The question everyone is asking wrong I had a call last week with someone running a small logistics business. She had three people doing admin, was drowning in invoices, and wanted to know whether to hire a fourth person or "get some AI thing" to handle it. The question itself was wrong. She was framing it as either/or when the real answer depended on about six factors she had not considered. This is the conversation I have constantly. Lloyds just announced a £2bn cost-cutting plan built around AI. Chinese robotics firms are expanding into UK retail because of labour shortages. The pressure to automate is everywhere. But the pressure to hire is also everywhere, especially when you watch virtual interviews fail to show you who someone actually is. So how do you decide? You need a framework. Not a gut feeling. The four questions that actually matter When someone asks me whether to automate or hire, I walk them through four things: How repeatable is the task? If you can write down the exact steps and they rarely change, automation works. If the task requires judgment calls, context, or adapting to situations that look different each time, you probably need a person. Invoice processing? Highly repeatable. Client relationship management? Not repeatable at all. What is the error cost? Some mistakes cost you £50 and an apology email. Others cost you a client or a compliance fine. Automation is brilliant at consistency but terrible at catching edge cases it was not built for. I worked with someone last year who automated their customer support responses. Worked perfectly until a complaint about a faulty product got the standard "thanks for your feedback" reply. That one error cost them a £4,200 refund and a very public social media post. Does it need to grow with you? Hiring scales linearly. You need more output, you hire more people. Automation scales exponentially until it does not. Most small businesses hit automation ceilings faster than they expect because their processes are not standardised enough. What is the true cost of each option? People always underestimate the cost of hiring. Salary is maybe 60% of the real number. Add national insurance, pension contributions, equipment, training time, management overhead. A £30,000 salary easily becomes £42,000 in actual cost. They also underestimate automation costs. The software subscription is cheap. The implementation, the training, the maintenance, the time lost when it breaks: that is where the money goes. The matrix I actually use I built a simple decision matrix for this. Two axes: task complexity on one side, volume on the other. Low complexity, low volume: Do it yourself. Not worth automating or hiring for. Low complexity, high volume: Automate. This is where automation shines. Data entry, appointment reminders, invoice generation, report formatting. High complexity, low volume: Hire a specialist on contract. You do not need someone full-time, but you need human judgment. High complexity, high volume: Hire full-time. This is where you need people who can think, adapt, and handle the weird situations that will definitely come up. The logistics business owner I mentioned? Her invoice processing was low complexity, high volume. Perfect for automation. But her supplier negotiations were high complexity, medium volume. She needed a person for that, just not a fourth admin. She ended up automating the invoices and hiring a part-time operations manager. Total cost was actually lower than hiring another admin full-time, and her output went up. Where most people go wrong The mistake I see constantly is automating the wrong layer of a process. Someone will automate their lead capture but still manually qualify every lead. Or they will automate their email sequences but have no system for tracking which sequences actually convert. Automation works best when you automate entire workflows, not random tasks within a workflow. Otherwise you create handoff problems between the automated parts and the manual parts. Those handoffs are where things break. The other mistake is hiring for tasks that should not exist. Before you hire or automate, ask whether the task needs to happen at all. I have seen people hire someone to manage a reporting process that nobody actually reads. That is not a hiring problem or an automation problem. That is a process problem. A real example from last month A client came to me running a small consultancy. Eight people, growing fast. She wanted to hire an office manager to handle scheduling, expenses, and general admin. We ran through the matrix together using the Decision Matrix tool on alira.london. Scheduling was low complexity, high volume. Expenses were low complexity, medium volume. General admin was a catch-all that included some genuinely complex tasks like client onboarding. The answer was not "hire an office manager." The answer was: automate scheduling with Calendly and a few integrations, automate expense tracking with Dext, and hire a part-time client success person to handle onboarding properly. She saved roughly £18,000 annually compared to hiring a full-time office manager. More importantly, the client onboarding got better because someone was actually focused on it instead of squeezing it between admin tasks. The timing question With interest rates holding at 3.75% and economic uncertainty everywhere, cash flow matters more than it did two years ago. That changes the calculation. Automation has higher upfront costs but lower ongoing costs. Hiring has lower upfront costs but higher ongoing commitments. If your revenue is unpredictable, automation gives you more flexibility. If your revenue is stable and growing, hiring gives you more capability. Neither is universally right. It depends on your situation. What to do this week List every task someone does more than five times per week. Not job descriptions. Actual tasks. "Process invoices" is one task. "Answer client emails" is another. Get specific. Score each task on complexity (1-5) and volume (1-5). Anything with complexity under 3 and volume over 4 is a strong automation candidate. Anything with complexity over 3 and volume over 3 probably needs a person. Run one task through the Decision Matrix at alira.london. Pick the task you are most uncertain about. The framework will force you to think through the factors properly instead of going with your gut. The goal is not to automate everything or hire for everything. The goal is to put the right solution on the right problem. That is how you build something that actually works.