The AI landscape doesn't move in one direction — it lurches. Some techniques leap from experiment to table stakes in a single quarter; others stall against regulatory walls, technical ceilings, or organisational inertia that no amount of hype can dislodge. Knowing which is which is the hard part. The State of Play cuts through the noise with a rigorously maintained index of AI techniques across every major business domain — classified by maturity, evidenced by real-world adoption, and updated daily so you always know where you stand relative to the field. Stop guessing. Start knowing.
A daily newsletter distilling the past two weeks of movement in a domain or two — delivered to your inbox while the index updates in the background.
Each dot marks the weighted maturity of practices within a domain — hover for a brief summary, click for more detail
AI that automates spreadsheet tasks including formula creation, data analysis, pivot tables, and formatting from natural language. Includes Excel/Sheets AI assistants and formula generation; distinct from natural language to SQL which queries databases rather than manipulating spreadsheets.
AI-driven spreadsheet automation works in narrow, structured workflows -- and struggles nearly everywhere else. That split defines the practice's stalled position at bleeding-edge. Accounting and finance teams have extracted real value: 60% adoption rates, measurable time savings, and fast payback cycles. But these wins have not generalised. Gemini in Sheets achieves 70.48% accuracy on complex real-world tasks, nearing human baseline; independent testing of Copilot in Excel shows formula generation succeeds 7/10 times on first attempt. Yet these capability gains mask a widening governance crisis: CSA survey (418 security professionals) finds 82% of organizations discovered ungoverned shadow agents, 65% experienced security incidents, 61% reported data exposure. Finance ROI remains trapped in a measurement gap: Deloitte survey of 1,300+ finance leaders shows 63% deployed automation but only 21% report clear, measurable ROI. Formula injection attacks (Ramp Labs vulnerability) and permission-drift risks demonstrate the automation is outpacing governance capability. Until shadow-agent visibility, lifecycle management, and formula-level security are resolved, adoption will remain confined to teams with well-defined, repetitive data workflows in compliance-light domains.
May 2026 announcements confirm vendor feature velocity but reveal critical governance and security gaps replacing feature limitations as the central barrier. OpenAI shipped ChatGPT for Excel/Sheets sidebar add-in GA (May 8) with manifest XML deployment option for regulated enterprise environments blocked from app stores. Google's Workspace Intelligence and Fill with Gemini (April 22) continue expanding automation across Sheets, enabling conversational dashboard creation and data entry at 9x manual speed. Microsoft's "Edit with Copilot" agentic feature (GA March 2026) enables multi-step Excel edits across Python, OpenAI, and Anthropic Claude models. Yet independent testing reveals performance gaps: Copilot in Excel achieves 70% formula generation success rate and 12-second latency on 8,000-row datasets but requires mandatory verification. Security and governance have become disqualifying: CSA survey of 418 security professionals found 82% of organizations discovered ungoverned shadow automation agents, 65% experienced security incidents, 61% reported data exposure from AI automation. Ramp Labs' Sheets AI suffered a critical vulnerability (May 2026) where prompt injection via hidden spreadsheet text allowed agents to exfiltrate financial data through formulas without user consent.
Finance sector deployment metrics show real ROI but masked by measurement gaps. AR automation demonstrates concrete adoption (40% payment acceleration, 90% error reduction, 91% success rate among mid-market firms). Mayo Clinic's RPA deployment (2.4M transactions automated, 84,000 staff hours saved annually) and City of Los Angeles' licensing automation (45 days to 6 days, 94% backlog reduction) provide case-study evidence of organizational payback. Yet Deloitte's 2026 Finance Trends survey (1,300+ leaders) reveals 63% deployed financial automation but only 21% report clear ROI—a measurement and benefit-realization gap where automation costs are visible but benefits are diffuse (labour freed to higher-value work) and difficult to track. The strongest third-party signal remains: GPT for Work leads ecosystem with 7M+ installations and 4.9★ rating. Outside finance, adoption continues to stall: only 3% of enterprises achieve meaningful AI transformation; 56% of CEOs report no AI ROI. The central obstacle: ungoverned shadow agents, formula accuracy failures under conditional formatting (83%), and data exposure risks now supersede licensing friction as the blocker to mainstream trust.
— ChatGPT Excel/Sheets sidebar add-in GA across all tiers with manifest XML deployment for enterprise environments bypassing app-store restrictions; addresses adoption friction for regulated enterprises.
— Ramp Labs Sheets AI vulnerability: prompt injection via hidden text allowed agents to exfiltrate financial data through formulas without consent. Demonstrates critical operational risk in formula-based automation.
— AR automation deployment metrics: 40% payment acceleration, 90% error reduction in reporting, 91% success rate among mid-sized businesses, 71% CFO prioritization. Finance-specific adoption and outcome data.
— Deloitte Finance Trends 2026 survey (1,300+ leaders): 63% deployed financial automation but only 21% report clear ROI; widespread deployment masked by measurement barriers and diffuse benefit tracking.
— CSA survey (418 security professionals): 82% discovered ungoverned shadow agents; 65% experienced security incidents; 61% reported data exposure. Spreadsheet automation agents lack foundational visibility and lifecycle control.
— Named deployments (Mayo Clinic: 84,000 staff hours saved annually, 18:1 ROI; City of Los Angeles: 45 to 6 days processing, 94% backlog eliminated) showing measured transition from spreadsheets to automation workflows.
— Independent testing of Excel Copilot on 8,000-row datasets: data analysis queries return results in 12 seconds; formula generation succeeds 7/10 times. Documents both capability and limitations (large datasets slow performance; verification mandatory).
— Major Google feature release adding end-to-end spreadsheet creation, side-by-side editing, and workspace intelligence for complex optimization problems.