC-backed agent startups is dominated by a focus on horizontal applications — across sales, customer support, and other enterprise and general productivity workflows.
• Emerging industry applications and opportunities: While few agentic companies focus on single industries, companies are emerging to target workflows across financial services, industrials, and more. 

Download the full report to get all of the data and analysis.

The post Future of the workforce: How AI agents will transform enterprise workflows appeared first on CB Insights Research.

Brands and retailers are increasingly investing in tech-enabled customer service solutions that can help support and convert customers — either online or in stores — in a timely and cost-efficient manner.

Using CB Insights data, we examined tech markets across customer service for consumer & retail leaders and ranked them across two metrics — market momentum and industry leader activity — to help companies decide whether to monitor, vet, or prioritize these technologies.

Want to see more research? Join a demo of the CB Insights platform.

If you’re already a customer, log in here.

The post Why consumer & retail leaders are prioritizing agent support tools appeared first on CB Insights Research.