In collaboration with the finance company Mercury, Andreessen Horowitz published its inaugural AI Spending Report on Thursday.  Like the previously released Top 100 Gen AI Consumer Apps, the research examines the top 50 AI-native application layer businesses that startups are investing in using transaction data from Mercury.

According to Olivia Moore and Seema Amble, partners at a16z, the data indicates that businesses continue to use a variety of AI solutions for certain activities, and new apps are emerging and disappearing at a rapid pace.

“A wide variety of tools are available,” Amble stated.  “In each category, it hasn’t just gathered around one or two.”

Additionally, the survey reveals significant expenditures on “human augmentors” or “copilots,” which can increase worker productivity, indicating that startups are not yet prepared to make the whole transition to agentic workflows.

“I believe that shift will occur, where we’ll see fewer copilots and more end-to-end agent tools, as computer use becomes more of a mode and there’s more potential for end-to-end agentic flows to be built,” Amble stated.  “Especially since so many people want to try them.”

Related: As Bangalore’s Office Costs Surge, Startups Opt for Virtual Offices

Major labs dominated the top of the ranking, as was to be expected, with OpenAI at the top and Anthropic at number two.  With Replit at number three and Lovable at number eighteen, vibe-coding tools were also well-represented.  Emergent came in at number 48, and Cursor at number 6.  At number 34 was Cognition, which runs more enterprise-focused coding tools including Devin and Windsurf.

Lovable outperformed Replit on pure traffic alone when a16z generated a comparable list for consumer habits because many people were utilizing it to develop projects.  However, the lack of corporate functionality in Lovable is one of the reasons why startups are investing less money on it than on Replit.  The list’s diversity, however, seems to indicate that there is space for several distinct businesses operating simultaneously.

“The future of vibe coding is up in the air,” Moore stated.  Does one location end up being the greatest place to vibe code when the spaces begin to converge?  Or is it possible that there will be an additional four or five large vibe coding companies for other applications?  The solution to it is yet unknown to us.

The use of consumer-focused solutions like CapCut and Midjourney by startups also caught Moore off guard.

Moore noted, “We’re seeing that a lot of these [consumer] companies are getting yanked into enterprise faster and faster because they make such engaging consumer tools that people adopt and use as individuals and bring into their teams and workplaces.”

At least 60% of the entries on the list were horizontal applications, with the remaining 40% being vertical apps.  Three categories comprised the most well-known vertical software companies: customer service, recruiting, and sales.  However, the survey also discovered that AI was advancing in a number of areas that had proven difficult for earlier startup generations to conquer.

“In the era of artificial intelligence, what might have been service firms or consultancies are now software companies,” Moore stated.

As an illustration, Amble cited Crosby Legal, which can swiftly go over a legal contract for a user, taking the place of what used to be a meeting with an internal general counsel who would have been searching through ideas and research.  Instead of replacing entire workforces and talent suites with automated workers (end-to-end agents), she noted, the majority of the technologies are now utilized to help employees (like a co-pilot) make decisions more quickly.

She said that AI tools can perform far more tasks, such as outreach, more quickly than humans, and that “you’ll see that mix shift more toward end-to-end agents and away from co-pilots as the tech gets better and we’re actually able to build out full agent co-workers.”

Additionally, there were numerous note-takers on the list, including HappyScribe, Read AI, and Otter.ai; no one choice predominated.  This is what Amble meant when she stated that startups are still choosing “their own flavor” to determine which tools they want, rather than a single solution that has yet to dominate the market.  Employees benefit as well since they may choose the apps that best suit their needs rather than relying on a one-size-fits-all solution that is “pushed down from the top,” according to Amble.

The report’s final major finding was the growing convergence of commercial and consumer enterprises.  People who have started businesses are using their favorite personal applications to help grow their businesses, and people are bringing their personal applications from home to work.  Previously, there would have been a distinction between the two: a predetermined stack for startup development.

Amble and Moore used the example of Canva, a well-known consumer software with a substantial business user base.  Canva didn’t even include an enterprise plan for years.  However, businesses are more inclined to combine individual and organizational use cases as they grow more difficult to differentiate.

“You can now sell into both of them; your TAM [total addressable market] is no longer one or the other,” she added.  According to her, businesses creating these goods may also “professionalize” more quickly, which entails assembling corporate teams such as go-to-market, sales, and support so they can begin making sales and generating enterprise revenue sooner rather than relying on individual customers.

In the upcoming years, Moore and Amble anticipate a rapid change in the list.  In an effort to remain accessible and relevant, older businesses are now implementing AI technologies, while new competitors provide fresh concepts.

“For legacy players, legacy practically translates to ‘what was twelve months ago,'” Amble stated.  Will the same note-taking apps still be available if we repeat this in a year?  Or will a brand-new set be released?