The wrong question Most people I work with ask this question backwards. They start with the solution: "Should I get software or a person?" Then they try to figure out which one fits. That is like asking whether you need a hammer or a screwdriver before you have looked at what you are fixing. The right question is: what is actually breaking? What task is not getting done, or getting done badly, or eating time that should go elsewhere? Start there. The answer often becomes obvious. What automation is actually good at Automation excels at tasks that are repetitive, rule-based, and high-volume. If you can write down exactly what should happen in every scenario, software can probably do it faster and cheaper than a person. Think: sending the same email to every new lead. Copying data from one spreadsheet to another. Generating invoices when an order comes in. Posting to social media on a schedule. I helped someone last year automate their client onboarding. They were spending about 90 minutes per new client sending welcome emails, creating folders, adding contacts to their CRM, and scheduling a first call. We built a system that does all of it in under two minutes. That freed up roughly 12 hours a month. For a one-person operation, that is a full working day back. But here is the thing: that only worked because the process was the same every time. No judgement calls. No exceptions. No conversations. What people are actually good at People are good at things that require judgement, context, and adaptation. If the task changes based on circumstances, or requires reading a situation, or involves building a relationship, you need a person. Think: handling a customer complaint where something genuinely went wrong. Deciding which opportunities to pursue. Having a sales conversation. Managing a team member who is struggling. Lloyds Bank announced this week they are cutting £2 billion in costs through an AI-powered strategy. That sounds dramatic. But if you read the detail, they are automating back-office processes, not customer relationships. The pattern holds: routine tasks get automated, human judgement stays human. I have seen people try to automate things that should not be automated. One client wanted to use AI to respond to customer service emails. The result was a disaster. Customers could tell immediately. Complaints increased. They ended up hiring someone anyway, but now with damaged trust to repair. The real trade-off Automation costs money upfront and saves time over the long run. Hiring costs time upfront and saves capacity immediately. If you need something done tomorrow, you probably need a person. If you need something done a thousand times over the next year, you probably need automation. There is also a flexibility question. A person can adapt to new situations. Automation does exactly what you built it to do, nothing more. If your business is changing fast, automation can become a liability. You build something, the process changes, now you have to rebuild. I use the Decision Matrix tool on alira.london when I am working through this with clients. You list out the criteria that matter: cost, speed to implement, flexibility, quality of output, ongoing maintenance. Then you score each option against them. It sounds simple, but it forces you to be specific about what you actually need. The hybrid answer The best setups I see are usually hybrid. Automation handles the predictable parts. People handle the exceptions and the relationships. A recruitment agency I worked with automated their initial candidate screening. Software reviews CVs against job requirements and flags the ones worth looking at. But a person still does the actual interviews. Andy Burnham made a point this week about virtual interviews not showing personality or passion. He is right. Some things require presence. The automation does not replace the recruiter. It gives the recruiter time to do the work that actually matters. This is the pattern across industries. Chinese robotics firms are expanding into UK retail, handling warehouse logistics and stock management. But the shops still need people on the floor. The robots move boxes. The humans solve problems. When to automate first Automate first when: The task is identical every time You can write down the rules without exceptions The volume is high enough that time savings compound Getting it wrong does not damage relationships A good test: if you could hand someone a checklist and they could do the task perfectly without asking any questions, it can probably be automated. When to hire first Hire first when: The task requires judgement or reading context You are not sure what the process should be yet The work involves building trust or relationships Mistakes are costly and need human accountability Another test: if you would need to explain "it depends" more than twice while describing the task, hire someone. The cost question People ask me about cost constantly. The honest answer is that automation is usually cheaper per task, but more expensive to set up. Hiring is usually cheaper to start, but more expensive over time. A basic automation setup might cost £500 to £2,000 to build properly. A part-time hire might cost £1,500 per month. Run the numbers over six months, twelve months, two years. The break-even point varies wildly depending on volume. Jaguar Land Rover announced job cuts this week. That is the macro version of this same calculation: when does automation become cheaper than people? For large manufacturers, the maths often favours machines. For small teams, it is more nuanced. What to do this week First, list the five tasks that take up the most of your time each week. Be specific. Not "admin" but "updating the CRM after calls". Second, for each one, ask: could I write down the exact steps without any "it depends"? If yes, that is a candidate for automation. If no, that needs a person. Third, pick one task from the automation list and map out what the process actually is. Every step. Every decision point. The 5 Whys tool on alira.london can help here if you are not sure why the task exists in the first place. Once you have the map, you will know whether automation is realistic or whether you have been fooling yourself about how simple it is.