AI’s Consulting Confession: When “Revolutionary” Tech Needs Human Handlers
So the business model answer to the LLM era of so-called AI is… consulting?
OpenAI recently launched enterprise consulting services, charging at least $10 million per client, deploying “Forward Deployed Engineers” who embed directly with organizations, a model that mirrors Palantir’s approach of rebranding IT services integration with military-grade language. This came right after OpenAI landed a $200MM deal with the DoD.
Meanwhile, Cursor significantly raised prices as demand for AI-assisted coding exploded, with developers complaining and cancelling, hopping to the next model that hasn’t yet been pressured to start charging something close to the actual cost of delivering the product.
The Wizard Behind the Purple Curtain
These aren’t just business model pivots. We’ve learned this lesson before… before IBM Watson.
In the early years, companies had a hard time escaping the “services as software” trap. You’d sell an enterprise contract, but behind the purple curtain, it was mostly people doing the work, Wizard of Oz style.
Industry whispers about IBM Watson back then were always the same: “You know it’s just a lot of people cleaning and labeling data.” Watson eventually shut down, but not before burning through billions, proving that human-powered “AI” doesn’t scale.
The VC Economics Paradox
The VC playbook was clear: if you wanted to be a real software business that could grow fast, you had to find repeatable use cases that didn’t require armies of humans, build a sticky interface, and design proper SaaS economics.
Inside a VC-backed firm during the early 2010s wave of machine learning AI, when I sold a $2MM enterprise services deal, VCs would frown. When I sold a $150k software deal, they’d pat me on the head, because software was supposed to scale without people in the cost structure; therefore, it was “higher valuation” revenue.
My competitors selling $24k and $10k “seats” got acquired for rich sums because they’d cracked the code: fast-turn, low-margin, no-humans-in-COGS SaaS businesses, with low-friction “just below the Director budget threshold where she didn’t have to get approval to pay for the deal” sales motion.
Full Circle, 200x the Price
I should have just called my human-heavy enterprise deals Forward Deployed Engineering and told those VCs that premium consulting is the business model.
Apparently, we’ve come full circle, except now we’re charging 10x more and calling it the most revolutionary technology transition in the history of all things, at risk of replacing all of our jobs, but creating lots of consulting gigs in the short term.
What This Really Reveals
The services push reveals that AI often requires significant human contribution, built on the extraction of energy, water, and human creativity, as well as underpaid data labelers and model trainers. It also relies on large defense contracts to generate a business model, the kind that people actually pay for, rather than freemium, subsidized consumer chatbots.
Most tellingly, it reveals who we are and what we believe is worthy of value.
What patterns are you seeing in AI business models? Share your observations in the comments.