Before you buy an AI tool for your finance team, read this

There's a pattern I've seen play out more than once in the last two years. A CFO attends a vendor demo, sees something that looks impressive, and walks away thinking: we should have that. A budget gets allocated. A contract gets signed. Six months later, adoption is low, the team is frustrated, and nobody is quite sure what they actually bought.

The technology wasn't necessarily wrong. The sequence was.

Before your business spends a penny on AI for your finance function, there's a conversation that needs to happen first. It isn't with a vendor.

---

Start with the people doing the work

The most valuable intelligence in your finance function sits with the people closest to the data. The analysts building reports. The team managing links between systems. The people who know exactly where the process breaks down because they're the ones patching it back together every month.

These are the people who can tell you where the real pain is. Where human intervention is highest. Where time disappears into tasks that feel like they should be automatic but aren't. Where inconsistencies between systems get quietly absorbed into someone's Friday afternoon.

Before you talk to any vendor, talk to them. Find out where the friction is. Because the answer to that question determines which type of technology you actually need, and they are not all the same.

---

Understanding what you're actually buying

This is where most purchasing decisions go wrong. AI is not one thing. The term gets used to describe technologies that work in fundamentally different ways and solve fundamentally different problems. Here's what the landscape actually looks like:

RPA (Robotic Process Automation)

RPA uses software bots to mimic repetitive human actions. Copying data between systems. Moving numbers from one spreadsheet to another. Running the same reconciliation process that someone on your team currently does manually every week. It doesn't learn or think. It just executes, consistently and at scale.

If your team is spending significant time on repetitive, rules-based tasks such as data entry, system transfers, or routine reconciliations, RPA is often the right starting point. It's unglamorous, but it fixes real problems and frees up real time.

ML (Machine Learning)

Machine learning algorithms identify patterns in data and use them to make predictions or flag anomalies. Forecasting models that improve over time. Anomaly detection that spots unusual transactions. Demand-based scenario modelling that updates as new data comes in.

ML works well when you have clean, structured, historical data in sufficient volume. That last part matters. If your data is inconsistent, incomplete, or siloed across systems that don't talk to each other properly, ML will amplify those problems, not solve them. Getting your data foundations right before applying ML isn't optional. It's a prerequisite.

GenAI (Generative AI)

This is what most people mean when they say AI right now. The technology behind tools like ChatGPT, Claude, and Microsoft Copilot. It's exceptionally good at language: summarising variance commentary, drafting board narratives, explaining complex financial results in plain English, generating first drafts of reports and presentations.

Where it adds genuine value in finance is on top of a process that already works. Use it to accelerate the communication and narrative layer, the part that currently takes your team hours of writing and editing, so they have more time for the thinking underneath.

Where it doesn't add value is as a substitute for structured analysis, precise calculation, or clean data processes. It is a reasoning and language tool, not a finance system.

AGI (Artificial General Intelligence)

You may hear this term in a sales conversation. AGI refers to machine intelligence that can perform any intellectual task a human can, reasoning across domains and adapting to new problems without being trained on them. It does not exist in any commercial product available today. If a vendor uses this term to describe what they're selling, treat it as a red flag.

---

The right sequence

Armed with that understanding, the purchasing question becomes much clearer. Go back to what the people on the ground told you. Where is the pain?

If the answer is repetitive manual tasks and broken data flows between systems, start with RPA. Fix the pipes before you worry about what flows through them.

If the answer is that the data is reasonably clean but forecasting is slow, inconsistent, or too backward-looking, ML-based analytics tools are worth exploring. Focus on whether your data infrastructure can actually support them.

If the answer is that the analytical work is solid but the communication of it takes too long and the narrative layer is where time gets lost, GenAI tools layered on top can genuinely help.

The mistake businesses make is jumping to GenAI because it's visible and exciting, when the underlying problem is an RPA problem. Or investing in ML when the data it would need to learn from is a mess.

---

One question worth asking before any vendor conversation

Can our current tools already do what we're being shown, and does our team know how to use them?

Most enterprise finance platforms have capabilities that go largely unused. Before you buy something new, find out what you already have. The answer is often surprising.

Start by checking which version of your current software you're running. A simple upgrade, often included in your existing licence, can unlock meaningful new functionality. Finance platform vendors release enhancements regularly, and many businesses are sitting on capabilities they've never activated simply because nobody checked.

Beyond the upgrade, make sure someone in your team is actively tracking what's coming. The leading FP&A and finance platforms publish product roadmaps, run customer forums, and host user days specifically to share what's in development. These aren't just marketing events. They're genuinely useful signals about where the technology is heading and whether your current investment will keep pace with your needs. If nobody in your finance function is engaged with those communities, you're making procurement decisions in the dark.

AI can transform a finance function. The businesses that get the most from it are the ones that approached it in the right order, fixing the foundations, understanding the technology, and buying with a clear problem in mind rather than an impressive demo.