FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Professional indemnity explained
Caseware AiDA is an AI digital assistant inside Caseware's software, whose UK launch Caseware announced on 15 October 2024. It matters because when an assistant's answer on methodology or documentation is wrong, the firm that relied on it usually carries the professional responsibility.
Part of: Audit firm PI cover
In short
Caseware describes AiDA as an AI digital assistant that summarises long PDFs, answers questions about a user's documentation and on methodology and compliance, and helps with audit evidence and workpapers. Caseware says no prompts are stored or fed back into the model. Its own head of analytics and AI has said AiDA "should always be subordinate to users" and "you still have to check it". For audit and accountancy firms, the PI point is that the firm remains accountable for the work. ISQM (UK) 1 and the FRC's 2026 AI guidance both point to review, competence and documented judgement.
AiDA is Caseware's AI digital assistant. In plain terms, it is a chat style helper built into Caseware products that answers questions and drafts material for the user. Everything below is Caseware's own description, or press reporting of what Caseware said.
Caseware's pages we read do not name the underlying AI model, the hosting region or any security certification for AiDA. If those points matter to your risk assessment or your clients, ask Caseware directly and keep the answer on file.
Caseware's announcement says that "no prompts are stored or fed back into the model". AccountingWEB reported that no user prompts are stored or shared and no company data is used to train the model. Those are Caseware's statements; a firm cannot verify them itself, so treat them as supplier assurances to record in your due diligence, not as a guarantee.
On review, the clearest statement comes from Danielle Supkis Cheek, Caseware's head of analytics and AI, quoted by AccountingWEB: "AiDA is a digital assistant and should always be subordinate to users", "Auditors shouldn't hand over their judgment to the system" and "AiDA helps you get to that first stage, the first draft, but you still have to check it". That is consistent with the FRC's position, below, that the human auditor is always accountable. The tool supports the work; the firm signs it.
For the wider picture of AI assistants used for audit methodology, see AI methodology chatbots for auditors.
An assistant that answers a technical question quickly is useful, but a wrong answer applied consistently across files can spread an error further than one person's mistake. The table sets out common uses and what a reviewer would usually check.
| Use | What can go wrong | What a reviewer usually checks |
|---|---|---|
| Methodology question (for example, sample sizes or procedures) | Answer does not fit the firm's own methodology or the engagement facts | The firm's methodology and the relevant ISA (UK) requirement |
| Technical accounting question | Answer reflects the wrong framework or an out of date position | The standard itself, such as FRS 102, and any firm technical guidance |
| PDF summary of a contract or board minutes | Summary omits a clause or condition that matters | The source document for the points relied on |
| Draft workpaper or deliverable | Draft reads well but is incomplete or not supported by evidence | Evidence on file, reviewer sign off and documented judgement |
The table is a general guide only. For technical research, see AI technical research on FRS 102.
ISQM (UK) 1 is the FRC's quality management standard for firms that perform audits or reviews of financial statements and other assurance or related services engagements. Firms had to design and implement systems of quality management under it by 15 December 2022. Paragraph 32(f) expects appropriate technological resources to be "obtained or developed, implemented, maintained, and used", and paragraph 32(g) expects intellectual resources, such as methodology, to be "consistent with professional standards and applicable legal and regulatory requirements". An AI assistant that answers methodology questions touches both.
On 30 March 2026 the FRC published Generative and Agentic AI Guidance. It says it "codifies good practice", that it "does not alter" accountability under auditing standards, and that "the human auditor is always accountable". The guidance itself expects outputs to be reviewed by "humans in the loop" with "appropriate competence", professional scepticism and awareness of automation bias. It does not name any commercial AI tool, so it neither approves nor criticises AiDA. ICAEW members are also bound by the ICAEW Code of Ethics, whose 2025 update, enforced from 1 July 2025, includes changes relating to technology.
More detail: certifying AI tools under ISQM (UK) 1 and FRC AI guidance explained.
Accountants' and auditors' PI usually responds to claims that the firm was negligent in its professional services. Using an AI assistant does not usually change that: the claim is about the work the firm delivered, subject to the policy terms. Cyber cover usually deals with data breaches and system compromise. Where an assistant processes client documents, both policies can be relevant. See PI cover for AI mistakes, audit firm PI cover explained and cyber insurance for accountants.
A senior asks an assistant how to treat a lease under the firm's framework and pastes the answer into a file. The answer is wrong for the facts, the reviewer does not test it, and the accounts are later restated. A claim against the firm for the cost of restatement is a professional services claim, and PI usually responds, subject to the policy terms.
An assistant's summary of board minutes leaves out a covenant breach. The audit team relies on the summary and does not read the minutes. If a going concern issue is missed, the claim is about the audit, not the software, and the question will be whether the review met the expected standard.
For more on how these claims are handled, see AI errors, liability and PI for auditors.
See also AI audit tools and PI for audit firms, our accountants PI insurance guide and the AI tools by profession hub.
Apex arranges PI and cyber cover for audit and accountancy firms; a broker reads every enquiry.
Caseware's UK announcement says no prompts are stored or fed back into the model, and AccountingWEB reported Caseware saying no company data is used to train the model. These are supplier statements. Record them in your due diligence and ask Caseware for contractual terms if your clients need firmer assurance.
Usually the firm that relied on it. Caseware itself says AiDA should be subordinate to users and its output must be checked, and the FRC says the human auditor is always accountable. A claim would be about the quality of the firm's work, and PI usually responds subject to the policy terms.
No. The FRC's March 2026 Generative and Agentic AI Guidance names no commercial AI tool, and it is good practice guidance rather than an approval scheme. Firms remain responsible for deciding whether a tool is appropriate under ISQM (UK) 1.
If your proposal form asks about AI use, answer it accurately. Using an assistant does not usually change the scope of PI cover, but check your wording for any AI exclusion, condition or endorsement, and ask your broker if unsure.
AccountingWEB reported that AiDA is available to Caseware's cloud and desktop users, with desktop functionality possibly limited depending on the task. Check the current position with Caseware for the products your firm uses.
Tell us which tools your firm uses, what they produce and how output is checked. We will review how your PI and cyber cover respond and what insurers will ask at renewal. No fee and no obligation. Or call 0117 325 0027.
Book a PI programme review Get a PI quoteApex Insurance Brokers Limited is authorised and regulated by the Financial Conduct Authority. Registered in England and Wales, company number 07014570. This page is general information, not advice on your individual circumstances, and it does not guarantee that cover will be available or on what terms.
