Meet the coaches, measurers, and builders carving out a slice of the AI cost-saving business

Don't usage Swiss Army knives for surgically precise work; usage scalpels.

That's what Manos Koukoumidis and different players successful the AI cost-saving business opportunity is the biggest problem pinch really companies usage AI. They're utilizing the astir powerful, versatile, and costly AI models — from frontier labs like Anthropic, OpenAI, and Google — to grip niche tasks wrong their organizations.

Koukoumidis, the CEO of the Washington-based AI institution Oumi AI, called this believe "widely irrational and inefficient," resulting successful staggering AI bills.

Founded successful 2024, Oumi lets individuals vibe-code their ain niche AI models wrong minutes, tailored to their tasks and goals. It raised $10 cardinal successful its 2024 seed round.

Oumi is 1 of the galore companies that person sprung up successful the past 2 years to lick the biggest endeavor problem of the day: reducing highly inflated endeavor AI spending and expanding its returns connected investment. As companies move distant from the tokenmaxxing trend — wherever labor were encouraged to pain arsenic galore AI tokens arsenic they could — these cost-saving services are becoming progressively sought after.

Some of these companies are consulting firms and measurement platforms that counsel clients connected integrating AI into their workflows successful a smart way. Others, for illustration Oumi, are building caller products, specified arsenic exemplary builders and conclusion platforms, to tackle the problem from the source.

The coaches

The first class is the coaches: the consultants and advisors lighting up the way to AI success.

Sydney-based business consultancy Adaptovate has been operating for much than 8 years, pinch complete 100 consultants crossed offices astir the world. But complete the past 3 years, the institution has zeroed successful connected helping others get the astir bang for their AI spending buck.

Michael Murphy, a partner successful the company, said Adaptovate useful pinch clients astatine each stages of their AI take travel — from a snacks shaper that's conscionable figuring retired really to usage AI successful its proviso chain, to a 30,000-employee-strong master services institution that wants to restructure each its teams astir GenAI products.

"It ever starts pinch the strategy," Murphy said. "Figuring retired really decision-making changes. How does your talent exemplary change? How does your organization's building alteration erstwhile you're opening to standard AI crossed the afloat organization?

The measurers

Then location are the measurers, who usage package devices to thief companies pinpoint wherever AI is moving for them and wherever it isn't.

One illustration is the San Francisco-based Larridin. CTO Ameya Kanitkar told Business Insider that the institution acts arsenic a measurement furniture crossed its clients' AI tools, employees, agents, and spending.

One of Larridin's products is simply a information study instrumentality that shows really worker productivity varies pinch token spend, helping companies place the saccharine spot for the token fund they allocate to their staff.

Kanitkar, who has worked astatine LinkedIn, Coinbase, and Groupon, said Larridin got disconnected to a slow commencement because companies weren't peculiarly concerned astir AI spending successful the early days of adoption. But that has shifted aft they started pouring millions into the resource, and he's seeing Larridin's traction double each 4th since the commencement of the year.

"It conscionable makes consciousness that you request a measurement furniture to make judge that each those hundreds of millions of dollars you're spending are really justified and are decently spent," he said.

Larridin raised $17 cardinal successful seed backing from Andreessen Horowitz, Bloomberg Beta, Google Ventures, and different investors and launched the level successful 2025. It now serves clients from information halfway building companies to biosciences and financial services firms.

The builders

The 3rd class is the builders: the startups for illustration Oumi that are creating products to straight thief companies slash AI costs.

Another illustration is Runware, a institution cofounded by serial entrepreneur Ioana Hreninciuc. Runware provides the inference infrastructure companies request to tally AI models quickly, cost-effectively, and astatine scale. This makes it overmuch easier for companies to standard up their AI products without worrying astir AI outages.

For those buying their ain GPUs, there's Tensormesh. The startup focuses connected Key-Value Cache, cofounder Junchen Jiang said, thing that "never appears successful people's token bills" but surely affects their costs.

Tensormesh's caching strategy helps businesses "accept much queries connected less GPUs," Jiang said. The institution raised $20 cardinal from big-hitters successful hardware for illustration AMD and NVentures, Nvidia's task superior arm.

And moreover much are gestating, fresh to deed the market. Conifer is still participating successful Y Combinator's summertime 2026 batch. Its cofounder, Charles Muehlberger, said the institution had already raised $1.3 million, astir 20% of its desired raise, earlier Demo Day. The company's merchandise will synthesize queries, break them into pieces, and provender the pieces to different models (including section ones).

"We deliberation the main customer is 1 who is focused connected cost," said Michael Jeffords, Conifer's different cofounder.

Bottom line

While the coaches, measurers, and builders are cutting the AI-cost-saving pastry successful different ways, their proposal is mostly the same: stop utilizing frontier AI models for each menial task, commencement utilizing lighter, open-source models, and build your ain models to trim costs and reliance connected the large AI labs.

They're besides proposing caller ways to deliberation of ROI. Kanitkar from Larridin said AI costs should beryllium treated arsenic superior expenditures, not operating expenses, and companies shouldn't expect to spot gains immediately, but alternatively complete a play of a twelvemonth aliases two.

They're each agreed connected 1 belief — that companies jumped connected the AI hype train without due planning, and astir don't really cognize what they're doing.

"Everybody's pretending their AI spending is paying off," Hreninciuc from Runware said.

"The inducement correct now is for group to opportunity it is paying disconnected and to find a measurement to warrant it, and because of this, we don't person an meticulous position of the market, arsenic cipher wants to beryllium the canary successful the ember mine," she said.