FCA authorised · FRN 724952 0117 325 0027Quote & buy →
Aipedia: Payroll bureaux
A bureau that uses AI to migrate a new client's payroll can inherit, or create, errors in year-to-date tax, National Insurance and pensions. Whether it is liable depends on what it agreed to check and whether a competent bureau would have spotted the problem.
Part of: AI in payroll bureaux
In short
Onboarding a new client means moving employee records, pay elements and year-to-date figures from the old provider or spreadsheet into the bureau's system. AI tools now map fields, match employees and flag inconsistencies. Errors can come from the old data or be introduced in the move, and they then affect every pay run until found. The bureau owes the client reasonable care and skill, and the contract should say what checks are included. PI usually responds to a negligence claim for migration errors the bureau introduced or should have caught, subject to the wording, but not to errors that were already in the data and outside scope.
A migration involves employee details, tax codes, student loan plans, pension scheme membership, pay elements, attachment orders, holiday balances and year-to-date figures for tax, National Insurance and pensions. Bureaux use AI tools to read exports and spreadsheets, map old pay codes to new ones, match employees, and flag gaps or inconsistencies. Some tools propose a full configuration for the new payroll.
Migrations happen under time pressure, often mid-year, and the first live pay run is the test. Anything the tool mapped wrongly will appear there or, worse, run quietly for months.
The bureau owes its new client reasonable care and skill under section 13 of the Supply of Goods and Services Act 1982. The scope of that duty depends on the onboarding terms: did the bureau agree to check the inherited data, or only to load what it was given? Without clear terms, a client may argue that a competent bureau would have spotted obvious problems.
Migration is a large transfer of personal data, including health and financial information, so Article 28 processor terms, security and any AI tool provider's retention terms matter. HMRC's student loan guidance is one example of how inherited data matters: it says over-deductions in the current tax year should be refunded and year-to-date records amended.
Illustrative scenario: a bureau takes on a client mid-year from another provider. An AI tool maps the previous payroll's codes and imports year-to-date figures. It treats a taxable allowance as non-taxable because of how the old system labelled it, and it drops pension year-to-date figures for a group of employees with a different scheme. The first few pay runs look normal. At year end, tax and pension figures do not reconcile, and the client faces corrections, employee queries and adviser fees.
Other examples include duplicated employees, missed attachment orders and holiday balances reset to zero.
Migration errors are hard to unpick because two providers are involved. The client may not know which system produced a wrong figure, and the previous provider may no longer be engaged or may dispute responsibility. A bureau that cannot show what it received and what it changed is likely to find itself defending errors it did not make.
If the client shows that the bureau introduced the error, or should have caught it within the agreed scope, PI usually responds to compensation and defence costs, subject to the policy terms. Errors already in the data may be the previous provider's responsibility, and the client may pursue them instead. Disputes about which provider caused what are common, so records of the data received and checks done are valuable. Tax the client always owed and any HMRC penalties are usually treated differently; see fines and penalties under PI.
These steps make migrations safer:
Apex arranges professional indemnity for payroll bureaux. A broker can check that your wording covers onboarding work and how it treats errors inherited from another provider. See also AI in payroll bureaux.
Not automatically. It depends on what the bureau agreed to check and whether a competent bureau would have spotted the problem. Clear onboarding terms and a record of the data received help separate inherited errors from new ones.
Usually, where the bureau negligently introduced or missed an error within its scope and the client suffers a loss, subject to the policy terms. Tax the client always owed and any penalties are usually treated differently from compensation for the bureau's negligence.
At minimum, year-to-date pay, tax, National Insurance, student loan and pension figures, plus tax codes, attachment orders and holiday balances. A parallel run comparing the old and new payroll catches many mapping errors before they reach employees.
They can speed up mapping, but a person who understands the client's pay elements should review every proposed mapping. Labels in old systems are often misleading, and a tool cannot always tell whether an allowance is taxable or pensionable.
Tell us about your firm and the work you do. A broker will review how your current PI treats AI-assisted work, or arrange new cover. No fee for a quote. Or call 0117 325 0027.
Start a PI quote Request a callbackApex 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 about how AI is used in the profession and how insurance may respond. It is not legal, regulatory or insurance advice on your circumstances, and it does not guarantee that cover will be available or on what terms.
