For years, payroll has mostly lived out of sight. Many organizations still treat it as a background task, something that only reaches senior leaders when a crisis appears. In 2026, that approach is under real pressure.
New HMRC rules and wider Employment Rights Act changes in the UK are bringing pay accuracy and timeliness into sharper regulatory focus.
CEO & Co-Founder of HealthboxHR.
At the same time, the EU AI Act is formally treating many HR and worker management systems as “high risk”, with stricter expectations around documentation, oversight and governance.
Sitting between those two developments is one of the unsexy, yet most exposed, parts of the enterprise stack: payroll.
What used to be seen as a chore is now a test of how seriously companies take data quality, automation and resilience.
Payroll slips and people pay the price.
Payroll errors are often talked about in abstract terms, but the impact on employees is immediate and concrete. In a recent survey of 2,000 UK workers, one in five said a wrong or late payslip had already caused them to miss a bill or regular payment, and 18% had been pushed into borrowing through credit cards, overdrafts, loans or friends and family because their pay was incorrect or delayed.
Around a third of employees (32%) said they could not cope if their main pay was wrong or late even once, and the strain is heaviest on Gen Z and millennials, who are more likely than boomers to say errors have driven them into debt, made it harder to cover essentials and affected their sleep or mental wellbeing.
The retention risk is just as serious. In the same research, 61% of workers said they would be likely to look for a new job if pay mistakes or delays continued for six months. Among Gen Z and millennials that figure rises to around three in four.
In sectors like science and technology, where skills are hard to replace, that level of flight risk turns payroll accuracy into a strategic issue, not just a procedural one.
Taken together, these numbers show that payroll now behaves much more like critical infrastructure than background administration. When people are paid correctly and on time, most never think about the gear behind it.
When that process breaks, it quickly becomes a problem of financial stability, wellbeing, reputation and retention. That is why payroll belongs in the same category as security and finance at board level, rather than being buried as a low‑status back office task.
Why 2026 exposes brittle payroll stacks
The underlying problem is less about individual mistakes and more about the architecture that makes them likely.
Across mid‑market and multi‑site organizations, a familiar old picture appears. HR software sits in one platform, while rotas, time and attendance are managed in separate tools or spreadsheets. Payroll software runs in a standalone system that only partially integrates with either of them. Then, on top of all that, exceptions, corrections and approvals are all pushed around via email and informal workarounds.
Legislation is always changing, so HR and Payroll teams must continuously wrap their heads around new rules across multiple different environments - which increases the risk of something being missed. Take the Employment Rights Act changes, or HMRC’s reforms to payroll and tax, as two examples. In a unified stack, these can be updated centrally and applied consistently. In a fragmented setup, each change becomes a project that relies on coordination across teams and systems that were never designed to move in step.
In practice, the main regulatory exposure does not usually come from misunderstanding the law. It comes from trying to implement that law on top of brittle infrastructure. The combination of new HMRC requirements, employment law changes and tighter expectations around AI will flush out that fragility. Spreadsheets that felt “good enough” during calmer periods will struggle under closer scrutiny.
AI is landing on top of that fragility
While regulation tightens, AI is also being added across HR and payroll workflows.
Vendors are releasing anomaly detection capabilities to highlight unusual payments, natural‑language tools so managers can query HR or Payroll data in plain English, and assistants that handle routine employee questions about holidays and pay.
These are logical places to apply AI, because they involve repeated, structured tasks at scale.
Under the EU AI Act, however, many of these systems now fall into a “high risk” category. That brings expectations around clear technical documentation, human review of important outputs and visible audit trails.
Those requirements are manageable on top of a solid platform. They are far harder to meet when data flows depend on manual exports, copied spreadsheets and point‑to‑point integrations that have grown organically over time.
It is possible to have a sophisticated anomaly detection model watching payroll data, yet still see serious errors slip through because the underlying inputs are incomplete or inconsistent. AI cannot compensate for missing foundations.
For technology and finance leaders, the key question has shifted. The focus is no longer on whether to use AI in Payroll, but on the quality of the stack AI is applied to, and on the controls that surround it.
What a resilient, AI‑ready payroll stack looks like
If older payroll approaches are reaching their limit, a stronger alternative starts with how the stack is put together.
The first shift is towards a genuine single source of truth. Core people data, roles and rules should live in one consistent record that HR, time, attendance and payroll all draw from.
That can still involve more than one application, but there needs to be one authoritative view rather than several loosely aligned versions.
When that foundation is in place, updating statutory sick pay, holiday entitlement or new HMRC rules becomes a focused configuration task, not a manual rewrite scattered across tools.
It also makes it far easier to explain to auditors and regulators how pay has been calculated in a specific case.
The second shift concerns the role of AI. In Payroll and HR, AI is well suited to scanning large volumes of data, spotting unusual patterns and answering routine questions.
It can highlight potential payroll or pension anomalies, bring emerging absence or scheduling trends to the surface and respond to standard employee queries so fewer reach support teams.
Even so, people should remain in control wherever jobs, pay or compliance are involved. AI should make work faster and clearer, while human reviewers retain responsibility for interpreting results and making final decisions.
The final shift is about visibility and ownership. A resilient stack allows teams to trace a payslip or scheduling choice back through the data, rules and approvals that produced it, including any AI‑generated recommendations.
That requires clear logging, version control for configuration and transparent change histories. It also calls for visible sponsorship at senior level, with HR, finance, IT and security involved in key decisions about platforms and automation.
In practical terms, that group should be able to answer straightforward questions: what data a tool depends on and how reliable it is in real life, how new errors or bias will be spotted and corrected over time, and what the plan is if the system fails during a critical period such as payday.
When those answers are clear, AI becomes part of a stronger payroll foundation rather than another fragile layer on top of an already stressed stack.