The AI industry continues to release tons of new models, and companies looking to stay competitive are racing to adopt them for their purposes. In fact, nearly 10% of businesses plan to spend a whopping $25 million this year on AI initiatives, according to tech consulting firm Searce.
But while lots of money is being spent on AI, it’s unclear the ROI is there. Half of all AI leaders aren’t sure how to calculate or demonstrate the value of AI projects, according to Gartner.
Ex-Airbnb data scientist Chetan Sharma makes the case that figuring out AI ROI is no heavy lift with the right tooling. Sharma is the co-founder of Eppo, an experimentation platform that lets customers evaluate and customize AI models for specific use cases. Beyond its model evaluation suite, Eppo provides a general A/B testing platform and service for apps and websites.
“With new AI models launching weekly and companies pouring millions into them, A/B testing offers a cost-effective way to evaluate their effectiveness without overspending,” Sharma said. “Eppo helps companies identify which models truly deliver value and enables smarter, more sustainable decisions in an environment of rapid innovation and escalating costs.”
Eppo competes with a number of experimentation and A/B testing startups in the market, including Split, Statsig and Optimizely. Big tech giants like AWS, Microsoft Azure and Google Cloud also offer a growing number of model fine-tuning and evaluation tools.
But Sharma says that Eppo stands apart from the crowd thanks to features like its “contextual bandit” system, which automatically spots new variants of customers’ websites, apps or AI models and actively explores the performance of those variants by serving increasing load or traffic to them.
“Experimentation drives velocity and accelerates growth by stripping away bureaucratic — and often incorrect — decisions by committee while tightly tethering initiatives to growth metrics, killing bad ideas fast while canonizing good ideas for reinvestment,” Sharma said. “Eppo’s approach to live ‘online-eval’ tests of AI models answers whether premium models improve metrics.”
Eppo, which launched out of stealth in 2022, now has “several hundred” enterprise customers in its roster, including Twitch, SurveyMonkey, DraftKings, Coinbase, Descript and Perplexity, according to Sharma. Alexis Weill, Perplexity’s head of data, told Dakidarts that Eppo has allowed Perplexity to “significantly scale” the number of experiments it runs concurrently.
Investors seem pleased. This week, Eppo closed a $28 million Series B round led by Innovation Endeavors with participation from Icon Ventures, Amplify Partners and Menlo Ventures. Sharma says that the new cash, which values Eppo at $138 million post-money and brings its total raised to $47.5 million, will be put toward bolstering Eppo’s marketing and AI experimentation capabilities, enhancing its analytics offerings and scaling its go-to-market efforts.
San Francisco-based Eppo currently has 45 employees and expects to end the year with 65.
“The demands of efficient growth along with the rise of AI has combined to an adapt-or-die mentality that forces companies to become experimental,” Sharma said. “And due to the gaps of legacy vendors, most of the experimentation market had chosen to staff large in-house teams and build over buy. With so much employee movement and layoffs, these in-house teams are no longer sustainable, leading to companies seeking out Eppo to replace expensive or orphaned in-house builds.”
