Finance has a strange job. We are not the ones making most of the business decisions, yet we are the ones expected to explain them. All of them, including the ones we had no part in making.

I have been sitting with this observation for years now. It used to be manageable, as we were able to trace the decision back to its source in most cases in the past. Now, with AI generating more of the outputs that Finance is asked to explain, it has become a real problem worth naming.

The Gap Nobody Talks About

Here is what happens in almost every organisation that has started using AI in Finance. A model runs, produces the number that goes into a deck, and in the meeting, the CFO asks,

 "Where did this come from, and can we trust it?"

Everyone looks at Finance.

This used to be easier to handle. When logic lived in a spreadsheet, you could open the file and point to the cell. The model might’ve been wrong, or the assumption might’ve been off, but at least you could find it. With AI, logic is distributed across data, systems, and processes, and some decisions were made weeks earlier in the meetings Finance was never part of. The accountability stays. The traceability does not.

Finance is still being held responsible for outputs we had no hand in building. That is the gap!

And it grows every time another function deploys a model that embeds a financial definition without asking us what that definition should be.

What Integration Actually Looks Like

The sales team is running models that embed the way revenue is counted. The product team is defining what an active customer is. The operations team is deciding how capacity translates into cost, etc. These are financial definitions being set by non-Finance teams, and nobody is coordinating them. When the outputs of those models eventually reach Finance, we spend our time reconciling numbers that were never designed to agree.

I call this the missing integration layer. And fixing it is not about Finance taking over other functions. It is about being present earlier in the process, when definitions are still being set.

What do I mean by that?

For most of its history, FP&A sat at the end of the process. Business activity happened, data flowed in, and Finance made sense of it: we analysed, reported, and presented. That was the job, and honestly, we built entire careers on doing it well.

Integration means something different. It means Finance is embedded in the process before the output exists. When the sales team is designing its pipeline model, a Finance person is in the room asking: when this feeds into a revenue forecast, which definition of pipeline do we use? Or when the data team is building a forecasting model, Finance is part of the conversation about which signals to train it on and whether the outputs will match what the board actually tracks.

And the goal here is to own a financial logic layer that runs through all of them, not to own all of these models. Because Finance is the only function that sees the full picture across the whole business and its functions, that cross-functional view is the most useful before the model is trained, not after the output is in a deck.

So, the shift I am talking about is from explaining decisions to shaping them. That is a bigger role than FP&A has traditionally played in the past, and it requires a different skill set to pull off.

The Skill Set to Earn the Seat

You can walk into a data science planning meeting on good terms. But after that, you must earn the invitation. The room needs to see that Finance is adding value, not just attending.

Hard Skills

It all starts with AI Literacy. Just enough to understand what kind of tool you are working with. That Forecasting model learns from historical patterns. That Classification model sorts inputs into categories. And that Large Language model generates text based on statistical associations. These are not the same thing, and treating them as interchangeable leads to bad questions, misplaced trust, and missed issues with output. AI literacy is knowing enough to engage with the people who build these models.

Part of the literacy is Prompt Engineering, and I want to be specific about why it matters. The quality of what an AI tool returns is almost entirely determined by how you ask. A vague prompt gets a generic answer. A well-constructed prompt that includes context, constraints, and the right framing gets you something that you can actually use. Finance professionals who learn to write good prompts are getting real analytical work out of AI tools. And those who have not learned yet are mostly generating text that they have to rewrite later.

Basic Statistics and Predictive Analytics go hand in hand in practice. You need enough statistics knowledge to read an output critically, to know that correlation is not causation, that averages hide outliers, and that 80 per cent confidence is not the same as being right 80 per cent of the time. Paired with predictive analytics, that knowledge becomes more useful: the ability to ask which assumptions drive the output, where a small change in input produces a large change in results, and where you need to run scenarios before it goes to the CFO. These are not technical questions; these are finance questions, and they are exactly what gets Finance invited back into the room.

Data Fluency is simply the ability to trace a number back to its origin: where the data came from, what was filtered, which definitions were used, and what time period the model was trained on. In other words, it is about knowing enough to ask if the output reflects the actual business or a convenient simplification of it.

Automation Awareness and Basic Cybersecurity go hand in hand. AI only creates value when it is part of the workflow where decisions are made, not sitting in a separate dashboard that people check only when they remember. We always had plenty of scattered dashboards. Sometimes we even had dashboards about dashboards, so AI value is created when it flows through the whole workflow, simplifies it, automates it, and provides clean data for Finance professionals to focus on insights rather than manual tasks. And of course, one poorly considered data share into an unapproved platform can shut down an entire initiative for months. We work with sensitive data every day and the habit of care has to carry over into the AI Era as well.

Soft Skills

Let’s talk about the past for a second. For most of us, before AI, the career path in Finance was pretty straightforward. Build the technical skills first, like good modelling, reporting, analytics skills and of course, becoming an Excel Guru while doing that. That’s what would get you from Junior Analyst to Manager. Soft skills mattered too, but they were the second chapter, something you would develop when you were ready to move into the leadership role.

But it doesn’t work the same way anymore. The job of Integrated FP&A requires both at the same time, from the beginning. You can have strong technical skills and still fail to get into the rooms that matter if you cannot communicate effectively with people who do not speak “Finance”. You can be great with people and still not earn credibility in a data science conversation if you cannot engage with the substance. Hard skills get you invited, and soft skills determine whether you get invited back.

Storytelling has become more valuable than ever. AI gives teams more signals than they can ever react to. The bottleneck is no longer finding the insight. It is choosing which one matters and framing it in a way that actually changes a decision. A CFO does not need a list of interesting findings. They need the one thing that reframes how they are thinking about the business. Finance needs to be the function that makes that call and communicates it in a way that actually lands.

And what gives that judgment weight is Business Acumen. A model can find optimisation, but it can’t tell us what’s off limits this year, which trade-offs leadership won’t make, or what the Board has been worried about for the past two quarters. That knowledge leaves with Finance.

Cross-functional collaboration in this context is not what we normally think of it. It is a working method. Finance being in the right rooms earlier only happens if other functions want us there. That means showing up as a genuine partner, someone who helps them build something that will hold up, rather than someone who shows up to review and criticise their work after the fact.

Adaptability is about staying functional in a moving environment at a pace that is changing today. Finance teams that treat AI readiness as a one-time project will quickly fall behind, because the environment will keep changing. Those who build it into how they work, rather than treating it as a separate activity, will tend to stay relevant.

And then, there is a human part that AI doesn’t have, and may never have: Ethical Judgement, because not everything that counts can be counted. Finance has to be willing to ask tough questions: what are we actually rewarding with this output, what behaviour does this incentivise, and what will we think about this decision in a year? These are not analytical questions. No model will raise them on their own.

The Bigger Idea

And when these technical and human skills work together, Finance will stop being the function that explains what happens and become the function that helps determine it.

That is the real shift in our role. Not a bigger reporting mandate or a fancier tool. A different position in the organisation, earlier in the process, with more influence over the outcomes that Finance will eventually be asked to explain.

For FP&A leaders thinking about where to start: the most practical first step is to find out which models in your organisation are already generating outputs that carry financial consequences, and ask whether finance was part of building them. Not auditing them after the fact, but actually being part of the team that builds them. That conversation, with the right people, is usually where the integration work actually begins. Start there, before the next model produces an output.

I tell my team that we are no longer here to watch the game from the sidelines and report the score. We should be on the field while the strategy is being called. Integration is what makes this possible. The skills are what earn the right to be there.