A Short to Long-Term Plan for AI adoption in Finance for CFOs

AI adoption in finance has reached an inflection point. A majority of finance professionals are now experimenting with AI in some form, yet most usage stays shallow: summarising reports, drafting emails, running simple analysis. The deeper opportunity, compressing month-end close, enabling real-time forecasting, and repositioning finance as a strategic partner, is still out of reach for most teams. The question is no longer whether to act. It is how to act in sequence.

This article draws on CFO Connect's State of AI in Finance 2026 report to map what is holding teams back, what the leaders are doing differently, and how to structure a credible 30/90/365-day plan to move from experimentation to embedded advantage.

Why AI adoption in finance is still shallow (and what separates leaders from laggards)

A General Atlantic poll found 45% of finance teams "experimenting, with limited pilots underway" and only 17% actively using AI in their finance workflows. Looking at AI in finance industry adoption today, a clear divide has emerged between two groups.

The tinkerers use ChatGPT or Copilot for everyday efficiency. They rely on manual copy/paste and lack the data connectors or governed infrastructure that would let AI go deeper. The integrators, a much smaller group, have built unified finance data hubs, connected AI directly to their ERPs, and deployed well-functioning internal agents. For them, AI nearly automates entire processes and delivers insights that were previously impossible.

"Most adoption remains focused on administrative workflows," says Pauline Babel, CFO at Spendesk. "The opportunity now is to extend AI deeper into core finance functions: accounting, forecasting, compliance, and strategic decision-making."

Three barriers explain why most teams stay stuck at tinkerer level.

Data fragmentation. Finance data is scattered across ERPs, billing systems, CRMs, contract repositories, and spreadsheets. These silos block deep automation. Phil Sharp, Interim CEO and CMO at Subscript, captures the frustration precisely: "Nine out of ten finance leaders think their data is an absolute hot mess. Everyone whispers it like they're the only one facing this problem, but I hear it five times a day. Messy data is the norm, not the exception. The real question becomes: how do you operate when data will always be somewhat messy?"

No protected time. "The biggest challenge is time and focus," says Ido Peled, Head of Finance Data and Technology at Adyen. "Finance professionals are busy with closing and reporting. To innovate, we must create protected time for experimentation." Monthly close cycles absorb whatever spare capacity might otherwise go toward AI setup.

No playbook. 68% of CFOs say they have been slow to adopt AI because they don't know where to start. Jessica Pillow, an HR leader overseeing 7,000+ team members across 120+ countries, captures the mood: "There's definitely no playbook. We're all sitting around wondering how we have this AI, and it's not going anywhere. What we're doing at our company is treating it like a product: constantly testing, iterating, and making small, incremental movements."

What finance automation is delivering for teams today

The gap between tinkerers and integrators is widening because the gains from genuine finance automation are compounding quickly.

Speed, parallel processes, and real-time decisions

Classic finance operations are sequential: close the books, hand off to FP&A, begin analysis. AI breaks that waterfall. "Finance teams are drowning in operational work," says Axel Demazy, CEO of Spendesk, "chasing receipts, closing books, and reconciling transactions. Accounting closes, then hands it off to FP&A, who spend a week figuring out what just happened."

With AI, those flows run in parallel. "AI moves finance from backward-looking reporting to augmented decision-making," Demazy continues. "Real-time data, cloud ERPs, and AI compress month-end into a continuous close. Variance analysis goes live. Forecasts update as signals change. Accounting and FP&A run as a single, always-on cycle."

The close itself compresses dramatically. Babel notes: "Before, closing books in D+1 or D+2 was reserved for large corporations. AI is now unlocking so much automation that it becomes possible for companies that cannot afford heavy and expensive tooling. This is the case at Spendesk, where we can cut the month-end close down to a couple of days."

Yubo She, Head of Technical Accounting at OpenAI, describes the role shift that follows: "AI accelerates prep work, letting humans focus on strategy and oversight. And preparers become reviewers and exception managers."

Cost structure, job satisfaction, and new strategic capabilities

Dan Zhang, CFO at ClickUp, offers a striking projection on the finance cost structure: "Today, 80% of finance costs are payroll. Within three years, a growing share will come from AI tooling and model usage." That shift does not mean redundancies. It means finance headcount becomes more specialised, focusing on analysis, business partnering, and strategic work, while AI absorbs high-volume, low-value tasks and scales cheaply with usage.

Job satisfaction improves alongside it. Zhang adds: "A lot of the younger generation of workers want their company to give them the tools so they don't need to do all those manual, soul-crushing jobs. Those classic entry-level tasks. Employee satisfaction is a great leading indicator for if this AI tool is going to be well adopted in your organisation." Rebeca Bichachi, Product Marketing Director at Oracle NetSuite, agrees: "Top talent seeks meaningful work, not endless manual tasks. AI helps unlock that shift."

Business partnering improves too. Ido Peled describes how Adyen democratised knowledge access: "We built a knowledge hub where we put all the documentation, videos, and other sources of information. We allow them to ask the tool a question and get an immediate answer. So that not only frees up time for us, but it also lets them get instant answers no matter where they are in the world."

Perhaps the biggest shift is in entirely new capabilities. Gabriel Hubert, CEO and co-founder of Dust, frames it: "AI increases productivity in three ways: doing things faster, doing them better, and doing things you couldn't before." Zhang gives a concrete example of that third category: "One use case is the AI BDR, engaging the long tail of small customers we've never touched before because the unit economics doesn't make sense. That's a zero to one unlock; a whole new capability. That's where I would put resources."

As Julien Lafouge, CFO at Photoroom, puts it: "CFOs can collaborate with the rest of their teams in thinking about productivity: how much faster can we make great decisions as a company? If it takes three months to make a strategic decision on a new product, or to enter a new market, or to change a part of our infrastructure, can we do that in two months instead?"

Six real-world examples of AI integration in finance

These case studies, drawn from the State of AI in Finance 2026, show what actual AI integration looks like across high-growth and enterprise teams.

Spendesk: spend planning and continuous close

Spendesk uses automation to speed reconciliation and coding throughout the month rather than at month-end. Live budgets let teams track spend against plan as it happens. The result is a finance function that operates on current data at all times.

"This isn't just about efficiency," says Babel. "It's about repositioning finance from operational to strategic. Moving from reporting the past to shaping the future." Demazy frames the model shift: "The new operating model for finance is forward-looking by default, embedded where decisions happen, and measured by impact, not report volume."

OpenAI: generative AI in finance operations

The use of generative AI in finance is perhaps nowhere more advanced than at OpenAI itself. Their Contract Reader Bot extracts terms, applies ASC 606/IFRS 15 logic, and auto-generates journal entries, dramatically reducing manual effort across the close. The team has since extended AI into reporting drafts, disclosure writing, controls testing, and deal-prep analysis.

The outcome: a finance team operating with roughly 22% of the headcount of comparable tech firms, and delivering faster. "Our first step was mapping all data sources and aggregating them where possible," says She. "We've done both: merging critical data for control-heavy areas and using aggregators for private-company-level risk processes."

Adyen: Finance Data Core

Rather than adding AI tools to isolated workflows, Adyen rebuilt its data infrastructure via a multi-year effort to centralise all financial data into its Finance Data Core. Reconciliation, accounting memos, and reporting now run against a single, reliable foundation. Controllers and FP&A analysts rotate directly into technical teams to co-design use cases, building trust on both sides.

"Adoption is fundamentally a trust issue," says Peled. "If finance teams don't understand how AI outputs are generated, they won't use them."

[GAP: no specific quantified outcome metric for Adyen in the source material; "measurable impact in reconciliation and exception handling" is qualitative only.]

Zapier: mandated adoption at scale

At Zapier, AI usage is a company-wide expectation rather than a personal choice. "Transformation requires mandates, not nudges," says CFO Ryan Roccon. "At Zapier, around 98% of employees use AI tools." Their finance team measures ROI through objective metrics (cycle times, pull requests merged, issues resolved) and subjective surveys, providing a framework other CFOs can replicate.

ClickUp: platform consolidation

ClickUp's approach centres on replacing scattered point solutions with AI capabilities built natively into one platform. "The main challenge isn't finding AI tools," says Zhang. "It's having too many. We replaced five separate AI note-taking tools with a native one built in ClickUp." Beyond consolidation, ClickUp has built internal agents for investor-relations workflows, scenario modelling, internal reporting, and companywide knowledge retrieval.

Microsoft: ready-built Copilot agents for CFOs

Microsoft 365 users have two ready-built AI assistants available immediately. The Researcher Agent compiles schedules, flags important emails, and surfaces relevant documents from a user's calendar and files. The Analyst Agent handles data-heavy tasks: comparing Excel files for trends and anomalies, performing complex analysis using Python, and generating charts automatically. Tasks that can take analysts hours now take 5 to 10 minutes.

[GAP: no named CFO or finance-leader quote specific to Microsoft Copilot ROI in the source material.]

A 30/90/365-day roadmap: from experimentation to embedded advantage

First 30 days: identify and test

Focus narrowly. Select one high-friction workflow (spend categorisation, variance analysis, reconciliations, or management reporting) with clear inputs and outputs. A narrow, well-defined use case reduces risk and makes results visible to stakeholders quickly.

Before purchasing anything new, audit your existing tech stack, including your FP&A software, BI platforms, and productivity suites. Many organisations underuse embedded AI features they already pay for. The Microsoft Copilot agents above are a practical starting point for any Microsoft 365 subscriber.

Start tracking impact beyond time saved. Establish early metrics for decision speed, forecast accuracy, error reduction, and stakeholder satisfaction. These build the business case for future investment.

90 days: build capability and governance

Launch a focused automate/upskill/govern programme. Define which processes will be automated, what skills the team needs to develop, and how AI use will be governed. Identify AI champions within the finance team: curious, credible professionals who can bridge finance, IT, and data. Create governance frameworks covering data usage, model validation, access controls, and human oversight. Well-defined guardrails build trust with auditors and regulators without shutting down innovation.

6 to 12 months: scale and embed

Build a governed finance data core with clear ownership, standard definitions, and strong controls. This is the backbone for scalable AI across reporting, planning, and decision support. Redesign roles around AI capabilities: update role definitions, performance metrics, and career paths to reflect the shift toward judgement, insight, and business partnership. Make AI literacy a baseline expectation, not a specialist skill.

Identify pilots that have delivered measurable value and standardise them across teams and regions. At this stage, AI moves from innovation to infrastructure.

The skills finance teams need for AI in accounting and finance

Two skill clusters are emerging as non-negotiable for finance professionals navigating AI in accounting and finance.

Technical skills. A CFO Connect Summit 2025 poll found that 56% of finance professionals identify workflow automation (tools like Copilot and ChatGPT) as the most important AI-related skill for their careers over the next two years, followed by data literacy at 21% and cross-functional collaboration with AI at around 14% (State of AI in Finance 2026).

Source: CFO Connect Summit 2025 poll, State of AI in Finance 2026

The must-have technical skills identified in the report are LLM literacy (ChatGPT, Copilot, Gemini), workflow automation, data literacy and governance, and systems integrations. The report frames LLM literacy as the foundational layer: knowing how to prompt, validate outputs, and ground responses in source data is what unlocks the rest.

For teams sourcing AI in finance course material externally, the most useful curricula are workflow-led: identify a high-friction task first, then choose the tool. A separate practical question keeps surfacing in CFO Connect discussions: when teams ask which AI in finance courses to recommend, the report's authors point to programmes built around real finance workflows rather than generic AI overviews.

The same principle applies to finance automation courses, where the strongest programmes teach process-mapping before introducing specific platforms like Zapier, Make, or n8n. Teams evaluating the best FP&A software during this phase often find that the right platform is the one that integrates with this workflow-first thinking, not the one with the longest feature list.

Soft skills. Bichachi notes: "Finance leaders increasingly value problem-solving and storytelling skills. Only 23% of CFOs now rank deep accounting knowledge as the top hiring priority." As AI handles more analysis, the distinctly human value lies in interpreting results, framing decisions, and influencing stakeholders. Mike Tsang, Finance Director at ARIA (the UK government's Advanced Research and Invention Agency), puts it directly: "Soft skills like curiosity and rigour are timeless. But AI amplifies their importance."

What finance leaders who have got it right do differently

Define your ambition before investing in FP&A tools. Gabriel Hubert asks: "What's our target productivity gain? 30 percent? 50 percent? Do all department heads know what's technically possible today?" Dan Zhang uses a three-bucket framework: one-to-ten automation (time savings from repetitive tasks), zero-to-one unlocks (new capabilities previously impossible), and C-to-A quality boosts (better decisions, not just faster outputs). "If we only measure AI by time saved, we miss its real value: capabilities and quality."

Set expectations from the top. "Start bottom-up to spark adoption," says Roccon, "but pair it with a top-down edict to drive real transformation." At ClickUp, AI use is reinforced through weekly highlights, monthly awards, and quarterly hackathons. Adoption through persuasion alone rarely reaches scale.

Pair finance with engineering. The most successful implementations come from cross-functional collaboration rather than finance working in its own silo. Lafouge describes the Photoroom approach: "If a tool costs less than $100 and saves at least one hour, buy it and test it - no approval needed." That combination of autonomy and oversight breeds innovation without chaos.

Use a value/effort matrix to prioritise. She at OpenAI evaluates initiatives across four quadrants: low value/low effort (repetitive controls), high value/low effort (reporting and disclosure drafting, the quick wins), and high value/high effort (long-term LLM and ERP integration, the strategic bets). Move fast on impactful, achievable wins while laying the foundation for deeper integration.

Frequently asked questions

What are the key features to look for in an automated financial system?

Based on the case studies in the State of AI in Finance 2026, the most important characteristics are: a single, centralised data source (avoiding the point-product fragmentation trap); direct ERP integration so AI acts on live data rather than exports; governance controls built into the platform rather than added afterwards; and embedded AI features rather than separate tools that add complexity. The common failure is a patchwork of point products with no common data layer. Adyen's Finance Data Core and ClickUp's platform consolidation approach both illustrate the same principle: fewer, better-connected systems outperform many specialised ones.

What AI in finance course or training options should finance teams prioritise?

Formal certification matters less than structured, on-the-job practice. Leading companies have operationalised skills development through quarterly hackathons, mandatory usage expectations, and internal AI champions rather than standalone courses. OpenAI, Zapier, and ClickUp have all made AI fluency part of the job description rather than an optional extra. Teams that treat AI as a product, constantly testing and iterating, close the skills gap faster than those waiting for an approved curriculum.

What are the best AI tools used in financial services today?

The State of AI in Finance 2026 highlights five platform types delivering measurable results: large language models (ChatGPT Enterprise, Microsoft Copilot, Gemini Enterprise) for drafting, analysis, and summarisation; ERP-embedded AI (Oracle NetSuite) for automated reconciliation and reporting; knowledge management platforms (Dust) for cross-functional information retrieval; workflow automation tools (Zapier) for connecting systems and automating handoffs; and consolidated work management suites (ClickUp) that embed AI natively across finance workflows. The consistent finding across case studies is that integration beats isolation: tools connected to live finance data outperform those relying on manual input.

Where to go from here

Patrick Puck, GVP of AI Product Strategy at Oracle NetSuite, captures where leading finance functions are heading: "AI is fundamentally reshaping the finance function, driving greater efficiency, accuracy, and strategic insight. AI is allowing finance teams to move beyond reactive reporting to focus on strategy, collaboration, and growth to help their organisations navigate an increasingly complex business landscape with confidence."

The data in the State of AI in Finance 2026 is unambiguous: teams that have built a governed data core, protected time for experimentation, and a structured adoption roadmap are already operating with capabilities that were unavailable 18 months ago. For deeper CFO Connect coverage of practical prompts, checklists, and implementation guides on AI adoption, see https://www.cfoconnect.eu/resources/finance-insights/ and the State of AI in Finance 2026 report at https://www.cfoconnect.eu/resources/reports/state-of-ai-in-finance-2026/.