By AIGAS
AIGAS Weekly | 12 November 2026
Research window: 5–11 November 2026; checked Wednesday morning.
AI agents are moving closer to everyday financial workflows. That can remove administration and help firms focus on judgement, but it also makes permissions, approvals and evidence more important.
This week, new accounting-vendor announcements, ICAEW guidance and questions about the independence of AI oversight all point in the same direction. Our AI governance framework for accounting firms is designed around that practical balance: enable useful adoption, while keeping professional responsibility visible.
Sage and Canopy show why release status matters
On 5 November, Sage described an expanded AWS collaboration focused on agentic financial workflows. Sage was unusually clear that its announcement did not introduce a new customer-facing product or availability date. A separate Sage Intacct update described an Agent Marketplace, permission controls and a planned limited rollout of an accounts-payable workspace. Availability varies by capability and market. Read Sage’s collaboration announcement.
On 10 November, Canopy said its Coworker agent was available to customers as an Experimental Feature, similar to an open beta. Canopy says firms can decide which actions run automatically, separate job creation from approval, and review session and activity logs. It also says Coworker data is processed in the US and that an opt-out from using inputs and outputs to improve the feature is still being finalised. Read Canopy’s announcement and control notes.
These are supplier statements, not independent tests. UK firms should confirm regional availability, beta terms, data location and the exact actions needing approval. Our Vendor Transparency Programme gives suppliers a consistent way to present that evidence.
ICAEW puts confidentiality back at the centre
ICAEW published regulatory guidance on 9 November warning that entering client information into public AI tools can create an unauthorised disclosure risk. It says firms should review policies and training regularly, and clarifies that an AI-policy breach is not automatically reportable to ICAEW; the firm must assess whether the Code of Ethics may have been breached. Read the ICAEW guidance.
The practical point is that a policy cannot simply say “do not enter confidential data”. Staff need approved alternatives, clear examples and a route to report mistakes quickly. Firms should also distinguish between a consumer chatbot, an enterprise account and AI embedded inside practice software: the data terms and controls may be materially different.
Independence needs structure as well as good intentions
On 8 November, the Associated Press reported concerns about the independence of the US government centre responsible for AI-safety evaluation. AP said the centre lacked a permanent leader, had stopped publishing research and had been asked to report informally to a White House office. The account relied partly on anonymous former officials; the White House did not answer the specific questions. Read the AP report.
In the UK, a parliamentary answer on 10 November set out the AI Security Institute’s conflict controls. Staff and advisers must declare relevant interests; conflicts apply across procurement and grant decisions; and the Cabinet Office Permanent Secretary is accountable for their management. Read the parliamentary answer.
For accounting firms, the lesson is proportionate but important. A reviewer is not independent merely because they have a different job title. Record their interests, authority, evidence access and ability to challenge the person sponsoring the tool.
This week’s action for your firm
Choose one AI feature that can act across client work and check:
- Status: Is it generally released, in beta or only a roadmap item—and is it available in the UK?
- Data: Where is client information processed, retained and used to improve the service?
- Challenge: Who can approve, stop and independently review the workflow, and are conflicts recorded?
Capture the answers in your free AIGAS AI Tool Register and assign an owner to gaps. Agentic AI can be valuable, but clear boundaries are what allow firms to use it with confidence.