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 for financial operations, reporting, planning, and risk management. Over half the practices are good practice: fraud detection, expense management, invoice processing, and financial forecasting have mainstream adoption. Regulatory compliance and audit automation are advancing. The domain is tightly clustered around good-practice with minimal bleeding-edge — finance favours proven, auditable tools over experimental ones.
The headline: AI in finance works on paper but rarely pays in practice -- only about 7% of finance chiefs report real returns despite 60% having deployed it. The bottleneck is not the technology; it is your data, your controls, and whether the output can survive an audit.
Nearly every finance task you can name -- processing invoices, spotting fraud, closing the books, scoring credit -- now has AI tools that demonstrably work, and almost every finance team is using something. But adoption is not the same as payoff. Roughly 88% of pilots involving agents (software that acts on its own without being prompted) never reach full production, and 95% of deployments show no measurable bottom-line impact. The companies pulling ahead are not the ones with the cleverest AI; they are the ones with clean data and strong governance who can prove to an auditor how a number was produced. The rest are stuck in pilots. If you have deployed AI and cannot yet point to a hard financial return, you are firmly in the pack, not behind it -- and the gap to close is organizational, not technical.
The Big Four just demonstrated, in public, what happens when AI governance fails. KPMG withdrew a research report after major institutions including UBS and the UK's NHS contradicted its claims -- only 5 of 45 sources held up. Deloitte Australia refunded $291,000 of a government report containing fabricated citations and quotes from court judgments that do not exist, and EY Canada retracted another with most of its citations made up. If the largest, best-resourced firms cannot stop hallucination (when an AI tool confidently makes things up) in their own published work, treat any AI-generated finance document as requiring a human reviewer before it leaves the building.
Autonomous finance agents became standard vendor equipment, not a premium feature. Oracle alone announced more than 600 AI agents built into its finance software, with over 1,000 already running in customers' systems; the bill-payment platform BILL.com now automates 1.2 million invoices and has made AI its top priority. The capability is now table stakes -- the question for you is no longer whether to buy it but whether your processes and controls are ready to switch it on safely.
The rules stopped being suggestions. Europe's AI Act now sets a firm December 2027 deadline for credit-scoring systems to prove they are fair and explainable; the US audit regulator confirmed that any AI touching your numbers, estimates, or disclosures now falls inside your formal financial controls; and six US states restricted insurers from letting AI decide coverage without a human. Compliance is moving from optional policy to mandatory operational control, and your AI vendor choice should now be driven by audit-defensibility, not benchmark scores.
EU AI Act credit-scoring deadline, December 2, 2027. Any AI used in lending or creditworthiness decisions will need formal bias testing, explainability, and a conformity assessment. If you lend, start the governance work now -- conformity assessments take many months and the floor is binding, not aspirational.
Your auditors will start asking how AI touched the books. With AI-influenced figures now formally inside financial-controls (SOX) territory, expect audit-trail and human-accountability questions in your next cycle. Confirm now that every AI-assisted number can be traced and explained, or budget for the controls to make it so.
The "trust tax" will decide which pilots scale. The cost of checking AI output for compliance is what keeps most finance pilots from going live. Vendors are racing to offer governed, audit-ready platforms -- favor those over raw capability, because the platform that is easiest to audit will scale fastest and cheapest in a regulated function.
Finance needs the same answer twice; AI does not naturally give it. Auditable finance work requires that identical inputs produce identical outputs, but today's AI is probabilistic and can return different results on the same task. That is why routine, high-volume work (like accounts payable) automates well while judgment-heavy work (like journal entries, under 5% automated) does not.
The barrier is your data and controls, not the model. Better AI will not fix fragmented data, unclear success criteria, or missing governance -- the actual causes of failed deployments. Money spent on data quality and process discipline returns more than money spent chasing a more powerful model.
Being first carries hidden risk. The lender Upstart, an early AI-credit leader, faces securities lawsuits alleging its model failed to handle changing economic conditions -- a reminder that even leading production systems can carry model risk you cannot see until it breaks.
Go deeper: the full Finance & Accounting briefing -- the longer analytical write-up, plus every practice we track in this domain with its maturity rating, the tools to consider, and the evidence behind our assessment.