The Dress that Nobody Needed

My wife isn’t a financial expert. She doesn’t have a massive amount of savings, and she doesn’t really understand the stock market. Yet despite this, she often tells me she's saved us money.

That’s not earned us money. She hasn’t actually found money. She hasn’t successfully fought a utility company on our behalf and secured a refund after six months of emails and threats involving the Financial Ombudsman. It’s none of the scenarios that would generally have put money into (or back into) our account. No, instead she's saved us money by buying something – and usually something we didn't need!

As far as I can tell, her logic seems to work like this. A dress was originally £300. Nobody in our house would ever spend £300 on a dress (I certainly wouldn't, as I don’t buy clothes for my wife, and I’m not the dress-wearing male type). But for some reason the dress is reduced to £120. And suddenly we're standing in a shop discussing the £180 we've apparently saved. At some point in the conversation I inevitably find myself wondering whether I've accidentally wandered into an alternative dimension in which spending £120 makes your bank balance go up. What makes it worse is that she will probably only wear the dress once and then decide she doesn't like it. Our local charity shop is full of clothes like this, mostly funded by us! However, valuing my life and certain parts of my anatomy, I usually don’t push the point.

The thing that fascinates me here isn't the maths, which is perfectly sound. Three hundred minus one hundred and twenty does indeed equal one hundred and eighty. It's the comparison that bothers me. We're comparing reality with something that was never going to happen. Nobody was standing there contemplating a £300 dress – even my wife has limits. Before the discount sticker appeared, the dress never existed in our lives at all. It was just another item hanging in a shop somewhere. Yet the moment a large percentage reduction appears, we stop comparing buying it with not buying it and start comparing it with a completely imaginary purchase that nobody was ever planning to make.

the moment a large percentage reduction appears, we stop comparing buying it with not buying it and start comparing it with a completely imaginary purchase that nobody was ever planning to make.

Retailers understand this perfectly well, of course. Entire industries are built around convincing us that hypothetical savings are somehow equivalent to actual money. Black Friday is essentially a national celebration of this idea. Every year people spend vast amounts of money to save money. It's extraordinary when you stop and think about it, but we usually don’t. We've somehow reached a point where buying a television you didn't want because it's 40% off is regarded as financial wisdom. If that logic worked everywhere else, I could buy a Ferrari reduced from £250,000 to £150,000 and proudly tell everyone I'd made a hundred thousand pounds. The bank manager might view matters slightly differently, but details like that tend to spoil the story. (Actually this is probably an even better example of buying something we don’t need, as I am currently unable to drive!)

You might be asking what all of this has to do with analysis, research or AI. Isn’t that what I usually write about? The reason is that this scenario comes to mind whenever I hear people talking about AI productivity. And the more I think about it, the more I think we're in danger of making exactly the same mistake.

Everywhere you look there are claims that AI can do in ten minutes what used to take three hours. Sometimes it's three days or three weeks. The examples vary and don’t really matter that much – it’s the message that’s important, and that’s always the same. Time has been saved. Productivity has improved. Efficiency has increased. The future is here, and it's arriving at remarkable speed.

I’m not saying that’s wrong in itself. Sometimes it's perfectly true. I mean, if you genuinely need something doing, then doing it faster is probably useful. I don't think there's much point pretending otherwise, no matter how much we might be AI averse. If I need a report, and AI helps me produce it in ten minutes rather than three hours, that sounds like a good thing. If I need a presentation and it helps me build a first draft quickly, that's really useful. And if it helps me to analyse information more effectively or automate a repetitive process, that has to be useful too. I’ve said in countless articles that I use AI myself. So I'd look rather foolish arguing otherwise.

If I need a report, and AI helps me produce it in ten minutes rather than three hours, that sounds like a good thing. If I need a presentation and it helps me build a first draft quickly, that's really useful ... I'd look rather foolish arguing otherwise.

But I've also said in a lot of my articles that I’ve spent a large chunk of my career surrounded by reports, dashboards, analyses and performance measures. One thing you discover fairly quickly in that line of work is that the existence of a report doesn't necessarily mean there was ever a need for one.

Of course some reports are genuinely valuable. They answer important questions and support decisions. A good report helps people to understand what's happening and what to do next. But other reports seem to exist because they've always existed.

I have worked with reports so old that nobody could explain where they came from. Nobody knew who had requested them (or who actually read them). The original purpose had probably disappeared years earlier. Staff had moved on, and departments had changed names. Entire organisational structures had come and gone, probably more than once. Yet the report remained, faithfully produced every month like some administrative ghost wandering through the system long after everyone had forgotten why it was there.

And the strange thing is that nobody wanted to stop it. “Perhaps someone will need it in the future." ”You know someone will ask for it the moment we stop producing it”. And other similar responses.

Stopping things, I've discovered, is one of the hardest things organisations ever do. Starting things is easy. Starting things is exciting. New initiatives arrive with launch events, strategy documents and steering groups. Someone designs a logo. Someone else writes a vision statement. Before long there's a PowerPoint presentation full of stock photographs showing ethnically diverse people in business attire pointing enthusiastically at flipcharts, apparently thrilled by whatever transformational programme has just been announced. Stopping things, on the other hand, has very little glamour attached to it. No senior leader gathers staff together to celebrate the successful retirement of a report nobody has read for five years. There are no ribbon-cutting ceremonies for dashboards that have finally been switched off. Yet, as I argued in A Good Data Die, most information assets should probably have a planned end as well as a beginning. Reports, measures and dashboards shouldn't acquire immortality simply because nobody wants to be the person who kills them. The difficulty is that organisations often treat creation as progress and continuation as safety, even when both have long since ceased adding any value. AI may make that tendency even worse. If a report now takes ten minutes instead of three days, the temptation will be to keep producing it forever because it seems harmless. But multiplying something pointless by low cost doesn't make it useful. It just makes it easier.

Stopping things, I've discovered, is one of the hardest things organisations ever do. Starting things is easy. Starting things is exciting ... Stopping things, on the other hand, has very little glamour attached to it.

I don’t know if you’ve realised it, but nobody books a conference centre because they've stopped producing a spreadsheet nobody reads. But you know what? They probably should. I think I'd attend a conference like that. There could be awards. Categories could include "Most Pointless Report Successfully Eliminated" and "Dashboard Nobody Opened For Five Consecutive Years". We could hand out certificates. Maybe a small trophy shaped like a trash can. Actually now I’m getting images of 3-2-1 with Ted Rogers doing impossible things with his fingers. So maybe not the best suggestion.

Anyway, the point is that effort often creates a natural challenge function. If a report takes three days to produce, somebody will (or should) eventually ask whether it's worth spending three days producing it. The cost of producing it in itself encourages scrutiny. The moment you tell someone a task consumes a significant amount of time, awkward questions start appearing.

“Who reads it?" "What decisions does it support?" "What would happen if we stopped it?" “Who would miss it?" “Could we do something simpler?

These are healthy questions and ones I have asked my teams many times. And they're exactly the sort of questions organisations should be asking more often.

The trouble is that AI can make those questions disappear. I mean, if the report now takes fifteen minutes, challenging it almost begins to feel unreasonable.

"It's only fifteen minutes."

And that amount of time doesn’t seem unreasonable, so the report survives. In fact, because it's now so cheap to produce, people may decide three more reports are wanted alongside it (this is not the same as the original report, or the new reports, being needed). Then perhaps a dashboard, and maybe an executive summary of the dashboard. And given it's so quick to produce, we could have a weekly version of the report because monthly reporting no longer feels ambitious enough. No, forget that; let's have the report in “real time” irrespective of whether anyone needs the information.

In this case the thing that's being saved isn't necessarily time. Sometimes what's being removed is the last remaining incentive to ask whether something needs doing at all.

Sometimes what's being removed is the last remaining incentive to ask whether something needs doing at all.

I've seen this pattern before. Most of us have, if you've been around a while and seen technology develop.

Take the example of email. Email was supposed to save time, and in fairness, it did at the start. Writing and sending a message became dramatically easier. You no longer needed paper, envelopes, stamps or a functioning postal system. You didn’t have to wait two weeks for a response in the mail.

The technology worked brilliantly. The only problem was that humans responded by sending emails about absolutely everything. The old barriers disappeared. The small moments where somebody might have paused and thought, "Does this really need saying?" vanished completely. The cost of sending something via the postal system disappeared (not that emails didn’t cost money, but they weren’t as in-your-face as a stamp on an envelope). Suddenly there were emails, replies, replies to replies, reminders, follow-ups, update emails and pre-meeting emails explaining the purpose of the meeting that existed largely because of the earlier emails that had been sent.

I've sat in meetings whose sole purpose appeared to be discussing an email chain that should never have existed in the first place. So every step was more efficient than before, but the overall process wasn't. And I think that's the bit that tends to get overlooked in conversations about productivity. We focus on the efficiency of individual activities rather than whether the overall system is getting any better.

We focus on the efficiency of individual activities rather than whether the overall system is getting any better.

Imagine a factory that discovers a way of manufacturing unwanted products ten times faster than before. Technically productivity has improved. The machines are more efficient. Output has increased. Targets are being exceeded. But unfortunately nobody wants the products. In this case efficiency hasn't solved the underlying problem. It's merely allowed the organisation to create unwanted things at unprecedented speed.

As an epidemiologist/statistician, I sometimes wonder whether we're in danger of doing something similar with data and information.

In this day and age most organisations already have more information than they know what to do with. So they produce more reports. They build more dashboards. Someone defines more metrics. Which often means even more data. They send more emails, often questioning the reports and dashboard and metrics. There are more meetings. More updates and more briefings. There's more strategy documents, more policy, and more guidance.

Scarcity of data isn't the problem. Our attention is. You see, the challenge isn't producing information anymore. Instead, it's deciding what deserves to be looked at. I’ve said this in a number of my previous articles. But still, many discussions about AI seem to assume that producing even more information even faster must automatically be a good thing.

I'm not convinced by any of this. Or perhaps more accurately, I'm not convinced it's always a good thing. Sometimes it is, but sometimes it clearly isn't. What worries me is how quickly we've started treating output as a proxy for value. For years we've been told, often quite rightly, that people shouldn't be judged by the number of hours they spend sitting in an office. What matters is what they produce, not whether they're visible at a desk from nine until five. That sounds perfectly sensible, and in many ways it was a much-needed challenge to an outdated way of thinking about work. Somewhere along the line, though, output seems to have become tangled up with volume. More reports must be better than fewer reports. More presentations must indicate greater productivity. More dashboards, more analyses, more content, more emails, more updates. We talk about outputs as though their existence alone proves their value. Yet quality and relevance have quietly slipped out of the conversation. A person who produces one report that changes an important decision may have contributed far more than someone who produces twenty that nobody reads, but only one of those examples generates a large number that looks impressive on a performance dashboard. AI risks accelerating that tendency because it makes volume astonishingly cheap.

Somewhere along the line, though, output seems to have become tangled up with volume. More reports must be better than fewer reports. More presentations must indicate greater productivity.

The danger is that we end up congratulating ourselves for producing more and more things at greater and greater speed, while paying less and less attention to whether any of them are useful. And that’s just busywork. Busywork is what happens when we become more concerned with producing outputs than delivering outcomes. The work gets done, the boxes get ticked and the metrics move in the right direction, but nobody can quite explain why the activity was necessary in the first place. The outcome isn't necessarily a more productive organisation. It may simply be a noisier one.

So the more I think about it, the more I realise that the most important productivity question has nothing to do with hours saved. Those calculations are easy. Before AI, three hours. After AI, ten minutes. Subtract one from the other, and of course everyone applauds. But the difficult question has to come first. “Why are we doing it? ”. Not how quickly can we do it, or how cheaply can we do it, but why are we doing it at all?

That's a much less exciting question because sometimes the answer turns out to be slightly embarrassing. Sometimes the answer is tradition. It could be that it's a comforting habit. Or it might be that somebody senior once requested something and nobody has had the courage to question it since. And in many cases nobody really knows. And if nobody really knows, making the process faster doesn't necessarily represent progress. Instead, it may just mean that we've become more efficient at doing something pointless.

Which brings me back to the dress.

The £180 saving only exists because we're comparing reality with an imaginary purchase that wasn't going to happen. Remove the hypothetical £300 price tag from the equation, and what you actually have is a £120 purchase and a slightly fuller wardrobe.

I occasionally wonder whether some AI productivity claims work in much the same way. "We've reduced a three-hour task to ten minutes." That’s fine. But if nobody needed the task in the first place, then we haven't saved two hours and fifty minutes; we've spent ten minutes producing something we'd have been better off ignoring.

if nobody needed the task in the first place, then we haven't saved two hours and fifty minutes; we've spent ten minutes producing something we'd have been better off ignoring.

My wife has a wardrobe full of bargains. Every single one made perfect sense at the time. Every single one represented an opportunity that couldn't be missed. I have a sneaking suspicion AI is doing something similar to organisations. Not because the technology is flawed, but because at the end of the day human beings are human beings.

So the irony is that none of this is really about AI. It's about us. AI didn't invent busywork, pointless reports, redundant dashboards or meetings that leave everyone wondering why they were invited. We sadly managed all of that perfectly well on our own. But what AI does is remove some of the friction. It lowers the cost. It makes it easier to produce one more report, one more update, and one more summary of the update explaining the report. Every individual decision seems harmless because the effort involved is so small. After all, it's only ten minutes. It's only a few clicks. It's only another dashboard. Yet when thousands of people make the same calculation, organisations gradually fill up with information in much the same way wardrobes fill up with bargains. Each item seemed sensible at the time. Each one was justified. Each one represented a saving of some sort. Nobody quite noticed the accumulation until there was barely room for anything else. And that's why, before we congratulate ourselves on how quickly AI can produce something, perhaps we should spend a moment asking whether it needed producing at all.

The most expensive thing in any organisation may not be the work that takes three days. It may be the work that takes ten minutes, gets repeated thousands of times, and leaves behind a wardrobe full of reports like the dress that nobody needed.