Everyone is already so over AI. It’s the normal post-trend hangover: we marvelled, we bought, we prompted, we fatigued. Of course, AI isn’t going anywhere, but as I wrote recently, it’s now moving into that slightly boring phase where we use it but we’re a tiny bit less enamoured.
Everyone is complaining about how ‘AI-sounding’ emails and blogs and social media have become. That isn’t a bad thing, exactly. It’s a bit like coming up for air after slightly drowning ourselves in AI-first everything. Like everything else, we need this in moderation.
Leadership teams have spent the past two years (at least — many for the better part of a decade or longer) treating AI as a disruptive event, and rightly so. It is one, still — that has not ended. The potential is enormous, we’ve only begun to tap it, and I won’t even bother trying to imagine what the true 2.0 stage is. So the attention it has commanded at the top of organisations has been earned. (I’d worry more about a leadership team that hadn’t been consumed by it.)
But. I do think there’s about to be a slight pivoting of AI at this point. It isn’t even really about slowing down. It’s about a bit less experimentation, and more about integration.
Normalisation of a new technology is perfectly typical and totally boring.
Most organisations are still in the first-date phase with AI: embracing it, getting to know it, testing what it will and won’t do. What comes next is less glamorous. The relationship has to settle into the infrastructure. And the leadership conversation has to move from experimental discovery of what it will do someday to a quieter (more boring) question: what is the best use within our current operations.
If you’re old enough to remember late-90s corporate culture, you saw this when the world wide web began to normalise into our lives. Email and e-commerce arrived as the story. You ordered anything online (nervously) just to see if it really did go from pushing a button to arriving in your mailbox.
And then it stopped being interesting. Mail order catalogues slowly and quietly began to fade. We bought not only books and CDs (!) but plane tickets and hotel reservations as we stopped being fascinated by online shopping and simply shopped. Then, at scale, the genuinely disruptive thing began to happen. Business opportunities online began to be bigger and more transformative than merely an online version of their analogue predecessor.
AI at this stage remains extraordinary, but we are still a long way from this technology changing how humans actually decide, judge, and behave. The C-suite is right to keep pushing on the future of AI but they don’t have to get ahead of it. In fact, I think when companies begin to integrate AI into their current operations, we will begin to see a more natural evolution of what it can do next.
The thing leadership might need to reconsider right now is using AI at the cutting edge of its capability. That’s a reflection both of its still considerable limitations, and our still developing controls on checking its boundaries. And if we layer one AI decision on top of another, we risk compounding uncaught errors.
Consider how much of what reaches a decision today is already AI-derived. An analyst produces numbers with an AI tool. Her manager folds them into a deck drafted by AI. The SVP runs the deck’s figures through an AI modelling program. And the executive team then relies on that output to make the call. There is some risk in every link of that chain. There was always human risk too.
But we’ve spent generations of organisational culture building systems for catching human error. We know what an overconfident junior person sounds like. But AI sounds polished and precise. We limit reports to volumes and frequencies humans can produce (and digest). But AI encourages us to ask for infinite amounts of data and we can’t spot check that volume for accuracy.
Integration before innovation: building AI Into infrastructure is the first efficiency
To be clear, I don’t believe companies have handed over their operations wholesale and consciously decided to let an AI model make the call. But that also isn’t how it happens, I don’t think. Nobody decides to let AI take over. The reliance on the tool bleeds into the operating rhythm one convenient handoff at a time. And then ‘AI-assisted’ becomes ‘AI-driven’ and no one even knows.
Which leaves leadership teams with the less glamorous ambition: make AI part of the infrastructure, deliberately, rather than trying (yet) to get it closer to the decision-making. Use AI to analyse data but have more robust checks on the output. Let it inform ideas but document which of those options were not limited by a human brain, which usually filters for certain ethics and histories and impacts that AI cannot. And for now, reserve choice itself — from which market to which hire — for the humans who will have to answer for it.
