AI in treasury isn’t a revolution. It’s an evolution

Let’s get something straight.

AI in treasury isn’t a future fantasy. It’s already happening. 

Quietly. Cautiously. But undeniably.

I’ve spoken to teams using it right now – not theorising or waiting for perfection – they’re just getting on with it.

Here are some real life examples drawn from my podcast interviews:

  1. NASDAQ – They’ve built bots that scrape their balance sheet and automate FX trades. What used to take three days now takes 20 minutes.
  2. Twoday Group – They’ve created rules-based tools to automate treasury ops, and in some cases, replace the need for a TMS altogether.
  3. Roche – Britta Döttger’s team are already using AI to manage short-term cash: how much to invest, for how long, and in which currency.
  4. ME! – Even I use it too. AI helps us crunch salary survey data faster, prep notes before interviews, and gives us first drafts of job ads. (Don’t worry – I still do the rewriting. I haven’t outsourced my humour.)

But here’s what matters most:

AI isn’t replacing treasury people. It’s refocusing them.

It’s taking the grunt work – reconciliations, reporting, spreadsheet hell – and freeing you up to focus on the stuff that actually moves the needle.

Like Dana Laidhold at NASDAQ put it:

“We don’t deal with data. We deal with analysis.”

Exactly.

This isn’t about becoming a coder. It’s about becoming a translator. The best treasurers I know aren’t building bots, but they DO understand what those bots can (and can’t) do.

They’re asking better questions. Spotting better opportunities. And leading better conversations with the CFO.

So if you’ve been sitting on the sidelines thinking “we’re not ready for this yet”… here’s my advice:

Don’t wait for perfection.

Start small. Automate one annoying task. Try a free tool. Ask around. Play with it.

Because the teams that are doing that now?

They’re the ones who’ll be leading treasury tomorrow.

Best regards,

Mike

P.S. Is your team using AI in any way? Cash forecasting, reconciliations, anomaly detection? Hit reply and tell me. I’m always up for hearing what’s worked… and what’s been a total flop!

Related articles