The CFO walks out of the board meeting and calls you within the hour. AI came up three times, apparently. The board wants to know what the finance function is doing about it. Your CFO, fresh from that conversation, isn’t asking whether you have a view. They’re asking what you’re going to show them, and by when. If your organisation is exploring AI in finance, this is a conversation you’ll want to be ready for.
You know the process and where the exceptions live, which reports are still stitched together in spreadsheets, which data fields nobody trusts because three different teams populate them three different ways. You also know that whatever you say next will either buy you room to do this properly or commit you to a timeline built on hope.
This is the moment most credit and collections leaders get wrong because they respond to pressure with reassurance instead of with a plan.
The problem nobody wants to say out loud
Most credit and collections functions are being asked to apply AI to processes that are not clean, running on data that is not ready. Peter McKenzie made this point well at the O2C Transformation Forum: layer AI onto a broken process and you don’t fix the process; you only accelerate it. It’s a lesson every organisation exploring AI in finance should understand.
You get the same errors, exceptions and manual rework – only faster and now with a licence fee attached. Speed doesn’t fix a bad foundation. It exposes it more expensively.
Lead with the conclusion – the Minto Pyramid
When it comes to AI in finance, knowing this isn’t enough. You have to say it, upwards, in a way that lands. This is where most finance professionals sabotage themselves. Trained to build a case from evidence to conclusion, they walk into the CFO conversation and start with data quality, then process maturity, then systems architecture, arriving at their actual point somewhere around minute eight. By which time the CFO has stopped listening and started forming their own view of what you’re capable of delivering.
Peter McKenzie taught the room a better method: the Minto Pyramid. Lead with your conclusion. Follow it with your key arguments. Support those with evidence, only as far as the audience needs it. Senior people don’t have the patience for a slow build. They want the headline first, and they will ask for the detail if they want it.
Power isn’t given to you in this conversation
There’s a second layer to this and it’s about where influence comes from. Peter McKenzie’s framing at the same session was blunt:
- power is 20% granted and 80% taken.
Nobody is going to hand you the authority to shape the AI conversation. You take it or someone else in the room does.
The credit leader who treats the CFO’s call as pressure to be absorbed and managed is ceding ground they never had to give up. The credit leader who treats it as an opening to demonstrate judgement is doing something else entirely – they’re building the kind of credibility that gets them invited into the next strategic conversation.
Agree to everything and deliver a mess in six months and you have spent trust you won’t get back easily. Come back with a clear-eyed assessment and a narrow, credible plan and you have just made yourself the person the CFO trusts with the next hard question.
What to do before you take the call
So, before you take that call or before you return it, do these three things to build a credible approach to AI in finance:
- Assess your process readiness honestly and write it down. Not a vague sense that “data quality is a challenge” – specifics. Which fields are reliable and which aren’t, and why. Where the exceptions concentrate. This isn’t a research project; it’s two hours with the people who touch the process daily, and it gives you the evidence layer for the conversation you’re about to have.
- Find the one-use case that could work now, with the data you already have, not the data you wish you had. It’s probably narrower than what the board is imagining – payment prediction on your cleanest segment, dispute categorisation on a well-structured subset, cash application on accounts with reliable remittance data. Small and credible beats broad and speculative every time this gets reviewed in six months.
- Frame the ask so that you’re setting the agenda, not just responding to one. Don’t go back with “here’s why we can’t do this yet.” Go back with “here’s what we’re doing, here’s what it will prove, and here’s what I need from you to do it properly.” The first version manages expectations downward. The second positions you as the person driving this, which is the position you want to be in.
The CFO doesn’t need you to be excited about AI. They need you to be right about it. Successful AI in finance starts with clean data, realistic use cases and the confidence to challenge unrealistic expectations.