Why the next hire might not be a person.
Most operations roles blend repeatable process with human judgement. Here is how to compare automation with hiring honestly, whether you are being asked the question or asking it about your own job.
When a workflow outgrows the person doing it, the default is usually to hire. That may still be right, but the comparison worth making first is between the human role and a controlled automation option, before the job is posted.
Should the next hire be a person or an automation?
Split the job spec first. The rules-based, structured, high-volume part is the automation candidate; judgement, relationships and accountability are not. Compare the loaded cost of the role with the honest cost of the alternative, human review included, and count released capacity as cash only if a budgeted hire is genuinely avoided.
Read the job spec honestly
Take the most recent operations job spec. Mark the work that is sequenced and rules-based: sorting tickets, reconciling returns, pulling numbers into a report, or progressing a supplier checklist. Then mark the work that depends on judgement, relationships, negotiation or accountability. Only the first category is a credible automation candidate.
McKinsey's State of AI in 2025 reports widespread experimentation but limited scaling. Deloitte's 2026 enterprise research identifies customer support as the area where respondents expect the highest agentic impact, with supply chain and knowledge management also prominent. Those findings identify candidate areas, not guaranteed savings.
The maths on one role
Start with the actual loaded cost of the role: salary, employer contributions, benefits, software, recruitment and cover. Compare it with the honest cost of the alternative: subscriptions, the hours spent setting the workflow up, monitoring and upkeep, the oversight someone has to give it and the human work that remains. Do not treat capacity released as cash saved unless a budgeted hire is genuinely avoided.
The shape matters more than a headline saving. Automation can own defined steps while a person owns exceptions and decisions. Use the payback method to separate capacity, avoided cost and recovered cash, and do not count the same benefit twice.
Where agents quietly go wrong
Agents fail when they are handed judgement work dressed as routing. Stanford Digital Economy Lab's Enterprise AI Playbook studied 51 successful implementations and found that readiness, process design, trust and data infrastructure were usually harder than the model. That supports a cautious operating principle: automate defined work, retain meaningful human review and measure exceptions.
The example is the boring one. Software flags a revenue dip as significant when an experienced operator would know it is seasonal. Software suppresses an outlier that does not fit its heuristics, and leadership gets a tidier but materially less accurate picture. The thing that makes a good operations manager is that they spot what does not fit the rules. That does not transfer. Hand agents the rules-based work, the synthesis, and the routing. Keep contextual judgement, exception filtering, and weak-signal detection with a person.
Lean teams are the new operating model
Agents can change how a team allocates work, but public research does not justify a universal headcount ratio. BCG's AI Radar 2026 reports that nearly all surveyed CEOs expect agents to produce measurable returns in 2026. McKinsey's 2025 survey finds AI high performers are nearly three times as likely as others to report fundamental workflow redesign. The next workflow that gets too big should therefore trigger a design decision, not an automatic hire or automatic automation.
The shape that wins is small teams running agents, not no teams.
DSIT's January 2026 research found that 71% of surveyed businesses reported lack of identified need as a barrier to adoption and 60% reported limited AI skills, expertise and knowledge as a barrier. A clear use case and named owner are therefore practical starting controls.
Where the maths still says hire a person
Three categories remain human-led. In research published on 18 March 2026, the British Chambers of Commerce Insights Unit, with the University of Essex and Atos, found that 54% of UK firms were actively using AI, while 95% of SME AI users reported no workforce-size impact over the past year and 86% said job roles were unchanged. Current adoption is primarily evidence of augmentation, not wholesale substitution.
Senior judgement roles.
Heads of department, finance directors, commercial directors. Exception handling, relationships, the long memory of what went wrong.
Customer-facing trust.
Account management for your biggest customer. Sensitive complaints. The bespoke side of B2B sales.
Genuinely creative.
The first creative pass on a campaign. The editorial line on a launch. Agents help draft; they don't yet originate at a publishable level.
What the next hire actually looks like
What changes when agents own the rules-based slice is the role you wanted to hire for. The next req gets sharper. Less doing, more deciding. More senior, harder to find. But the right role. The people already in the team can start shifting that balance this week, and our guide to working alongside AI day to day is where that begins.
Before you post the next ops job
- If a substantial part of the role is rules-based work with structured inputs, test automation before approving the hire.
- Use real loaded employment and automation costs, including human review.
- Keep judgement, relationships, accountability and consequential decisions with people.
Frequently asked questions
When should a UK SME consider automation before hiring?
Consider automation when a substantial part of the proposed role is high-volume, rules-based work with structured inputs, known exceptions and a measurable output. Keep judgement, relationships and accountability with a person.
How should I compare an AI agent with an operations hire?
Use the company's real loaded employment cost and the quoted build, software, monitoring, maintenance and internal oversight costs. Compare like-for-like tasks and include the value of human review. Do not assume released time becomes cash savings.
What roles should still be hired as people?
Prioritise people for senior judgement, customer trust, negotiation, accountability and genuinely creative direction. Agents can support those roles, but should not own consequential decisions.
What commonly blocks SME AI projects?
DSIT's January 2026 research found lack of identified need and limited AI skills were the two most reported barriers among surveyed businesses. A bounded use case, named owner, reliable data and explicit controls reduce those risks.
What is the agent-plus-champion model?
A central owner sets platform, security and governance standards, while an operational champion in each function owns the workflow, exceptions and outcomes. The exact structure should match the size and risk of the business.
References
McKinsey, The state of AI in 2025; Deloitte, State of AI in the Enterprise 2026; Stanford Digital Economy Lab, Enterprise AI Playbook; BCG, AI Radar 2026; Department for Science, Innovation and Technology, AI Adoption Research; and British Chambers of Commerce Insights Unit, University of Essex and Atos, SME AI adoption research, 18 March 2026.
The UK adoption figures quoted above, with their denominators, sample bases and caveats set out in full, are collected in our roundup of UK AI adoption statistics.
Want this built in your business?
This is the kind of system Clerq builds inside UK businesses - in your tools, around your controls, handed over working. Tell us what is slowing you down and the founder will reply personally. Every price is on the homepage menu, and there is no obligation.