24/06/2026
AI in Finance Is Built From the Bottom Up. Almost Everyone Starts at the Top.
Every finance leader I talk to wants to know how AI makes their team faster.
It's the wrong question.
The better one is: what slows finance down before AI ever enters the room? The answer is almost never the technology. It's two things — processes nobody designed to be understood, and the wrong people making the decisions.
I've walked into organisations running the best ERPs money can buy. Significant investment. Long implementations. Every module purchased. And half the team is still living in Excel. The software was bought. The process was never designed. When the process is unclear, the tool just executes that lack of clarity at scale.
Then there's the room. Leadership is excited, the board is aligned, the budget is approved — and the AP manager who processes invoices every day was never asked. Neither was the controller who owns month-end exceptions, or the analyst rebuilding the same broken report every week. These are the people who know where it breaks. Exclude them, and someone who doesn't live with the consequences decides what everyone else has to.
That gap is where implementations go to die. And it almost always starts in the same place: the back office.
# # The first stack: operational maturity
Look at where AI actually pays off, and it's not the glamorous top of the pyramid. It's the foundation.
**Invoice & AP processing** is the base. Fix this first — everything above depends on it.
**Reconciliation & close** runs on what AP produces. Clean inputs, clean close.
**Cash flow visibility** depends on accurate AP and AR data underneath it.
**Reporting & dashboards** are only ever as good as what feeds them.
**FP&A & forecasting** sits at the top. It looks strategic. It breaks the moment anything below it is messy.
Most CFOs start at the top — the forecasting, the dashboards, the board-ready strategy layer. That's the wrong order. The back office is unglamorous, but it's where AI returns capital first, because every layer above inherits its quality. Fix the foundation and everything above gets faster, cleaner, and more reliable on its own.
# # The second stack: control maturity
Here's the part nobody puts on the roadmap. The same bottom-up logic governs whether you can trust any of it — AI governance.
Governance isn't a policy document. A policy you can't enforce is just a PDF. And you can't audit what you never inventoried. It's six layers, and most teams skip the first five:
**1. AI Inventory** — You can't govern what you can't see. Run a shadow-AI pass, list every tool in use, tag each with an owner and a risk tier.
**2. Data Foundation** — Track where every input comes from. Screen for bias before it touches a model. Stale data is its own failure mode — monitor freshness.
**3. Data Security & Access** — Encryption, anonymisation, role-based access. Least privilege by default. Not everyone needs the model keys.
**4. Model Assurance** — Write a model card for everything in production: what it does, what it trained on, where it breaks. Then red-team it and watch for drift.
**5. Human Oversight** — Name who can override the model and who's accountable when it's wrong. In writing, before you need it.
**6. Compliance & Audit** — Regulatory mapping, alignment, audit trails. This is the layer everyone starts with. It only holds if the five below it exist.
# # The same mistake, twice
Both stacks fail for one reason: people start at the top, where it looks strategic, and skip the foundation that makes the top hold.
Operationally, that's chasing AI-powered forecasting on a back office that can't close cleanly.
In governance, it's writing a policy you can't enforce because nothing underneath it was ever built.
Most finance teams have an ERP. Almost none have designed the process. Most have a governance policy. Almost none have governance. The difference, in both cases, is the layers nobody sees — and the right people in the room to build them.
AI reaches its potential in finance only when the processes are defined and the right people are at the table. Once those two things are in place, the right tools become obvious.
Start at the bottom.
Which layer is your organisation actually working on right now — and which one did you start with? Curious to hear where others began. Get in touch with us for an assessment and whiteboard session to map it out for your organisation.
