The Finance Talent Shift Is Already Here: What AGOS Summit 2026 Added to What We Said in July
In "The Finance Talent Shift That Has to Happen Before Any AI Tool Lands," we said the sequence was backwards. Capability has to come before the AI tool, not after. At AGOS GBS Summit 2026, that argument showed up again, from a different angle. It wasn't about which job title needs redefining this time. It was about who gets to make the call once the job changes, and what actually has to be built before that person is ready to make it.
The decision moves down, not sideways
The instinct is to assume AI removes junior roles first, since that's where the routine work sits. What came up in the room instead: as AI absorbs more of the transactional load, decision-making shifts down, not up, from managers to the people actually closest to the work. That's a bigger ask than reskilling for a new task. It means someone earlier in their career now needs the judgement a decade on the job used to build, and needs it faster than the usual training pipeline can supply it. If that piece was about the shape of the job changing, this is about who's now trusted to run it. Most performance frameworks and promotion timelines still assume judgement is something you earn with tenure. That assumption doesn't hold anymore, and the organisations still hiring and reviewing against it are training people for a role that's already moved.
The mechanism, not just the mandate
The open question at the end of that piece was practical: has anyone started teaching the judgement calls before the software that surfaces them arrives. The summit answered with a specific mechanism, not just a policy line. Building that judgement takes three things at once: permission to question how something's always been done, space to actually experiment instead of squeezing it into evenings, and safety to be wrong without it costing someone. Take one away and the other two stop mattering. Permission without space is a policy nobody has time to use. Space without safety just moves the risk of being wrong onto the individual instead of the organisation. You can mandate AI adoption. You can't mandate the curiosity that makes it work, you can only build the room for it and hope people walk in.
What the job actually looks like now
On the ground, that shift already looks less like data entry and more like judgement work: less time reconciling numbers, more time asking what they mean. Some teams have gone further and started building their own small AI tools for specific problems instead of waiting for a central rollout to hand them one, a different kind of ownership than being trained on someone else's system. The goal one team described wasn't data literacy. It was data fluency, the difference between reading a dashboard and being able to argue with it. Literacy gets someone through the tool's onboarding. Fluency is what lets them catch the moment the model's confident answer is quietly wrong.
None of this changes the original argument. The sequence still has to reverse: capability before tool, not after. What the summit added was the how. Judgement moves to people who've never had to carry it before, and it gets built through permission, space and safety, not a training module. None of it sticks, either, unless leadership does the same work first — the clearest signal from the summit was that leaders who used the tools themselves outpaced leaders who handed the whole effort to a specialist team.
Same conclusion as before, sharper this time. The gap was never the technology. It's whether anyone below the CFO has been trusted, and trained, to actually use their own judgement before the tool starts asking them to.
If your team's still figuring out how to build this kind of judgment before the tool lands, that's the work we do with GBS leaders every day. Click the button below and we'll get in touch.
Comments