The first two years of enterprise AI were about adoption at any cost. The next phase is an audit, and investors are paying for the auditors.
Larridin, based in San Francisco, raised $30M in Series A funding led by Pelion Ventures, with Andreessen Horowitz, Gradient Ventures, GT Ventures, W23 and several others participating. The company has gone from a single enterprise customer last December to 50 today, according to chief executive Russell Fradin.
The platform hooks into internal systems and studies how employees actually use AI, including the prompts they send to different tools, then scores who is producing results and who needs training. Chief technology officer Ameya Kanitkar says prompts are analysed without storing their contents, so the output is a score rather than a transcript.
Fradin’s argument is drawn from advertising. He spent years at Comscore, where third-party measurement became the currency that advertisers and sellers negotiated over. He expects AI spending to demand the same referee, and a16z general partner Alex Rampell describes the gap as an absent “AI boss” telling staff what to do and not do with the tools.
The timing is not accidental. Companies spent much of 2026 pushing employees to use as many tokens as possible, with some tying AI usage to performance reviews. Ugly bills arrived mid-year, and the mood flipped toward model routing and cheaper open-weight models.
The company plans to double its 20-person team across sales, engineering and marketing.
For startups selling into enterprises, the lesson is blunt. The budget exists, but it is moving from experimentation toward proof, and tools that cannot show a return are the first line item cut.
