A new survey of 212 finance professionals finds that 54% of those aged 22–32 say they “love” Microsoft Excel—a higher proportion than any older generation. The finding, reported by TechRadar from Datarails research, lands just as Microsoft rolls out Copilot Agent Mode, embedding AI that can build models and explain formulas directly inside workbooks. The combination of cross-generational devotion and powerful new AI suggests Excel isn’t just clinging to relevance; it’s becoming a strategic platform for finance teams—and a governance headache for IT.
What the data actually says
The Datarails survey, which polled finance professionals across the US and UK, surfaced several headline numbers alongside the “love” stat: 89% of all respondents believe Excel will be as important or more important in the next decade, and most would hesitate to accept a role that bans the spreadsheet. Younger workers reported spending five to seven hours a day inside Excel, a figure that helps explain the emotional attachment—familiarity breeds competence, and competence breeds preference.
The survey also captured anecdotes about why enterprises stick with Excel. TechRadar notes that Airbus reportedly cites file-size limitations as a reason its finance team continues to rely heavily on Excel rather than cloud-first alternatives. This aligns with a broader truth: for large, complex models, desktop Excel’s local performance and compatibility with legacy macros and add-ins still outpace web spreadsheets.
Caveats apply. Datarails is a vendor of Excel-native FP&A tools, and its survey sample is modest (212 respondents). The numbers should be read as directional, not universal, especially since the full methodology and recruitment details aren’t exhaustive.
Why finance teams can’t quit Excel
Excel’s staying power isn’t inertia. Several practical factors keep it bolted to core workflows:
- The universal tabular model. Rows and columns map directly to business concepts—accounts, periods, products—making workbooks a near-universal handoff format for auditors, regulators, and stakeholders.
- Composability and transparency. Formulas, named ranges, dynamic arrays, and pivot tables make logic visible and auditable without code. An experienced user can inspect and alter a calculation on the spot.
- Local performance. When models involve thousands of calculations or VBA macros, desktop Excel often outperforms browser-based tools, especially for the heavy lifting of financial modeling.
- Entrenched expertise. Years of templates, reconciliation routines, and “Excel muscle memory” mean that migration costs—broken macros, retraining, and temporary productivity hits—are hard to justify.
- Reporting integration. Finance teams frequently use Excel as the final consolidation layer, delivering outputs in the format everyone expects.
These advantages explain why even younger workers, trained on Excel as the default analytics tool in university and early career roles, express such loyalty.
The AI infusion: Copilot and beyond
Microsoft is now layering agentic AI directly into the grid. The Copilot Agent Mode, currently in preview and tied to licensed Copilot programs, can plan and execute multi-step tasks: generate and explain formulas, build PivotTables and charts, and even construct entire model skeletons. Crucially, it shows its plan and proposed edits before applying them, allowing users to inspect and approve changes.
Third-party assistants are also entering the fray. Anthropic, for example, has introduced Claude for Excel in limited preview, offering cell-level explanations and finance-specific connectors. This multi-vendor landscape means enterprises may soon have to decide which AI assistant to trust inside workbooks—each with its own governance and licensing implications.
What these AI tools promise is compelling:
- Faster model building from plain-English prompts.
- Democratization: junior analysts can handle tasks that once required deep formula fluency.
- Better traceability when agents show their reasoning steps.
But the promises come with material caveats.
What this means for you
For finance professionals: Copilot will accelerate daily tasks—building dashboards, cleaning data, generating reports. But the output must be verified. Treat AI-generated formulas as drafts, not final deliverables. Your role will shift from constructing every formula manually to reviewing and stress-testing AI suggestions.
For IT administrators: Agent Mode requires cloud connectivity and tenant-level enablement. You’ll need to configure admin toggles to control which AI models are available, ensure data residency and non-training contractual clauses, and set up audit logging. Third-party add-ins demand separate vetting; procurement and security must validate vendor claims independently.
For business leaders: AI can boost productivity, but it also introduces risks: skill erosion, hallucination, and compliance headaches. Budget for Copilot licensing (consumers will see Copilot bundled into some Microsoft 365 plans for a fee) and invest in training that teaches staff to verify, not just accept, AI outputs.
The risks you need to know about
- Hallucination and errors. LLM-based assistants can produce plausible but wrong formulas or aggregations. If teams accept suggestions blindly, errors will propagate into financial reports. The only reliable defense is human-in-the-loop verification and reproducible regression tests.
- Skill erosion. If Copilot routinely writes formulas, junior staff may lose core spreadsheet literacy, becoming dependent on AI that might not always be available or correct.
- Data governance. Agentic features often send workbook data to cloud models. Organizations in regulated sectors must demand explicit guarantees—non-training commitments, specified data residency, encrypted transmission—before enabling them.
- Spreadsheet fragility. Complex workbooks with hidden dependencies, circular references, or untested macros remain brittle. Agents that change formulas at scale can create downstream contradictions unless acceptance tests are in place.
- Cost creep. Copilot licensing and premium model access are evolving; enterprises must anticipate vendor-driven price changes and budget accordingly.
How to get Excel AI right: an action plan
A pragmatic path forward is neither “rip out Excel” nor “accept everything AI suggests.” Instead, take these steps:
- Inventory critical workbooks. Map every spreadsheet used for statutory reporting, regulatory filings, and high-stakes decisions. Rank them by business impact and error sensitivity.
- Pilot on low-risk workloads. Start with non-regulated, repeatable tasks—quick dashboards, exploratory analysis, data cleanup. Measure time savings, error rates, and user confidence.
- Mandate human verification. Require acceptance tests, reconciliation checks, and peer review before any AI-generated model is used in published reports. Insist that Copilot deliverables include formula previews and stepwise explanations.
- Lock down tenant controls. Use Microsoft 365 admin toggles to restrict which models and external vendors are available. Contractually require non-training clauses, define data residency, and request audit logs.
- Build regression suites. Create golden datasets and automated reconciliation scripts that validate agent changes against known outputs.
- Invest in cross-skilling. Encourage a “learn by review” culture: require junior staff to inspect and document AI-generated logic. That builds both verification skills and deeper understanding.
- Consider hybrid architectures. Use Excel as a last-mile interface while housing source-of-truth data in governed data warehouses or FP&A systems. This preserves flexibility while reducing fragility.
Outlook: what to watch next
- Adoption of Agent Mode. How quickly enterprises enable the feature and what governance playbooks emerge will set the tone for AI in spreadsheets.
- Pricing and licensing shifts. Changes to consumer and enterprise Copilot tiers will directly affect adoption choices. Watch for bundled subscriptions and premium add-on fees.
- Evidence on error rates. Independent studies that measure whether agentic features reduce errors in production (not just speed in pilots) will be critical for building trust.
- Ecosystem expansion. Third-party assistants with finance-specific connectors and “Agent Skills” will challenge Microsoft’s integrated experience, forcing enterprises to evaluate a multi-vendor landscape.
Excel’s domination at 40 isn’t about nostalgia—it’s about a durable, composable model and deep institutional knowledge. AI promises to make that platform even more powerful, but only if organizations govern it carefully. By inventorying critical workbooks, piloting agentic features cautiously, requiring human verification, and hardening tenant controls, IT and finance leaders can capture the productivity upside while keeping the risks in check.