The robot pizza problem There is a pattern playing out in the restaurant industry right now. Robotic pizza makers, once heralded as the future, are failing. Not because the technology does not work. It works fine. The problem is that nobody asked whether a robot was the right solution to begin with. I see this constantly with people running their own thing. They hear about automation, they see the demos, they get excited. Then they spend three months and £4,000 building something that saves them twenty minutes a week. The maths never made sense. But the promise of automation is seductive, and seduction makes you skip the hard questions. The question you should ask first Before you automate anything, ask this: is the problem I am solving a problem of volume, or a problem of clarity? Volume problems respond well to automation. You have 200 invoices to send. You have 50 leads to follow up. You have a thousand rows of data to sort. Machines handle repetition without complaint. Clarity problems do not respond to automation. You are not sure which leads are worth pursuing. You do not know why your invoices keep getting queried. You cannot explain why some customers stay and others leave. No tool fixes confusion. You just move the confusion faster. I worked with someone last year who wanted to automate their client onboarding. They had built three different Zapier workflows, connected four tools, and spent six weeks on it. When I asked what problem they were solving, they said "it takes too long." When I asked how long, they said "I don't know, maybe two hours per client." They had ten clients a month. Twenty hours. That is not a volume problem. That is a clarity problem dressed up as a volume problem. We spent two hours mapping out their actual process. Turned out they were asking for the same information three times in different forms, and nobody had noticed. They did not need automation. They needed to delete two forms. What automation actually costs People underestimate the true cost of building automation. The tool subscription is the smallest part. The real costs are: Time to build it. Most custom automations take 10 to 40 hours to set up properly. If your time is worth £75 an hour, that is £750 to £3,000 before you save a single minute. Time to maintain it. Tools update. APIs break. Edge cases appear. I estimate 15% of the build time per year in maintenance. Forever. Time to fix it when it fails. And it will fail. Usually at the worst moment. Usually in a way that is hard to diagnose. Opportunity cost. Those 40 hours could have gone toward selling, delivering, or thinking. What did you give up? I am not against automation. I build systems for a living at ALIRA. But I am against automation that exists because someone wanted to feel productive rather than be productive. The three things worth automating After years of doing this, I have a short list. If a task does not fit one of these categories, I question whether automation is the right move. Repetitive data entry. Moving information from one place to another, unchanged. Copying a form submission into a spreadsheet. Logging a payment in your records. This is pure volume work. Automate it. Scheduled communications. Follow-up emails, reminders, check-ins. If the message is the same every time and the trigger is predictable, automate it. But only if you have tested the message manually first and know it works. Reporting and dashboards. Pulling numbers from multiple sources into one view. This saves time and reduces errors. But only if you know which numbers actually matter. If you are not sure, you will just automate your confusion. Everything else deserves scrutiny. Client communication that requires judgement? Do not automate it. Decisions that depend on context? Do not automate them. Anything that touches money or trust? Think twice, then think again. When to build it yourself Sometimes the right answer is not automation but a manual process you actually follow. A checklist. A template. A weekly habit. I use the 5 Whys tool on alira.london when I am trying to figure out whether something needs a system or just needs attention. You ask why five times, and usually by the third or fourth answer, you realise the root cause is not what you thought. Often the root cause is that nobody decided who owns the problem. No tool fixes that. The pension overpayment scandal that hit military veterans this week is a good example. Hundreds of people were overpaid due to clerical errors, then told they owe thousands back. That is not a technology failure. That is a process failure. Someone, somewhere, did not check. Automation would have moved the wrong numbers faster. The honest test Before you build any automation, answer these four questions honestly: Can I describe the process in exact steps, with no judgement calls? Does this task happen at least 20 times a month? Have I done this manually, successfully, for at least three months? If this automation breaks, what is the cost? If you cannot answer all four, you are not ready to automate. You are ready to document, test, and refine. That is not less valuable. That is the work that makes automation possible later. What to do this week Audit your current automations. List every tool, workflow, or integration you have built in the last year. For each one, write down how much time it saves per month and how much time it cost to build. Be honest. If the maths does not work, consider killing it. Pick one manual process that frustrates you. Do not automate it. Instead, write out every step on paper. Look for steps you could delete entirely. Most processes have at least one. Set a threshold. Decide now: you will not automate anything unless it saves at least 5 hours per month. Write that number down somewhere you will see it. Next time you get excited about a new tool, check the number first.