I don't need an analyst to spend two weeks reviewing a data room any more.

That's not a prediction, it's already true. AI can work through thousands of pages of due diligence and give me something useful the same afternoon. It can build a financial model. It can draft an information memorandum faster than most analysts can, and what it produces is usually a reasonable place to start. I use it constantly, and not only for the mechanical things. I think with it. I use it the way I'd use a colleague at my own level whose opinion I respect, which is a strange yet honest sentence to write.

I'm not necessarily writing this to defend my industry. Quite a lot of what we used to pay people to do can now be done by software, and pretending otherwise is simply denial.

I believe the conclusion most people draw from that is simple. If the work an analyst does can be done by a machine, hire fewer analysts.

We've watched a version of this here already. Between 2019 and 2024 the big four shed somewhere around seven thousand roles between them, Nedbank hardest of all, and the reason given was almost always the same. Customers moved onto digital channels, in-branch transactions fell away, and the branch network shrank behind them. That was a story about tellers and back-office staff and physical footprint, and most of it has already happened.

What's arriving now reaches a different kind of employee, and I don't think we've thought about it nearly as carefully. I'm yet to see anything published that shows deal teams shrinking at the bottom, so this is still a question rather than a trend. But the incentive is sitting there in plain sight, and incentives like that usually get taken.

I think that response would be wrong, and not for the reason people usually give.

Start with how anybody learns to do this job at all. You can read about a thousand transactions and still not understand why a particular mechanism was used instead of the obvious alternative. I learned structuring by putting structures together and then watching what happened to them. By proposing something that looked optimal on paper and discovering the market simply doesn't accept it that way. There is a version of every one of those lessons written down somewhere, and none of it did anything for me until I was sitting in a room where somebody explained, with varying degrees of patience, why the clever thing wasn't going to fly.

Experience is the only thing that translates theory into practice. I don't think that's a controversial claim. What gets discussed far less is the mechanism by which anyone actually gets that experience, and I was lucky in a very specific way.

I worked closely with my CEO early in my career. When I worked on something, he took me into the room, not because I was needed there. I sat in meetings where I was too nervous to speak and wouldn't have known what to say if I had been. Some afternoons the most useful thing I did was take notes. But I was in the room, I was expected to have a view, and I developed much faster than I would have at my desk.

Most analysts never get that, and the reason why is the part I think we have backwards.

They weren't historically kept out of rooms because they were too junior to be there. They were kept out because they were busy. Somebody had to be in the data room for two weeks. Somebody had to rebuild the model overnight so the version was ready by six. That work was real and it had to be done by a person, and the person doing it was at a desk instead of in a meeting.

The desk was the barrier, not the hierarchy.

Which means the thing everyone is anxious about may do the opposite of what they expect. If AI absorbs the two weeks, the analyst's time comes back. My time comes back too, and that's the part getting all the attention, but theirs is worth more. They have the most to learn and, until now, the fewest opportunities to learn it.

That raises a question I don't think we've properly answered. What do we do with the time we've just handed them?

Standard Bank's chief operating officer, Margaret Nienaber, said something along these lines when the group published on its own AI rollout. AI can give people time back, but what you do with that time is what actually matters. I agree with her. In our business I'd put it more narrowly than she did. You take them into the room.

I've seen what that exposure produces, and the clearest example I have wasn't an analyst at all. Some years ago I spent a long time in meetings with a woman who wasn't a dealmaker in any formal sense. Her job, on paper, was to take notes and make sure that whatever was agreed actually happened afterwards. Chase the people who owed things. Keep the process from slipping.

It sounds administrative but in practice she understood those transactions better than most people formally working on them, because she had been in every single conversation where they were shaped. She could manage upwards. She knew what her chief executive would want before he said it, which freed him to spend his attention on the parts of a deal that genuinely needed him.

When she engages with me, she does it as a peer, not as a nervous junior asking permission. I've thought about this a fair amount and I suspect she could move into an executive role in my world more easily than some analysts could, and she has never built a model in her life. Nothing about her technical training produced that. Sitting in the room did.

There's a version of this problem that starts much earlier than anyone's first job, though, and it's worth naming because it shapes who arrives in our industry in the first place.

When I was choosing what to study, the advice was near unanimous. Become a chartered accountant, because most of the chief executives in South Africa are chartered accountants. It was accurate. It was also advice about where the jobs had been, not where they were going.

I switched to straight finance and went into the market. A lot of the people I studied with did their articles. Several of them then moved into something else entirely, a hedge fund, a property team, a different division, and started again as analysts at twenty seven.

The interesting part is that they weren't short of practical training. Articles are three structured years of it, which is more than most finance graduates ever get. The problem was that not all of it transferred. They had spent three years accumulating experience in a destination somebody else had chosen for them, and when they arrived where they actually wanted to be, they were behind people who had been doing that specific thing since twenty two.

Exposure only compounds if it's exposure to the thing you'll end up doing. An apprentice electrician does practical after practical toward one clear outcome. We hand people a theory syllabus, a desk, and a career map drawn twenty years ago, and then we tell them the market has moved.

All of which leaves an honest problem with everything I've just argued, which is that I don't think anybody is going to do this voluntarily.

If AI saves two weeks of analyst time, the tempting move is not to take the same analysts into more rooms. It's to employ fewer of them and keep the difference. That's a decision that improves this year's numbers, and the cost of it doesn't show up for fifteen years, by which point the people who made the call have moved on.

No firm feels the absence of people it never trained inside a reporting cycle. So the rational thing to do, every year, is nothing. Which is roughly how a profession ends up short of senior people without anyone ever having decided to be short of them.

Doing it differently has to be a deliberate choice made against the incentives, and deliberate choices are a weaker force than a spreadsheet. I don't have a clean answer to that. I just think it's worth highlighting, because most of the writing on this subject skips straight past the part where somebody has to absorb a cost.

For whatever it's worth, I take my analysts into the rooms. Every deal they work on, they're in the meetings for it. I want them there, I encourage them to speak, and I'd rather they said something wrong in front of me than said nothing for two years. The difference it makes is visible, and it shows up in the things that are hardest to teach any other way.

If we simply cut analysts because the modelling is handled, we won't have an industry in twenty years. We'll have a small number of very experienced people and nobody behind them, and the value of what they know will go up for the worst possible reason.

Some of the people I bring into those rooms are too nervous to say anything. That's fine. So was I.