Finance: Powerful Decision Engine, Not Passive Scorekeeper

Finance has spent most of its life perfecting the art of looking backwards, but AI is forcing a sharper question: what if the finance function exists to decide what happens next, not just to report what already happened? We make the case that using AI to close faster is only a small win, and often a distraction from the bigger prize: faster, better decisions on pricing, capital allocation, working capital, risk signals, and scenario planning.

This episode explores how finance can move from scorekeeper to decision engine.

TLDR / At a Glance

• Decision speed and quality
• Sense, predict, judge, act
• Trusted finance data
• Human accountability
• AI Auditability
• Forecast accuracy measurement

We break down a simple, practical model for an AI-enabled finance decision engine: sense, predict, judge, act. AI strengthens sensing and prediction by turning live signals into analysis at speed, but we are clear about the boundary: judgement stays human, because accountability cannot be outsourced to a model. That shift changes the skills finance needs, moving the centre of gravity from preparation towards challenge, narrative, and commercial decision-making.

We also tackle the hard constraints that stop teams from getting measurable value from AI in finance and FP&A. Trusted data is the bottleneck, not the model, and poor definitions create “confident errors”. We explain how to build a minimum trusted data foundation for a specific decision, then scale from there. Finally, we cover why controls, audit evidence, and decision-quality measurement are not red tape but the mechanisms that create trust and let AI move into material work.

If you want practical guidance on how CFOs and finance leaders can redesign the loop, choose the right decisions to rebuild, and measure what matters, listen now. Subscribe, share with a finance leader who’s stuck in pilot mode, and leave a review with the one decision you would want 10% faster or sharper.

AI creates the possibility, but leadership design turns finance transformation into measurable business value.

Transcript

Copy Transcript Copied

Finance As A Decision Engine

SPEAKER_00 0:00

Finance powerful decision engine, not passive scorekeeper. TLDR slash at a glance. Finance's value is shifting from producing information to using it. The biggest reported gains are in decision speed and quality. Wide adoption, thin maturity, most have not redesigned around AI. A decision engine runs a loop. Sense, predict, judge, act. AI prepares the decision. Trusted dot is the fuel and the main bottleneck. Controls and measurement make the loop trustworthy, not slower. For most of its history, finance has been judged on how accurately and quickly it reports what already happened. The close, the management accounts, the board pack, all of it looks backward, and the functions craft has been to make that backward look reliable. AI is now being dropped into that world, and the instinct is to use it to do the same job faster, closing in days rather than weeks, and producing reports in minutes. That instinct captures a fraction of what is actually changing. This article argues that finance is moving from a function that records what happened to one that helps decide what happens next, and that the shift is real but conditional. It looks at where the value is genuinely appearing, why adding AI to existing work falls short, and what it takes to turn finance into a decision engine rather than a faster reporting line.

The Scorekeeper Instinct

SPEAKER_00 1:33

The Scorekeeper's Instinct. The appeal of using AI to speed up reporting is obvious. Closing faster and producing accounts with less effort is a real benefit, and it is easy to measure, which makes it the natural first move. The limitation is that it leaves the fundamental job unchanged. Finance still tells the organization what happened only sooner. The clock moves, the role does not. That instinct is now colliding with a much higher level of ambition. Deloitte found that 87% of North American CFOs expect AI to be extremely or very important to finance operations in 2026, which is not the language of a function looking to shave a few days off the close. When finance leaders talk about AI mattering that much, they are describing a change in what finance is for, not just how fast it does its existing work. The reporting upgrade is real, but it is the smallest part of the story.

From Information To Action

SPEAKER_00 2:34

From producing information to using it. The deeper shift is in where finance creates value. A reporting function produces information and hands it to others to act on. A decision function uses that information to shape the choices the business actually makes. How to price, where to allocate capital, when to move on working capital, how to respond to a covenant breach or a cost signal. AI is what makes the second posture practical at speed because it can turn raw signal into analysis fast enough to inform a live decision rather than a retrospective one. The evidence already points this way. In KPMG's 2026 Global Finance Survey, the largest reported improvements from AI were not in cost or headcount, but in decision making, with around seven in ten finance teams citing better decision speed and quality. That is a telling result. It suggests AI's most valuable current role in finance is not automating the production of information, but sharpening the decisions that information should drive. Put simply, the prize is not a faster scoreboard. It is a function that helps win the game. That reframing changes almost everything downstream, from which use cases matter to how value should be measured, and it is the reason the reporting speed framing is too small to be useful. Wide adoption, thin maturity. If the opportunity is this large, why are results so uneven? The answer is that most organizations have adopted AI without redesigning finance around it, and adoption without redesign produces pilots rather than performance. Tools get bought, copilots get switched on, and the underlying workflow, with its manual reconciliations and backward-looking cadence, stays exactly as it was. The gap shows up starkly in the data. Deloitte found that while 63% of finance departments said they had fully deployed and were actively using AI, only 21% said those investments had delivered clear measurable value. The distance between those two numbers is the distance between using AI and building finance around it. Closing it is not a technology problem, it is a design problem, and design is a leadership responsibility.

Sense Predict Judge Act Loop

SPEAKER_00 4:59

The decision engine runs a loop. It helps to be concrete about what a decision engine actually does. A reporting function produces an output. A decision engine runs a loop, and I find it useful to describe that loop in four stages sense, predict, judge, act. The function becomes an engine when it owns the whole loop, not just the report that used to sit at the end of it. The first two stages are where AI does most of its work. Sensing is detection, spotting anomalies, explaining variances, picking up operational and market signals as they appear. Prediction is the forward look, rolling forecasts, cash positioning, scenario ranges. In the language of the loop, AI prepares the decision, assembling and shaping the evidence faster and more continuously than a human team working from periodic extracts ever could. The third stage is the one that stays human. Judgment is where someone weighs the trade-offs, challenges the recommendation, considers what the model cannot see, and takes accountability for the call. This is deliberately a handoff, not a continuation. The value of the first two stages is realized only if a capable human owns the third. The fourth stage, action, is where the decision becomes a move the business makes, and where the loop closes by feeding the result back as new signal. Around all four stages sits a control layer that makes the loop visible, who owns each output, where human approval is mandatory, how inputs and decisions are documented, and how exceptions escalate. The single line worth remembering is this AI prepares the decision, finance owns the judgment, and controls make both visible. A function organized that way is a decision engine. A function that simply runs AI over its old reporting process is not.

Trusted Data As The Bottleneck

SPEAKER_00 6:58

The loop runs on trusted data. A loop is only as good as what flows through it, and this is where most finance functions are genuinely stuck. A decision engine needs data that is clean, connected, and consistently defined, because a recommendation built on fragmented sources and conflicting definitions is worse than no recommendation. It is a confident error. The bottleneck is rarely the model now, it is the data the model is fed. The scale of the problem is well documented. AFP found that 61% of FP and A teams cite unreliable data as a major obstacle to making technology work for them. The practical implication is not to wait for a perfect enterprise data program before starting, which can take years. It is to build a minimum trusted data foundation for the specific decision the function is trying to improve, harmonizing the financial, operational, and external signals that decision actually needs and expanding from there.

Human Judgement And Capability Gaps

SPEAKER_00 8:02

The judgment stays human. The loop deliberately keeps a human at its center, and that is not sentimentality. Material financial decisions carry accountability that cannot be delegated to a model, whether to the board, to auditors, or to regulators. AI can prepare the case with more speed and breadth than before, but someone still has to own the judgment, challenge the output, and answer for the result. The function center of gravity therefore moves from preparation toward judgment, challenge, and narrative. The trouble is that finance is not yet ready for that shift in emphasis. AICPA and CIMA found that 88% of finance leaders expect AI to be the most transformative trend in their field over the next year or two, while only 8% feel very well prepared for it. That gap is a capability problem, and it will not be closed by tools alone. It calls for deliberate investment in the data, fluency, and commercial judgment that the judge stage of the loop now demands.

Controls Measurement And Audit Evidence

SPEAKER_00 9:09

Controls and scoring make the loop trustworthy. There is a tempting belief that governance slows finance down, that controls are the price you pay for safety at the expense of speed. The evidence suggests the opposite. The functions that can demonstrate how an AI-supported decision was reached are the ones that trust the loop enough to use it for things that matter, and that trust is what allows AI to move from the margins into material work. KPMG's 2026 study makes the point with unusual clarity. Organizations that could produce AI audit evidence efficiently reported far better outcomes than those that could not, including markedly higher error reduction at 33% against 6%, and far greater confidence in scaling. Auditability, in other words, is not the break on value. It is part of how value is realized, because it is what turns a promising model into a control the business can rely on. The same logic applies to measurement, where finance has a curious blind spot. AFP found that only 14% of finance teams formally track forecast accuracy, which means most are not scoring the very capability AI is meant to improve. An engine you do not measure is one you cannot tune. A decision engine has to be judged on the quality of its decisions, through forecast accuracy, exception rates, cycle times, and the outcomes of the actions it informs, not on the volume of reports it produces.

CFO Starting Questions And Closing

SPEAKER_00 10:45

What this means for leaders. The shift from scorekeeper to decision engine is not automatic, and it does not arrive with the software. It is a choice to redesign the loop from signal to action, to put trusted data underneath it, to keep human judgment at its center, and to wrap it in controls and measurement that make it trustworthy. Organizations that make that choice are seeing the payoff. Lucid, working with PWC and AWS, rebuilt its finance function around AI and cut a forecasting cycle from weeks to under a minute, then used finance as the launch point for decision support across the wider enterprise. For a CFO, the practical starting point is not a tool but a decision. Which finance decisions would matter most if they were 10% faster or sharper? Where do current forecasts fail? And are they even being scored? What data definitions would break an AI recommendation today? And where must human judgment remain mandatory and why? Those questions point to a small number of decisions worth rebuilding the loop around, rather than a sprawl of pilots that never reach the income statement. The technology now makes a genuine decision engine possible in a way it was not a few years ago, but the technology is the easy part, broadly available to every finance function. What separates the leaders is the management work of redesigning how finance senses, predicts, judges, and acts, and of holding the controls and measurement that make the whole loop trustworthy. Technology creates the possibility. Management turns it into value. This concludes the article. You can also read this article on my LinkedIn page where I share regular insights on AI, strategy, and emerging technologies.