Start AI with the metric, not the painful process (Section)
The worst place to start an AI programme is your most painful process.
That runs against the usual advice, but Section's Michael Domanic makes the case well this week. Pain and size measure effort, not value. His own marketing team runs about 80 virtual events a year: visible, laborious, universally disliked logistics. Automating it would save roughly a fifth of one person's time and move no number that matters.
Leaders reach for painful processes because pain is visible. It has a queue, a complaint channel and a frustrated team. Value is quieter, and it rarely shouts in the meeting where the project list gets drawn up.
The second trap is scale. Leaders picture one agent taking over procurement or the month-end close. In practice a big process is a set of workflows, and a single workflow can need six agents handing work to each other. Start there and the project balloons, stalls and dies.
His alternative is disciplined. Name the handful of numbers a function exists to move, then find the work that drives them. For finance that points to procure-to-pay, because it touches free cash flow, days payable and operating margin at once. Then grade the steps on three tests: value, velocity to ship, and whether AI can do the step reliably today.
His best example is almost boring. A simple automation that checks every incoming invoice for a purchase order number and sends it straight back to the supplier if one is missing. A tiny build that removes a large share of the problems without human intervention.
I would put it more bluntly. Ugly processes attract attention. Valuable ones deserve it. They are rarely the same thing. The caveat worth holding: pain is not irrelevant, because a process everyone hates can be the one breaking your best people. But relief is a bonus, not a business case.
Which number would your next AI project actually move?
#Transformation #OperatingModel #AusBiz
Source: Section, Michael Domanic, 16 October 2026: https://lnkd.in/gGM4Ttne