The company that answers the question every board is asking - "is our AI actually doing anything?" - by counting tasks, one at a time.
Know and grow the ROI of your AI
The wordmark, plain against the light. No robot, no glowing brain - a company that sells measurement isn't going to sell you a metaphor.
Here is a fact about corporate AI in the mid-2020s: nearly every large company is buying it, and almost none of them can tell you what it returned. Workhelix started with that gap and, instead of selling more hype, decided to sell the number.
The pitch is almost aggressively unglamorous. A company signs on, and Workhelix does not begin by asking which division should be "AI-transformed." It begins by taking the organization apart - not into departments, which are too coarse to be useful, but into tasks. For a typical client that means examining more than 250,000 of them, and scoring each on a single, deflationary question: how much does a machine actually help here?
The answer, it turns out, is often "not much." Co-founder Erik Brynjolfsson - who runs Stanford's Digital Economy Lab and has spent a career on the relationship between technology and productivity - puts it plainly: there is a long tail of tasks where machines don't help that much, sitting right alongside other tasks where they help enormously. Most enterprise AI budgets are spent without knowing which is which. Workhelix's business is telling them apart.
This is a subtle and slightly subversive product. The AI industry's dominant sales motion is to promise that everything will change. Workhelix's motion is to hand a CIO a ranked list that says: this, yes; that, no; and this thing you were excited about, probably a rounding error. It is the rare vendor whose value proposition includes naming the places its own category is useless.
What makes the claim credible is who is making it. Three of the four co-founders are academic economists - Brynjolfsson, Andrew McAfee of MIT, and Daniel Rock of Wharton - the sort of people usually found publishing papers on exactly this question rather than incorporating around it. The fourth, CEO James Milin, came from Google and Amazon and was a founding member of the AWS private-equity sales team. The methodology is not a startup's back-of-envelope model; it is roughly a decade of published research pointed at a product.
The commercial model is contrarian too. In an era that worships pure software and its margins, Workhelix is proudly a tech-enabled services company. It ships the Nucleus platform, but it also deploys data scientists fluent in AI and econometrics to sit with customers and establish measurement cadences. The software finds the opportunities; the humans keep the scoreboard honest. Milin's team decided, deliberately, that measuring ROI is partly a human job - and priced accordingly.
The market appears to agree. The first dozen customers arrived without any paid advertising, on the strength of the founders' research reputation and word of mouth. Names on the roster now include Wayfair, Coursera, BAYADA, and Accenture - the last of which is investor, partner, and customer all at once. When a global consultancy that sells its own AI-transformation practice also pays you to measure AI, the measurement is doing something the decks are not.
There's this long tail of tasks that machines actually don't help that much with - and other tasks where the machines are very helpful.
Three things, none of them magic
Connects your workforce data and AI-usage history, then ranks AI opportunities by potential impact, flags your highest-performing AI users, exposes where AI is being underused, and estimates the time and cost it's saving.
Decomposes roles into ~250,000 discrete tasks per client and scores each for how suitable it is to generative AI - producing a prioritized roadmap grounded in a decade of productivity research, not vibes.
Data scientists with econometrics and AI backgrounds embed with your team to set measurement cadences and quantify the real performance impact and return of each AI initiative over time.
Illustrative only - directional depiction of Workhelix's task-scoring thesis, not client data.
Proportions illustrative; round total is $15M. See funding section for named investors.
Veteran of Google and Amazon and a founding member of the AWS private-equity sales team; previously co-founded a venture-backed AI platform in the Bay Area. The operator who turns the research into a company.
Director of Stanford's Digital Economy Lab and one of the most cited scholars on technology and productivity. His task-based framework is the intellectual spine of the product.
Co-Director of the MIT Initiative on the Digital Economy and principal research scientist at MIT Sloan. Longtime chronicler of how digital technology reshapes the business world.
Professor at the Wharton School and research fellow at Stanford and MIT. Keeps the methodology rigorous as it meets the messiness of real enterprises.
The Series A
In February 2025 Workhelix raised $15 million, led by AIX Ventures. The syndicate reads like a who's-who of people who think about AI for a living.
Who's Buying
Enterprises and more than a dozen publicly traded companies. The first dozen came without a single paid ad.
James Milin joins forces with economists Brynjolfsson, McAfee, and Rock to commercialize task-based productivity research.
The company publicly unveils its platform to pinpoint generative AI's business value for enterprises.
Led by AIX Ventures with Accenture Ventures, AI Fund, Bloomberg Beta, and a roster of marquee angels including Reid Hoffman and Mira Murati.