Upskilling Reality

    AI Came for FP&A in 2026 — Adoption Tripled, but the Core Work Is Still Manual

    July 29, 2026·7 min read

    TL;DR

    Finance has crossed the line from experimenting with AI to depending on it: 56% of finance professionals now use AI, up from 17% in 2023, and 90% of CFOs have automated some part of their workflow. But only 17% use AI inside core finance workflows — the modeling, close, and variance analysis that actually pays. Fully embedded AI is up 91% year over year, yet data readiness is the #1 blocker to ROI and 68% of CFOs don't know where to start. Here's the data, what an AI finance copilot actually is, and how to build one in a single sitting.

    Summarize with AIChatGPTClaude

    The year AI stopped being a pilot in finance

    For a few years, "AI in finance" meant a proof-of-concept nobody trusted with the actuals. In 2026 that changed. 56% of finance professionals now use AI — up from 17% in 2023. In three years the technology went from a curiosity a minority touched to a tool most of the function reaches for. And it isn't just individual contributors experimenting: 90% of CFOs report automating some part of their workflow.

    But adoption at the surface hides where the work still is. Of that 56%, only 17% use AI inside core finance workflows — the modeling, the close, the variance analysis that leadership actually reads. Most AI use in finance is still email drafts and meeting notes. The hard, high-value FP&A work is where the copilot hasn't arrived yet — which is exactly why it's the skill worth building now.

    The data

    Signal Figure Source
    Finance professionals using AI (up from 17% in 2023) 56% CFO Connect
    Of those, share using AI in core finance workflows 17% CFO Connect
    CFOs who have automated some part of their workflow 90% CFO Connect
    YoY growth in fully embedded AI in finance +91% Consero
    CFOs who don't know where to start with AI 68% CFO Connect

    Two numbers tell the whole story together. Fully embedded AI in finance is up 91% year over year — the teams that got past the pilot are compounding fast. And 68% of CFOs still don't know where to start — the gap between the leaders and everyone else is widening, not closing.

    💡The differentiator in 2026 isn't whether your team uses AI — most already do. It's whether AI has reached the *core workflow*: the variance commentary, the reforecast, the board deck narrative. That's the 17% who've actually moved the needle, and the 91% growth is coming from teams crossing exactly that line.

    What an AI finance copilot actually is

    Strip away the hype and an AI finance copilot is concrete. It's four repeatable capabilities pointed at the work you already do every month:

    1. Analyze actuals in plain English. Instead of a pivot table safari, you ask: "Why is Q2 opex $340K over plan?" and get an answer traced back to the line items that moved — headcount, a vendor true-up, a timing shift.
    2. Draft variance commentary. The copilot writes the first draft of the "what happened and why" narrative that goes into the monthly pack — the part that eats an analyst's Friday. You edit; you don't start from a blank page.
    3. Wire to your data. It connects to your actuals — the GL export, the FP&A model, the budget file — so the answers reference your numbers, not a generic template.
    4. A review step. Nothing reaches leadership unread. The copilot proposes; a human checks the logic and the numbers before it ships. The analyst moves from typing to judging.

    This is why management reporting and variance analysis are the fastest-paying AI use case in finance, with a 3–6 month payback. It's high-effort, high-frequency, and pattern-heavy — precisely the shape of work AI is good at, and precisely the work that consumes the most FP&A hours today.

    How to get started

    You don't need an engineering team or a platform migration. With Claude, you can stand up the loop yourself:

    1. Connect to actuals safely. Start with a read-only export — a CSV of the GL or the actuals-vs-budget tab. No live system writes, no credentials in a prompt. The copilot reads your numbers; it doesn't touch the source.
    2. Ask questions with the working shown. Don't accept a bare number. Prompt for the drivers behind every variance so you can check the reasoning: which accounts moved, by how much, and against what baseline.
    3. Auto-draft the variance commentary. Have Claude turn the driver analysis into the narrative paragraph your monthly pack needs — in your team's voice, at your team's altitude.
    4. Add a review step before it reaches leadership. Build the human check into the workflow, not around it. The analyst verifies the math and the story; the copilot never publishes on its own.
    5. Get your data ready first. Consistent account names, a clean actuals-vs-budget structure, one source of truth for the current forecast. This is unglamorous and it is the whole game (see the warning below).
    6. Iterate on one report, then expand. Nail the monthly opex variance before you touch revenue, cash, or the board deck. One working artifact beats ten half-built ones.
    ⚠️**Data readiness is the #1 blocker to AI ROI in finance** — the reason so many pilots stall. If your actuals live in inconsistent exports, your model has hard-coded overrides, and "the current forecast" means three conflicting files, no copilot will save you. Fix the inputs first. A clean, well-labeled actuals-vs-budget file is worth more than any model choice.
    ℹ️68% of CFOs don't know where to start — which means "start small and specific" is a genuine advantage, not a compromise. One report, wired to real data, with a human review step, is further than most of the market has gotten.

    Build it live

    The gap between "AI is transforming FP&A" and "I built an AI finance copilot" is one focused session. In Dexity's AI for Finance & FP&A workshop, you build your AI finance copilot live in 90 minutes — guided by a practicing FP&A analyst — connecting to actuals safely, asking questions with the working shown, and auto-drafting variance commentary with a review step before it reaches leadership. You walk away with it working, not with notes.


    Sources: CFO Connect — State of AI in Finance 2026 (56% AI adoption up from 17% in 2023; 17% in core workflows; 90% of CFOs automating; data readiness as #1 blocker and 68% unsure where to start; management reporting/variance analysis as fastest-paying use case at 3–6 month payback); Consero — 2026 CFO Survey (fully embedded AI in finance up 91% year over year).

    Anmol Gulwani

    Anmol Gulwani

    Dexity

    Connect on LinkedIn
    Questions or suggestions?hello@dexity.com