Upcoming Events:
AI for CFOs: Starting with AI in Finance Using Claude
Thursday 29 October 2026, 8:30 to 10:00 AM (UK time)
In person: WeWork, Moor Place, 1 Fore Street Avenue, London EC2Y 9DT
Free, limited seats
A working breakfast for finance leaders on moving from Claude access to measurable impact. The session covers setting up Claude, building skills, plugins and projects, and two use cases: month-end close and FP&A cash forecasting. Hosted by Selona.
Reserve your seat on Eventbrite. Each attendee, including colleagues, registers separately.
News of the Day -

Only 1 in 5 large-company CFOs uses advanced controls on AI spending. Deloitte's Q3 CFO Signals survey asked 200 CFOs at North American companies with at least US$1bn in revenue how they manage AI costs. Only 20% use AI-specific controls such as token-level cost tracking, AI gateways or caching. 53% mainly review AI costs after the money is spent, through bills and reports. When costs pass approved limits, 74% take several business days to several weeks to pause the usage. Pricing is moving the same way. Workday co-founder and CEO Aneel Bhusri told investors this month that usage-based pricing for its AI agents is "a far better model than a people-based subscription-based model, which will continue for a long time." Worth asking now: if your AI bill doubled next month, how many days would it take you to notice? Deloitte | The Register
The FCA has shown what good and poor AI practice looks like, and the FSB has set 12 practices for boards. On 19 October the FCA published 20 case studies across three areas: Consumer Duty, governance and accountability, and operational resilience. They "do not create new requirements", but they show what supervisors look for. Two points apply to any finance team: firms relying too heavily on a single AI provider, and human review of AI output turning into box-ticking instead of real oversight. Four days earlier, the Financial Stability Board finalised 12 "sound practices" for adopting AI, and it "strongly encourages" boards and senior management to reference them. India's RBI, SEBI and Ministry of Finance sit on the FSB, as do the UK's Bank of England, FCA and HM Treasury. Worth checking: who in your team signs off what the AI produces, and do they actually read it? FCA | FSB
68% of Asia Pacific CFOs report tangible value from AI. UK CFOs think they are a quarter of the way there. Deloitte's Asia Pacific CFO Pulse surveyed 363 CFOs across 14 markets, India among them. 68% reported tangible value from their AI investments. The report gives no India-only figure. In the UK, Deloitte's Q3 survey of 66 CFOs found 67% believe they are around a quarter of the way to the change they expect AI to deliver. 52% expect AI to reduce headcount in their own business over the next 12 months. The value is starting to show, but most CFOs think the bigger change is still ahead. Deloitte Asia Pacific | Deloitte UK
Article of the Week -

Before AI, Fix the Workarounds: Why Discretion Breaks Automation
AI adoption fails when layered over unwritten workarounds and tribal habits disguised as agility. Because automation requires consistent rules and structured inputs, automating chaotic workflows only scales errors faster. Before investing in AI, businesses must build process intelligence: map actual workflows, turn discretionary judgment into standardized rules, and eliminate exceptions.
To read more, click here.
The Shortlist -

Process mining: see how the work really runs
This week's article says map the real workflow before you automate it. These three tools do the mapping. They read your system's event logs and rebuild the path every invoice or order actually took, including the loops and the waiting.
Celonis is built for large companies and is strongest on SAP. Its Accounts Payable starter kit tracks touchless invoices, on-time payment, DPO, cash discounts and duplicate invoices, but it works with SAP source systems only. A duplicate invoice checker gives suspected duplicates a machine-learning confidence score and can block the payment. Pricing is quote-based. Expect a data project: other ERPs need extraction and modelling work first.

Microsoft Power Automate Process Mining suits teams already on Microsoft 365. Process and task mining come with Power Automate Premium at £11.50 / ₹1,250 per user per month (paid yearly, before VAT or GST). The catch: Premium includes 50 MB of process data per user, so a year of ERP history usually needs the add-on at £3,844.80 / ₹4,15,975 per tenant per month. Microsoft's finance templates cover accounts payable and procure-to-pay on SAP tables.

mindzie is the quickest way to test the idea. The free desktop edition handles up to 1,000 cases with no time limit, on Windows only. Its procure-to-pay view tracks touchless processing, blocked invoices and payment delays, and helps you find why early-payment discounts are lost. Paid editions are quote-based.

Verdict: three different jobs, not three competitors. Celonis if you run a large SAP estate and want ready-made AP apps. Power Automate if you already pay Microsoft and want it inside tools your IT team runs. mindzie if you want to test one process this month, for free, from a CSV export.

That’s it for this week.
Keep executing, not just experimenting. See you next week.
The CFO AI Hub Team
