AI guiding principles with Steve Elcock

Steve Elcock, Director of Product – AI and HCM, at Zellis, shares clear, practical principles to help HR teams adopt AI effectively.

AI has huge potential to support HR and payroll teams – but only if it’s applied with care. Speaking at the recent iPASS Annual Conference & Exhibition, Steve broke down what responsible AI adoption really looks like.

Here’s your roadmap to getting AI adoption right from the start.

In this blog

Start by creating a company AI policy

Focus on real use cases

Prioritise security

Embrace embedded AI solutions

Conclusion and key takeaways

Start by creating a company AI policy

You need to get the basics in place before you adopt an AI tool. This starts with creating a clear company AI policy – even a simple one.

At minimum, your policy should set out five pillars:

1. 1. Scope & purpose (define where AI can and can’t be used, and the business reason for each use)
   
2. 2. Governance (who approves, monitors and audits every AI workflow)
   
3. 3. Ethical rules (fairness, transparency and scheduled bias checks)
   
4. 4. Security & privacy controls (encryption, access levels, retention and supplier standards)
   
5. 5. Employee training & reporting channels (how staff learn to use AI safely, and where they flag concerns)

Creating an AI usage policy might not seem like the most exciting task, but it’s actually the catalyst that will fuel all your AI efforts. Once it’s in place, your team will know what to do (and what not to do).

They can then start experimenting with AI and implementing it in their daily workflows, safe in the knowledge that they’re not jeopardising the company’s policies or accidentally misusing the tools at their disposal.

Focus on real use cases

If you’re serious about using AI in HR or payroll, your first step is to ignore the hype and focus on solving genuine use cases. These use cases should have clear boundaries, measurable outcomes, and a clear benefit to employees, customers, and business performance.

Which processes take an excessive amount of time and energy? Which repetitive tasks would you gladly outsource if you could? Use these questions as a filter.

Look for tasks that are high-volume and rules-based. For example, with ELLA, Zellis’ 24/7 AI Assistant, HR teams can use prompts to handle a wide range of activities. These include:

• Content generation e.g. writing job descriptions – “Draft a job description for a team member working at our new Birmingham site.”

• Payroll admin – “Scan this P45 and pre-fill the starter record.”

• Sickness management – “Please book sick leave for Alice McGowan for the past 2 days and send notification to payroll”

• Recruitment search – “Show candidates who’ve worked at a competitor and hold a Russell-Group degree.”

• Policy queries – “What’s our maternity leave entitlement in plain English?”

With an AI assistant like ELLA, HR and payroll teams can handle these tasks in minutes rather than hours. No bouncing around systems or manually sifting through documents yourself. Simply ask a question, or type a command, and receive exactly what you’re looking for.

Think of AI as a smart layer on top of what your team already does well. Its impact depends on the quality of your data, the clarity of your processes, and the judgement of the people guiding it.

When those elements are in place, AI can help your teams be more efficient and effective.

Prioritise security

Security isn’t something you can bolt on later – it needs to be baked into your AI strategy from the start. When you’re evaluating AI tools for HR and payroll, you’re dealing with some of your organisation’s most sensitive data. Employee records, salary information, and personal details all require the highest level of protection.

Here’s what you need to ask before implementing any AI solution: Does this tool meet our existing security standards? Can it integrate with our current authentication and authorisation systems? Many standalone AI tools promise quick wins but lack the rigorous security controls you need for enterprise use.

Look for tools with enterprise-grade controls. For example, ELLA runs inside your existing Zellis tenancy, so it inherits single sign-on and role-based access control. This means that one employee can’t ask about another person’s salary because the core HR system already blocks that field.

Get security right from the start, and everything else follows. When your foundation is solid, your team can innovate without fear. That’s where the real transformation begins.

Embrace embedded AI solutions

The future of AI in HR and payroll isn’t about adding more tools to your tech stack – it’s about making your existing systems smarter. Embedded AI solutions like ELLA work within the platforms you already use, giving you all the benefits of artificial intelligence without the complexity and security risks of standalone tools.

These solutions are purpose-built for your industry. Instead of generic AI that tries to do everything, embedded solutions understand the specific challenges of HR and payroll. They know what a P45 looks like, they understand payroll cycles, and they’re designed around the workflows your team uses every day. ELLA, Zellis’ AI Assistant, is designed specifically for HR and payroll teams, seamlessly integrating into your existing Zellis platform to handle tasks like:

• Payslip explanations – “Why did my student-loan deduction rise this month?”
• Payroll anomaly detection – “Flag anyone whose net pay shifted ±10% vs last month.”
• CV parsing & intelligent search – “List applicants with >5 years QSR experience and CIPD Level 5.”
• Natural-language reporting / SQL – “Give me the total overtime cost by department for the 2024 financial year
• Document parsing & policy search – “Summarise our flexible-working policy in three bullet points.”

This is an example of AI that works the way you do. There’s no learning curve or disruption to your existing processes. It simply makes everything you’re already doing faster, smarter, and more effective.

Conclusion: Take a purposeful approach to implementing AI

The most successful AI implementations don’t happen overnight – they happen by design. By following these principles, you can build an AI strategy that protects your data, empowers your people, and delivers real results.

Your team gets more time for strategic work. Your company becomes more efficient. And you stay ahead of the curve without compromising on security or quality.

Key takeaways

• Start with a company-wide AI policy. Even a basic framework gives your team the confidence to experiment safely.
• Identify genuine use cases before choosing tools, focusing on solving processes that waste time and energy.
• Make security non-negotiable. Embedded AI solutions protect your data while delivering the functionality needed.
• Evaluate every tool against your standards. If it doesn’t align with your security requirements, walk away.
• Use AI to elevate your team. The best implementations free up people for higher-value, strategic work.
• Build for the long term. Thoughtful AI adoption creates sustainable competitive advantage.

Want to see what embedded AI looks like in practice?

Discover how Zellis is helping HR and payroll teams adopt AI with confidence.