Larridin, a startup selling software that shows companies what their AI bills buy, has raised a $30 million Series A led by Pelion Ventures, it announced on 2 November. Seed backers Andreessen Horowitz and Gradient Ventures returned, joined by six new investors including GFT, W23 and Thayer. Total funding is now $47 million.
The size matters less than the signal. Investors are paying for a product built around the question finance teams now ask about AI: what did we get for that money?
What Larridin sells
Larridin connects to a company’s AI tools, usage data and provider bills, then ties that spend to the work people and agents produce. For engineering teams, its published method scores each merged pull request (a reviewed code change accepted into the codebase) by model-assessed complexity. It discounts sloppy work and missing tests, then compares output with invoiced AI spend.
The software reads employees’ AI usage, including prompts. Chief technology officer Ameya Kanitkar told Business Insider it analyzes those interactions without storing their contents. We could not check that independently.
Named customers include SurveyMonkey, Vertiv, Klaviyo, Gainsight and EcoVadis. Chief executive Russ Fradin told Business Insider the company went from one enterprise customer last December to 50, and plans to double its 20-person team.
The numbers, and whose they are
Every figure below is Larridin’s own customer data, not audited by anyone else. The customers are large companies, not ten-person shops.
- Larridin claims a 47% average increase in “measurable team output”, using its own definition of output.
- At one customer, 92% of engineers used AI, yet one engineer ran up roughly 65% of the team’s AI spend in a single week.
- At another, two AI tools doing comparable work differed 66 times in cost per shipped change.
- In its August coding spend benchmark, the median engineer drew $213 a week in billed AI coding spend and the 90th percentile $911. Flat-fee plans never reach a metered bill, so true use is higher.
The release also says AI-assisted pull requests rose from 35% in April to 62% in September, across more than 1,000 engineers. Over the same period, the median wait for a first human code review grew by 50%.
The finding most useful to a buyer cuts against simple budgeting. Extra spend bought more output only for engineers already fluent with the tools. For one low-adoption group, output stayed flat from $21 to $421 a week.
Why investors want independent measurement
Jeff Kearl, the Pelion partner who led the round, told Business Insider that AI has become one of the five to ten biggest items on company ledgers. Fradin, a Comscore veteran, said that when money floods a category, someone builds an independent company to measure it.
That bet fits a run of deals in AI operations tooling, such as Traversal’s $120M round for AI site reliability work. It also fits vendors pricing agents by plan tier, as with Perplexity putting ongoing agent work on its Max plan.
Larridin is built for enterprises, and we found no public prices on its site. A small team is unlikely to buy it, but the problem scales down.
What to do this week
Pull last month’s invoices for every AI tool and API key your team uses, and split them by person or project. If one person or tool dominates, ask what work came out of it. Start with coding agents that run unattended, like the Kiro bots AWS built for recurring chores, and set a review trigger for when anyone’s weekly spend doubles.
Sources
Get the weekly email
The week's most useful AI news, a tutorial and an interview.
Martin Scott
Martin Scott is the editor for business and policy at AI Observer, covering startups and deals, regulation, health and science, and AI agency profiles. He follows the money and the legal text, and separates what a rule says from what people claim it says. His style is analytical and plain-spoken.
Spotted a mistake? Tell us on the corrections page.