Am I Embarrassed to Admit I Use AI?

If the provocative title caught your attention, welcome.

This Flow Journal has taken me a little longer than usual to write. Partly because I like to think that what I put out there might offer readers something worth considering, and partly because I've spent much of the past few weeks reading, researching and trying to get a clearer picture of where we actually stand with something everyone seems to have an opinion on, but surprisingly few people seem to fully understand: artificial intelligence.

At times, it reminds me of the early days of Covid, when we all suddenly became virologists. These days, everyone is an expert in AI, agents, APIs, MCPs and whatever new acronym has appeared in the last 48 hours. And I have to admit, the deeper I dig, the harder it becomes to tell what's genuinely possible today from what's still science fiction, or simply a very convincing sales pitch.

The pace at which these tools are evolving is staggering. I could spend this entire article walking you through what can be automated, which platforms to use and what results to expect, but there's a good chance some of that information would be outdated in a couple of weeks.

One thing, though, changes far more slowly than technology: the way we approach problems.

Before we automate the chaos, shouldn't we understand it first?

A few days ago, I was talking to someone who was looking for a consultant to help introduce AI into their work. As they explained what they needed, a rather messy picture started to emerge: documents scattered everywhere, processes that had never really been defined, ways of working that changed constantly, organizational headaches and a remarkable amount of information that seemed to exist nowhere except inside their head.

The hope was that someone could come in, apply a little artificial intelligence and somehow bring order to the whole thing.

Red Flag!

I suspect this expectation is far more common than we'd like to think. Spend five minutes scrolling through social media and you'd be forgiven for believing that AI can practically run your business while you enjoy a coffee, solve every problem you've ever had and, ideally, make you a few million along the way.

From what I've learned so far, the reality is considerably less spectacular, and a great deal more interesting. AI can certainly help us get through an enormous range of tasks faster, but if we want to hand over an actual business process, we first need to be able to explain how that process works. And this is the part I find fascinating.

Think about all the things you do at work without giving them much conscious thought anymore, simply because years of experience have made them second nature. You know when something needs your attention, which information matters, when an exception is justified and when an apparently routine situation is actually a warning sign.

Now imagine having to explain all of that to someone who has never worked alongside you. Every step, every decision, every exception, including all those little pieces of knowledge you've accumulated over the years without ever thinking to write them down.

There's something rather wonderful about the fact that one of the most sophisticated technologies we've ever developed is pushing us back towards such a basic exercise: understanding what we actually do all day.

In some ways, it brings to mind the principles behind Frederick Taylor's scientific management in the early twentieth century: observing work, breaking it down into individual activities, examining each step and looking for more efficient ways to get things done. The context, the tools and the possibilities are obviously worlds apart, but the underlying need to understand and structure work before trying to optimize it feels surprisingly familiar.

And there's another point worth considering: if we don't understand our own work well enough, how are we supposed to notice when AI gets it wrong?

Please, don't make things up.

One of the things that surprised me most when I first started experimenting with AI was how often I found myself spelling out instructions that I would probably never think to give another person.

Don't invent information. Tell me when you don't know something. Don't take action without my permission.

I remember finding that rather unsettling.

When we give a colleague a task, we take an extraordinary number of unwritten rules for granted. We expect them not to make up figures, to tell us if something goes wrong and to ask questions when they're unsure. Of course, we all know people are perfectly capable of lying, withholding information or cutting corners, but there are still certain standards of behavior we consider so obvious that we rarely bother to articulate them.

With AI, those assumptions become much less reliable. A language model can produce a perfectly convincing answer even when it doesn't have the information it needs, filling in the gaps with something entirely fabricated and delivering the result with a level of confidence that, frankly, some humans would envy.

Which got me thinking about something else.

How often does a manager assign a task, absolutely convinced they've explained everything clearly, only to discover that the person doing the work understood something completely different? How many decisions are made on the basis of incomplete instructions, unspoken expectations or criteria that exist only in the manager's head?

And when the outcome isn't what was expected, how often do we blame the person who did the work rather than question whether we explained the task properly in the first place?

Perhaps it's an occupational hazard, given that communication is my field, but the more I explore how AI can be applied to business processes, the more I wonder whether a significant part of what we're struggling with is, once again, a communication problem.

After all, learning to give an AI system useful instructions means learning to set clear objectives, establish boundaries, define responsibilities and explain what a successful outcome actually looks like. All things we should already expect from a good manager.

So here's a question I hadn't expected to be asking: could working with artificial intelligence actually make us better managers?

Because perhaps, in learning how to make ourselves understood by a machine, we'll finally realize how often we expect people to understand us without having explained nearly enough.

Everyone talks about AI. So why does nobody admit to using it at work?

I recently attended an information session at SUPSI about artificial intelligence in small and medium-sized businesses, where two statistics caught my attention.

According to the figures presented, around 80% of AI users in Swiss SMEs are using tools they've brought into the workplace themselves. That raises some fairly serious questions about data protection, security and corporate governance, but I'll leave those for another day.

The second figure interested me even more: 52% of people who use AI at work are reportedly reluctant to admit it.

Only a few hours earlier, I'd been part of a conversation in which someone made a rather dismissive comment: "Come on, you can tell that email was written by AI."

And, as usual, I started wondering why.

We're surrounded by people who can't wait to tell us how much they know about artificial intelligence. Knowing how to use it is increasingly presented as an essential professional skill, and hardly a day goes by without someone explaining how it's going to transform the workplace. Yet when people actually start using it in their day-to-day jobs, a surprising number would rather keep it to themselves.

There's something odd about that.

I've seen enough businesses to know how difficult it can be to persuade people to adopt new tools, even when those tools are supposed to make their lives easier. And yet here we have employees going out of their way to bring AI into the workplace themselves. That suggests they've discovered a benefit substantial enough to make them change their habits without anyone having to persuade them.

My guess is that a big part of that benefit is time. But what happens to the time they save? And if I finish my work earlier, what happens next?

Imagine someone who can now complete a task in two hours instead of four. They might enjoy a less stressful day, give more attention to complex work, improve the quality of what they produce or finally get around to something they've been putting off for weeks.

Or things could go rather differently.

Their manager might discover that the task now takes half as long and simply give them twice as much work. It certainly wouldn't be the first time an increase in productivity resulted in higher expectations rather than better working conditions.

And I can't help wondering whether that might explain some of the reluctance to admit using AI.

Perhaps people worry that revealing how much more efficient they've become will mean losing the advantage they've just gained. Or perhaps they fear their work will somehow be seen as less valuable because something that once required considerable time, knowledge and experience can now be done much faster with the help of a tool.

There's another possibility that I find even more interesting: does using AI make us look more competent, or does it make us feel as though we're admitting we couldn't do the job properly on our own?

If someone produces an excellent report with the help of artificial intelligence, do we give them credit for the quality of the result, or do we somehow consider them less capable than someone who spent three times as long writing it without assistance? And why?

We've always relied on tools to do our jobs better. Nobody questions an accountant's expertise because they use accounting software, or thinks less of a designer for working digitally rather than drawing everything by hand.

Yet AI seems to trigger a different reaction, perhaps because it's entering areas we've always associated so closely with our own intelligence, creativity and individual abilities: writing, analyzing, reasoning and coming up with ideas.

The irony is that using AI well requires exactly those abilities. Someone who really understands their profession knows what to ask, can spot mistakes and has the judgment to evaluate different answers. Someone without that expertise might receive an impressive-looking result and have absolutely no idea whether it's any good.

We risk reaching a point where we judge people's competence by how much they can accomplish without assistance, rather than by the quality of the decisions they make with the tools available to them.

And I think that tells us considerably more about our relationship with work than it does about our relationship with artificial intelligence.

Here we are again.

I started out wanting to understand what AI could actually do for businesses, and somehow ended up thinking about internal communication, how managers delegate, trust between employees and employers, and even how we decide what someone's work is worth.

None of these problems are new. In fact, they were around long before artificial intelligence entered the conversation, and perhaps that's what interests me most.

AI is opening up new possibilities, some of them genuinely extraordinary, but the moment we try to bring it into the reality of everyday work, we find ourselves facing questions we probably should have been asking for years.

Can we actually explain how we work? Can we tell the difference between a good result and one that merely sounds convincing? Do we know how to delegate properly? Have we created workplaces where people feel comfortable using tools that make them more efficient? And, perhaps most importantly, what exactly do we plan to do with all that extra efficiency?

I don't have satisfying answers to all of these questions yet, and I'm quite comfortable with that. This edition of Flow Journal deliberately leaves more questions than conclusions, because I think we're still at a stage where claiming to have everything figured out would be rather presumptuous.

One thing, however, is becoming increasingly clear to me: technology is moving incredibly fast, while we keep dragging the same organizational, cultural and very human problems along with us.

And perhaps, before we spend too much time wondering how intelligent AI is going to become, we should ask ourselves how much smarter we're prepared to become in the way we work.