Day three at Dublin Airport. I came over for two days of fractional AI leadership, the part where a firm’s AI ambition meets what regulation will actually allow. The version of that on LinkedIn is boardrooms and strategy days. The version I am living is two cancellations, three different hotels, and a receipts folder that now has its own filing structure.
Somewhere behind that departure board sits some of the most mature predictive modelling in commercial use. Crew rostering. Load balancing. Fuel burn, connection modelling, disruption forecasting. Decades of it, running at a sophistication most of what gets sold as AI today cannot touch.
None of it got me home.
What did was an Aer Lingus agent who read three screens, made a judgement call, and had me rebooked in ninety seconds.
Meanwhile, on my phone, something far less impressive was quietly earning its keep. Photographs of receipts, read, categorised, coded to the right engagement, checked against the policy. No transformation programme. No strategy deck. An evening of admin reduced to the length of a coffee queue.
Three layers
Three layers, and each one needed something different.
The optimisation model is extraordinary on the normal day and helpless off it. The judgement stayed human, and will for a while yet. And the value I actually banked this week came from the least glamorous capability in the stack.
Most AI business cases I read are written for the first layer. Most of the available return, right now, is sitting in the third.
Nobody puts that bit in the brochure.
Which layer is your AI programme actually funded for?
If you would like to look at where the real return sits in your own AI plans, I would be glad to talk it through. You can read more about our services or get in touch.
This article was first shared as a post on LinkedIn.