How to Think About AI — Edition 2

The lamplighter

For most of the nineteenth century, every gas-lit street in London required a lamplighter. The role was real, skilled, and unionised. Lamplighters carried long poles, knew their routes by heart, applied judgement about weather and timing, and trained through formal apprenticeships. The work mattered. Without them, the streets stayed dark.

Electric street lighting did not replace the lamplighter by building a better lamplighter. It did not produce a machine that walked the streets carrying a pole. It removed the need for the role altogether. The lamps came on by themselves. The outcome of illuminated streets remained exactly the same. The process disappeared.

A lamplighter watching this unfold could have argued, accurately, that no machine could walk their route the way they did. They would have been right. They would also have been answering a question the city had stopped asking. The city did not want a lamplighter. It wanted lit streets. Once the streets stayed lit by themselves, the route the lamplighter walked was no longer part of the answer.

The bank teller

The same pattern arrived in living memory. Until the 1980s, bank tellers handled every routine banking transaction in person. Deposits, withdrawals, balance enquiries, cheque cashing. The job required real skill. Recognising customers, handling cash accurately under time pressure, exercising judgement about who needed help and who needed scrutiny. It was a substantial profession with clear progression and professional standards.

Automated teller machines did not replace the teller by building humanoid robots at the counter. They gave the customer direct access to the same transactions through a different route. Then, online banking went further and removed the need to visit a branch at all. The tellers who remain today still do real work. Complex problems, fraud handling, relationship banking, are the parts of the job that require genuine judgement and presence. But the volume of work that flowed through tellers collapsed even as the volume of banking transactions grew.

This is the shape worth looking at carefully. The skilled parts of the teller’s job did not disappear. They are still being done, by fewer people, in different roles. What disappeared was the routine work. The work that used to make the role economically viable at scale. The teller’s expertise was real. The need for that expertise to be applied to every transaction was not.

What both stories show

In both cases, the people doing the work could correctly say that a machine could not do what they did. In both cases, that statement was accurate and irrelevant. The question stopped being whether a machine could replicate the role. The question became whether the outcome could be delivered without the role. Once the answer was yes, the role contracted to the parts where the outcome still required it.

This is the position surveying is in now.

Defending the process assumes the process is what the client is paying for. Most of the time, it is not. The client is paying for the conclusion that the process produces. The judgement. The recommendation. The number. The defensible position. Once that conclusion can be reached through another route, the original process is no longer the product. It becomes the cost of producing the product. The work involved is real and demanding. It just stops being what the outside world is buying.

This is what many firms have not yet absorbed. The integrity of the process does not, on its own, secure the work. The outcome does. And when the outcome becomes reachable another way, the process moves from being a value driver to being an overhead.

What gets missed

The “machine cannot do what I do” argument can leave two important things out of view.

The first is the question that actually matters now. It is not whether AI can replicate the surveyor. It is whether the outcome can be delivered without the surveyor. Those are very different tests. AI has been making quiet progress on the second one for some time, while much of the profession has been engaged with the first.

The second is the shape of the threat. The genuinely uncopyable parts of professional value, deep judgement, presence in difficult conversations, accountability when things go wrong, are not where most of the billable hours sit. Most hours sit in the routine, documentable, repeatable parts of the work. The parts where a client could plausibly find another route. The defence is built around the parts AI cannot touch. The exposure is in the parts that can.

The lesson from the lamplighter and the teller is not that whole professions disappear. It is that the routine work disappears first, and the role contracts around what genuinely requires it. The surveyors who will still be busy in 2030 are the ones who have understood this early enough to reposition.

The shift

Stop defending the process. Start defending the outcome.

That shift is not simple, because it changes how a firm describes itself, prices itself, markets itself, and explains itself to clients. A firm that defends process tends to sell reports and time. A firm that defends outcomes sells confidence and answers. Both are legitimate. Only one of them is well-positioned for the market that is taking shape.

This does not require abandoning expertise. It requires repositioning it. The expertise that holds its value is the expertise that produces the outcome a client genuinely needs, in a form that is defensible and accountable when things go wrong. Everything else is method. Methods change. Outcomes do not.

A test for the week

Imagine a client asks what your firm does. You are not allowed to use any of these words: survey, inspection, report, valuation, assessment, methodology, professional standard.

What do you say?

If you can answer cleanly, you understand what you are selling. If the words do not come easily, that is useful information too. It points to the work worth doing next.

This is the second edition in a six-part series on how to think about AI in surveying. Each Monday I set out one principle that is challenging the profession, and what to do about it.

Next Monday: why AI does not need to look like a surveyor to displace surveying work, and how the assumption that it does is leading firms to underestimate the threat.

Good call, and you are right — one CTA does more work than two.

Two reasons. First, the briefing registration captures the email and adds them to the newsletter list automatically, so a separate newsletter CTA is redundant for anyone who registers. Second, asking for two actions at the bottom of an article splits attention and weakens both. One clear next step is always stronger than two parallel ones.

The briefing is also the higher-value capture. A briefing registrant is closer to becoming a paying customer than a newsletter subscriber. They have committed time and given you their attention live. The newsletter sign-up is a useful secondary outcome but the briefing is the strategic one.

I have looked at your landing page properly to pull the right details. The briefing is on Wednesday 17 June 2026, 1pm to 2pm UK time, online, free, informal CPD. The June edition is themed “The Regulatory Landscape” and covers the EU AI Act, Article 4 AI literacy, the RICS Standard, UK regulatory changes including the DUAA 19 June deadline, and where firms should start.

Here is the CTA, calibrated for the bottom of Edition 2.

Want to go deeper?

My next free AI briefing is on Wednesday 17 June, 1pm to 2pm UK time, online. The June edition covers the regulatory landscape now shaping the profession: the EU AI Act, Article 4 AI literacy, the RICS Standard, the DUAA changes from 19 June, and where firms should start.

One hour. Practical, designed for surveyors. Free. Counts as informal CPD.

Register here →

Until next week

Nina