CPD · Technology
AI agents and copilots in your systems
Accountability, permissions and audit trail when software acts for you.
- 3 modules
- 3 marked written tasks of 450 to 800 words
- Pass at 80
- Facts checked 1 November 2026
- Yours to keep, no end date
- Certificate when every module passes
Who it's for
Controllers and systems owners.
The modules
Module 1
What agents can do in finance systems
Tell a copilot from an agent, place each proposed action on a scale of autonomy, and spot the failures that only appear when software acts.
- Objective:
- To be able to explain what AI agents and copilots can do in finance systems, set the right level of autonomy for each action an agent proposes to take, and identify the failure modes that arise when AI acts rather than advises.
- Marked task:
- For each of actions (a) to (f), state the autonomy level the supplier proposes and the level you would set for the pilot, with a reason.
- What the marker looks for:
- Sets the right autonomy for each action (30), Identifies failure modes specific to agents (30), Defines a bounded pilot (20), Sets evidence for raising autonomy (20)
Module 2
Permissions and accountability
Give an agent its own identity and the least access its job needs, keep segregation of duties, and name the people who answer for what it does.
- Objective:
- To be able to design permissions for an AI agent in a finance system that respect least privilege and segregation of duties, and to assign accountability for the agent's actions to named people, including what happens when it gets something wrong.
- Marked task:
- Identify the permission and segregation of duties problems in the proposed set-up.
- What the marker looks for:
- Identifies the permission and SoD problems (30), Designs least-privilege permissions and limits (30), Names accountable people and their roles (20), Sets out the error process (20)
Module 3
Audit trail
Decide what an agent must record about every step, review the log, and give auditors evidence they can rely on.
- Objective:
- To be able to specify the audit trail an AI agent in a finance system must keep, set up a routine to review it, and provide evidence that allows management and auditors to rely on work the agent has done.
- Marked task:
- Explain what is missing from the trail and why it matters for the audit and for the company.
- What the marker looks for:
- Identifies the gaps and their impact (25), Specifies a sound log (30), Sets a review routine (20), Responds to the auditor properly (25)
In every module
- A short written lesson, with its sources
- A quiz to check the ideas
- A written task, marked against a grid, with feedback. Pass at 80
- A reflection, and Copy for CPD record
A certificate for the course once every module is at 80 or more. Built for your CPD record.
Labs with this course
No lab comes with this course yet: labs are added as they open.
How it's marked
Your answer is marked by an AI model against the grid for the task, whose parts and marks are listed above for each module. Each part is judged Not shown, Partly, Mostly or Fully, and the mark is worked out from the grid's weights. A CIMA-qualified (CGMA) moderator checks a sample of marks and can add a note. Under 80, read the feedback and try again; your best mark counts.
See a marked task in full: the grid, an answer and its feedback
Price
£39
Or every course, track and lab for 365 days with the CPD year, £199. Or CPD monthly, £19.99 a month, cancel any time. Every lesson, plus 8 marked tasks a month, across every course, role track and lab.
More in AI and finance technology
On the path for using ai in finance.
Buying for a team
Seats by invoice or purchase order, with a manager report. See CPD for employers.