.png)
Every finance function has an AI story now. A copilot that summarises variances. A dashboard that refreshes automatically. A forecast that, technically, runs itself.
But ask that same function how long it takes to run a downside scenario when a key assumption changes — and the answer is still measured in days. Not because the AI tool is not capable. Because the planning process underneath it was never designed to move at the speed the tool now makes possible.
That gap — between AI being present in FP&A and AI actually changing the speed of planning decisions — is where the real story of AI in finance is playing out in 2026. Only 12% of finance organisations have deployed machine learning in FP&A forecasting at full scale — in active production, not piloting or testing, according to Bain's 2026 CFO Survey. The rest are either still experimenting or have deployed AI without redesigning the work around it.
The organisations in that 12% are doing something specific that the others are not. This post explains what it is.
The headline numbers look impressive. AI adoption in FP&A surged from 6% to a 41% increase in usage in 2025, marking a pivotal shift in finance capabilities, according to FP&A Trends Research Paper 2025.
The CFO Connect State of AI in Finance 2026 report finds that 56% of finance leaders now use AI — double the adoption rate seen in 2023. Deloitte's Q4 2025 CFO Signals survey found that 87% of CFOs at large companies say AI will be extremely or very important to finance operations in 2026.
But beneath those figures sits a sharper truth.
The satisfaction gap between scaled deployments and pilots is the real story. And it points directly to what separates the 12% getting results from everyone else.
Bain's research found that in many cases where AI has been deployed, finance teams run AI-generated forecasts alongside existing bottom-up planning cycles — two processes running in parallel, neither fully trusted, with the expected benefits largely unrealised. The AI was deployed. The work was not redesigned. Bain calls this "workflow debt." It is the single most accurate description of where most AI in FP&A initiatives currently stall.
The highest-value AI application in FP&A is predictive forecasting — using machine learning to analyse historical patterns, identify leading indicators, and generate forward projections that update continuously as new data arrives rather than being refreshed once per planning cycle.
Budget cycle time typically falls by 30–40% when AI tools are integrated into planning, according to EagleRock's 2026 survey. And once AI lowers the cost of re-forecasting, finance teams can run orders-of-magnitude more scenarios than before — instead of three scenarios per cycle, teams are running 50 or more to understand the full distribution of outcomes.
EPM platforms including Anaplan, Jedox, and OneStream — Keansa's core delivery partners — have embedded predictive forecasting capabilities directly into their planning environments. Anaplan's PlanIQ module allows users to incorporate internal and external data drivers and find correlations that would not otherwise be identified through manual analysis. Jedox's AI-augmented forecasting identifies patterns in actuals to generate accurate projections without requiring data science expertise.
The critical point — and one that vendor demonstrations consistently underemphasise — is that these models are only as reliable as the data feeding them. 35% of CFOs identify data trust as their top barrier to AI ROI, and only 10% fully trust their enterprise data. This is not a platform problem. It is a data foundation problem. And it is why we consistently start Keansa's FP&A implementations with a data readiness assessment before any AI capability is configured.
Scenario planning is where AI creates the most immediate and visible value for FP&A teams. The traditional approach — building separate models for base, upside, and downside cases — produces outputs that are outdated by the time they reach leadership, and takes long enough that scenario analysis becomes a quarterly exercise rather than a continuous capability.
AI improves scenario planning by generating fast, transparent permutations across price, volume, mix, headcount, and macro variables with side-by-side impact views on P&L, balance sheet, and cash flow. In an AI-enabled planning environment built on a driver-based model, running a new scenario is a matter of adjusting driver inputs rather than rebuilding the model from scratch.
This is a principle we examined in detail in our post on driver-based planning and in our companion post on scenario planning at scale. The combination of driver-based model architecture and AI-assisted scenario generation is what transforms scenario planning from a scheduled deliverable into a live decision-support capability.
One of the most time-consuming activities in FP&A is variance commentary — explaining why actuals differed from plan, which lines drove the variance, and what it signals for the outlook. For a business with multiple cost centres and business units, this can consume multiple days per month of analyst time.
AI and automation are enabling FP&A professionals to automate data ingestion, budget analysis, and narrative generation — enabling finance teams to focus on insight and action rather than data management. Natural language processing models can read a variance dataset, identify the primary drivers, and produce a first-draft management commentary in minutes rather than days.
The value is not that AI replaces the analyst's judgment about what the variance means. It is that AI eliminates the mechanical work of identifying and structuring what happened — freeing the analyst to focus on what to do about it. This is the most direct route to the outcome IBM's research describes: finance teams making quick decisions based on real-time data-driven insights and forecasts, rather than spending excessive time managing data and analysing figures.
Error and anomaly detection is already one of the top three AI use cases adopted across finance functions, according to Gartner's 2025 survey. AI anomaly detection runs continuously against financial data, identifying patterns that deviate from historical norms — and surfacing them before they compound into significant variances.
This directly addresses one of the most persistent gaps in planning performance. As we explored in our post on 7 planning KPIs every CFO should track, only 13% of organisations identify performance issues before they hit the financials. AI anomaly detection is the operational mechanism for moving into that 13% — catching drift in the data before it appears in the monthly close package.
It is a legitimate question. Finance functions have been promised transformation by ERP implementations, BI dashboards, and EPM platforms. Each delivered productivity gains. None of them fundamentally changed how quickly finance functions could move from a new signal to a new decision.
The distinction with AI is not speed of execution. It is the nature of what is being automated. Traditional automation follows rules and halts when an exception appears. AI in FP&A — specifically when embedded in a driver-based planning model on a connected EPM platform — can identify which drivers are moving, calculate the downstream financial impact across functions, generate the scenario comparison, and surface the decision that needs to be made, without a human manually initiating each step.
That is not automation of the same work. It is a different category of capability.
But — and this must be stated clearly — it only works when the underlying process has been redesigned. Finance teams that run AI-generated forecasts alongside existing bottom-up planning cycles end up with two processes in parallel, neither fully trusted. The AI was deployed. The work was not redesigned. The sequence matters: foundation first, process redesign second, AI third.
A head of FP&A at a mid-market manufacturing business learns that a critical input supplier has raised prices by 11%. Here is what happens in each environment:
This is what Keansa builds toward in every FP&A implementation — not AI as an add-on to an unchanged process, but as a compounding capability on top of a properly designed planning foundation. The same principle applies across our S&OP and Supply Chain engagements, where operational planning and financial planning need to move at the same speed.
Only about one in five organisations currently has a mature governance model for autonomous AI in finance, according to Deloitte's 2026 State of AI in the Enterprise research. This governance gap is the primary reason analysts expect a meaningful share of AI in FP&A initiatives to stall in 2026 — not because the technology fails, but because the guardrails were not established before go-live.
As EY's Aaron Shifrin, Americas Business Planning Reporting and Analytics Solution Leader, states: "The true edge in AI isn't just smarter models — it's smarter data. Data quality is what powers meaningful intelligence. That's why finance leaders must play a central role in shaping it."
For CFOs and finance transformation leaders evaluating AI in FP&A, four governance questions should be answered before any platform capability is activated:
These are not reasons to avoid AI in FP&A. They are the conditions that separate finance functions that adopt it successfully from those that stall.
The CFOs seeing measurable AI ROI chose embedded AI within their planning platforms rather than standalone tools requiring separate integration. They started with high-impact, low-complexity use cases: automated variance analysis, anomaly detection, baseline forecast generation. And they measured AI ROI against specific operational metrics — cycle time reduction, error elimination, analyst time freed for strategic work — not abstract transformation goals.
Three consistent patterns emerge across every case where AI in FP&A is generating real returns.
AI adoption in finance has reached 56% — doubled since 2023. But finance still ranks last among all business functions in AI deployment. And 45% of finance teams remain in limited pilot mode, with only 17% using AI in core workflows.
The gap is not ambition. 68% of CFOs say they have been slow to adopt because they do not know where to start. The gap is sequence.
The CFOs who are getting AI results are not the ones who bought the most advanced AI. They are the ones who built the infrastructure that lets any AI — even basic embedded features — produce trustworthy, actionable insights.
Finance functions that have built driver-based planning models, resolved their data quality challenges, and established the planning process KPIs to measure whether their process is improving have already done the hardest part. The AI capabilities available in modern EPM platforms are ready to compound that foundation.
For everyone else, the foundation is still the first step. And no amount of AI capability changes that.
The competitive divide in FP&A is not forming between organisations that have AI and those that do not. It is forming between organisations whose planning infrastructure is strong enough for AI to make a difference — and those where it simply makes the existing gaps more expensive to maintain.
Related Resources
AI will not fix a broken planning foundation. But on top of a strong one, it changes what finance is capable of.
Book a Free Planning Assessment →.png){kind=small}
.png){kind=small}
.png){kind=small}