Cortea

Cortea is pitching AI at the point in an audit where quality problems are cheapest to fix: before sign-off, and earlier still in planning and disclosure work. For a firm worried about file review findings, a consistent AI check across every engagement is an attractive idea, provided it is reliable enough that reviewers trust the flags.

What it does well. It is one of the few audit AI products that speaks the UK's language directly, referencing ISA (UK) 315, FRC guidance, UK GAAP and Companies Act disclosure requirements rather than only US standards. The scope now runs from planning and risk analysis through disclosure checklists, data validation and evidence testing, so it is more than a final review tool. The founders include a former auditor, investors include a former KPMG global head of audit, and testimonials from Forvis Mazars and dhpg show real firms using it. The clear auditor-decides stance and no-training commitment are the right defaults.

Where it falls short. Pricing is not published and every route in is a demo. Security wording is a little vague, with an ISO 27001 badge but only SOC 2 Type I referenced, so firms should ask for the certificates and reports. No integrations with audit file platforms such as Caseware are listed. Press coverage of flags raised across thousands of reports is the company's own figure and says nothing about false positives, which is what will determine reviewer workload.

Best for. UK and European audit firms with a quality or technical team that wants a consistent AI review and checklist layer across engagements. The right pilot is to run Cortea over completed and already reviewed files, compare its findings with what the human reviewers found, and count false alarms. Firms whose main need is evidence extraction in Excel should start with DataSnipper, and those after a full AI-led engagement platform should compare Fieldguide.

Verdict. A credible, well-backed AI quality layer with an unusually strong UK standards focus. Pilot it on completed files first and judge it on the ratio of useful flags to noise.