In my experience, if you really want to scare someone about the kind of dismal future our increasing reliance on technology will bring, you don’t need to paint dystopian sci-fi pictures of the unending battle between humans and machines, of robot uprisings or AI Armageddon, you simply need to utter those six little words that strike fear into the hearts of even the most digitally engaged – you just need to whisper “unexpected item in the bagging area”. Even typing it here sends shivers down my spine and has me reaching for the canned goods, survival candles, and a wind-up radio.


I find it amusing that something once hailed as a revolution in customer service is now being quietly phased out by the very same retailers that heralded its introduction. Far from the utopian retail experience we were promised, at best, these machines are awkward and obtuse, but at their worst, they’re maddeningly frustrating, seeming to delight in working just enough to get your hopes up before leaving you running out of the store teary-eyed and empty-handed, having spent 10 exasperating minutes playing the bagging area Hokey Pokey.


I won’t be sad to see self-checkouts go, but before they do, we need to recognize they offer us a crucial lesson that we’re going to need as we navigate a future filled with generative AI and other transformational technologies. The self-checkout isn’t just a flawed piece of technology; it’s a symptom of a broader issue – when technology is implemented to replace human experience rather than elevate it.


That’s the trap our entire industrial-era approach to work has been built on. We got very good at measuring output over input - speed, cost, throughput - and mistook that for progress. The self-checkout is efficient. It processes a transaction in less time, with less labour cost, than a person behind a till. It is also, for a meaningful number of shoppers on a meaningful number of days, a worse experience than the thing it replaced. Efficient and effective are not the same claim, and we keep building for the first while promising the second.


HR has its own version of this, and it’s been quietly proliferating for a decade. The applicant tracking system that filters out a brilliant candidate because their CV didn’t contain the right keyword. The annual review portal that turns a conversation about someone’s whole year into a dropdown menu. The benefits chatbot that answers “what’s my parental leave entitlement” with a link to a 40-page PDF. The exit survey nobody reads, filed into a system nobody opens. Every one of these was procured, at some point, on a business case built around efficiency - fewer admin hours, faster processing, lower cost per hire. Every one of them, at some point, made someone’s actual experience of work measurably worse, and the business case never had a line item for that.


Generative AI hands HR the same trap, except at a scale the self-checkout never had. It is extraordinarily good at automating the process side of work - screening CVs, drafting job descriptions, summarising engagement survey data, flagging attrition risk. All genuinely useful. All also, if we’re not careful, the fastest way yet invented to make the employee experience faster and worse at the same time. The question “what can we automate?” is the self-checkout question. It optimises for the machine’s job, not the human’s experience of the machine.


There’s a better question, and it’s the one this entire field of work should be built around: what does this make better for the person on the other end? The candidate waiting to hear back. The employee trying to understand their own pay review. The new starter on day three, still working out where the toilets are, being handed to a chatbot instead of a human.


Outside-in thinking means starting there, every time - not with the workflow you want to speed up, but with the outcome the person on the other end actually needs, and working backwards from that to whatever technology serves it best. Sometimes that technology is generative AI, genuinely elevating the experience. Sometimes the most human-centred answer is refusing to automate a moment that only works because a person is present for it. Both are legitimate outcomes of asking the right question. Neither is available to you if you never ask it.


Get this right, and the technology disappears into the experience - nobody thanks their HR system for working, because working is simply what it’s supposed to do. Get it wrong, and you’ve built a very expensive, extremely modern version of the bagging area hokey pokey, and your people will do it for years, quietly, without ever telling you why they’ve stopped trusting the systems you built for them.