AI Makes Trying Cheap and Trust Expensive

Most of the conversation about AI focuses on what it can do. For growth, I think the more important question is what it makes cheap. Building a prototype, drafting a proposition, analysing a market and writing code now take days rather than weeks. When something becomes cheap, its value falls, and value moves to whatever is still scarce.

I call this “the scarcity shift”. As AI lowers the cost of trying things, the parts of a business that are hard to copy become more valuable. The first is trust, the permission customers give you that I described earlier in the series as the trust radius. The second is context, the knowledge of how your customers work that comes from finishing more of their job. The third is position, your place in a customer's day and workflow. The fourth is accountability, someone prepared to stand behind the output when it matters. None of these can be generated by a model, and all of them are things established businesses tend to have and new entrants tend to lack.

Silver, the venture I co-founded this year, is built on exactly this idea. We help regulated professionals use generative AI without taking on unmanageable risk, starting with building surveyors. Drafting a building survey with AI is now relatively easy, and anyone can attempt it with a general-purpose tool. What isn't easy is doing it in a way a surveyor can put their name to, their regulator will accept and their insurer will cover. A report that lenders, buyers and insurers rely on is only as valuable as the confidence behind it.

So the value of Silver lies much less in the drafting than in the trust around it: the regulatory awareness built into the workflow, the audit trail showing what the AI did and what the professional checked, and the confidence it gives the people who carry the risk. We've spent far more of our time with the profession, its regulator and insurers than on the drafting itself, because that's where the scarcity is. The drafting will keep getting cheaper for everyone. The trust won't.

AI also changed what it cost to find that out. A venture like Silver would once have needed a large team and a long build before it could put anything in front of a customer. Today the underlying capability is available off the shelf, which meant we could test the hard question, whether professionals and their insurers would trust it, far earlier and far more cheaply than would have been possible even a few years ago.

That has two consequences for established businesses. The first is that the old reason for not experimenting, that it was too slow and expensive, has largely gone. The unfinished job is cheaper to finish, the accepted burden is easier to lift and the next stage of a staged bet costs less to test. The second is that features can now be copied almost overnight, so the product itself is a weaker defence than it used to be. The advantage lies in the relationship and the context around it.