How AI agents will change the economy. Start with the lessons.

Patrick Collison, co-founder and CEO of Stripe, posted a thread on how AI agents will change the economy. Tobi Lutke, CEO of Shopify, shared it with a short verdict: “Yes, and well put.” I think it is one of the most useful things written about agents this year, for consumers and for anyone who builds software for a living.

I want to start where most posts end: with the lessons.

Lessons learned

1. Quality beats distribution, more than it used to.For decades the debate was whether a better product wins if nobody hears about it. When every buyer has an agent that researches each purchase as if it had hours to spare, the better product gets found. Making the product better becomes a more effective strategy.
2. Information about quality becomes the real asset.People can’t aggregate quality signals on their own. Agents can, and they want all of them. Data is valuable for training models, but it is also valuable as the complement to cheap intelligence: the model is only as good as the evidence it reads.
3. Tricks that depend on inattention stop working.Overpriced upsells at checkout, coupons only some people find, subscriptions people forget to cancel. All of these rely on buyers not paying attention. Agents always pay attention.
4. Paying for placement loses power.A lot of the internet runs on whoever shows up first. An agent does not click the first thing it sees, so rankings and placement fees count for less and real evidence counts for more.
5. It all depends on whose side the agent is on.Patrick is clear that his argument assumes personal agents work for the buyer. The same is true inside companies: an agent is only as trustworthy as the people it answers to and the evidence it can show.

Agents are the end of bounded rationality as a business model

Patrick’s starting point is that companies were, in a sense, the original superintelligence. They can spend enormous collective effort optimizing against individual customers who are busy, distracted or underinformed. Herbert Simon called this bounded rationality, and a lot of businesses quietly profit from it.

Personal agents change that balance. They make each of us a little more rational in how we deal with companies. Some business models get weaker. Others get much stronger. Patrick’s own summary of it is the line Tobi picked out: agents will act as “a kind of structural subsidy for product quality.”

What this means for enterprise software

I have been selling software to large enterprises for nine years, and I can already see this shift in B2B. Enterprise buying was always more rational than consumer buying, but it was slow and expensive. Security reviews, questionnaires, legal back and forth. Now buyers bring AI into every step. In my last article I shared how an enterprise contract that once took three lawyers and a month now takes one lawyer and a week, with an LLM doing the heavy lifting.

When the buyer can evaluate everything, the vendor can hide nothing. Vague claims about productivity and “AI-powered” features stop working. What works is evidence: what your product actually does, measured, for customers like the one reading.

Engineering quality becomes the strategy

If agents reward better products, then the quality of how you build becomes a competitive advantage, not an internal metric. And this is happening at the same moment AI is changing how software gets built. More code is written by agents than ever. Output is easy to inflate. Quality is easy to lose.

That is the part I care about most. Engineering leaders need the same thing Patrick says buyers’ agents need: honest, aggregated signals about quality. Not how many lines AI wrote, but whether it held up in production, how much rework it created, and whether teams are shipping better products or just more of them. I have written about why AI adoption is not the same as AI ROI, and this is the same lesson from the other side of the market.

It is also why we built Waydev the way we did. Waydev measures AI adoption, impact and ROI across humans and agents, and with Waydev 15B, our own SDLC model, it runs tailored to each customer, private and air-gapped inside their environment. The quality signals that matter most are the ones you keep inside your walls.

In a world of agents, the companies that win are the ones that can prove their quality.

Alex Circei, CEO and Co-Founder of Waydev

Patrick ends with a hopeful note: that companies who make things they are proud of will find the world a little more on their side. I share that hope. The work now is to make sure you can show, with evidence, that what you build deserves it.

If you want to start measuring that, our guide to measuring AI adoption, impact and ROI and the Impact That Compounds playbook are good places to begin.