A small physiotherapy practice installs an AI chatbot on its website. It’s articulate, friendly and available around the clock. Within a week, it has explained the benefits of dry needling to a patient who wanted to cancel an appointment, and cheerfully assured another that the practice was open on a public holiday when it wasn’t.
Meanwhile, the receptionist still spends two hours every Friday copying appointment details from the booking system into a spreadsheet so the owner can see how busy the week was.
The practice automated something visible and complicated, and left something dull and repetitive exactly as it was. That’s a common pattern. Automation tools have never been cheaper or more capable, and the temptation is to start with whatever looks most impressive. The better place to start is usually much less exciting.
One Like insight: Automate the boring work first. It’s usually where the hours are hiding.
The potential is real
There’s little doubt that automation can take on a lot of everyday work. In 2023, the McKinsey Global Institute analysed some 2,100 detailed work activities across about 850 occupations. It estimated that generative AI, combined with other technologies, has the potential to automate work activities that absorb 60 to 70% of employees’ time today, up from its earlier estimate of about half.
That figure is easy to misread. It describes the technical potential to automate activities, not whole jobs, and McKinsey itself expected adoption to unfold over decades, with half of today’s work activities automated sometime between 2030 and 2060. Technical possibility is also very different from practical sense. Just because something can be automated doesn’t mean it should be, or that the cost and risk are worth it for a five-person business.
For small businesses, the more useful takeaway is that a large share of routine administrative and communication tasks are now within reach of affordable tools. The question is which ones to start with.
What to automate first
The best early candidates for automation tend to share four qualities. They happen often, so the time saved adds up. They follow clear rules, so a tool can do them reliably. They’re low risk, so an occasional mistake is cheap to fix. And they’re invisible to customers, so nobody feels the business has become less personal.
The physiotherapy practice’s Friday spreadsheet ticks every box. So do appointment reminders, invoice generation, payment chasing emails, backing up files, sending a review request after each visit, and moving enquiry details from a website form into a customer list. None of these will impress anyone at a networking event. Together, they can free hours every week.
The chatbot, by contrast, fails the low-risk and invisible tests. It talks directly to customers, and its mistakes, like confirming a closed practice is open, cost goodwill. That doesn’t mean customer-facing automation is always wrong, but it should come later, after you understand your processes well and can test carefully. As we explored in Why Brands Are Adding Mess on Purpose, customers tend to value a human touch most in exactly the moments they’ll notice.
One warning applies to all of it: automate a process only after it works well by hand. Automating a confusing or broken process simply produces confusion faster.
[Infographic placement] The automation sweet spot. A 2×2 grid. Horizontal axis: How often it happens (rarely → often). Vertical axis: Risk if it goes wrong / visible to customers (high → low). Top-right quadrant, highlighted green, labelled Automate first: reminders, invoices, backups, data entry, review requests. Bottom-right: Automate carefully, later: chatbots, quote responses. Top-left: Probably not worth it. Bottom-left: Keep human: complaints, sensitive conversations, bespoke advice.
Finding your hidden hours
For one week, ask everyone in the business to note any task they do more than twice that feels repetitive: copying, re-typing, sending the same message, chasing the same thing. At the end of the week, estimate roughly how many hours each one takes per month. The top three on that list, provided they’re low risk and mostly invisible to customers, are your first automation projects.
Then keep it simple. Many booking, accounting and email tools already include automation features that sit unused. Turning those on is often faster and cheaper than buying anything new. Test each automation for a few weeks, check its output regularly at first, and keep a person responsible for noticing when something goes wrong.
A few questions can help:
- Which tasks do we repeat every week that follow the same steps every time?
- Which of them would cause little harm if an automated version made an occasional mistake?
- Do our existing tools already offer automation we’re not using?
- Is the process working well by hand before we automate it?
Which brings us back to the physiotherapy practice, its eloquent chatbot and its Friday spreadsheet. Switching off the chatbot and connecting the booking system to an automatic weekly report would take an afternoon. The receptionist would gain two hours a week, patients would stop receiving creative answers about public holidays, and the owner would get her numbers on Monday morning without anyone having to type them.
Sources
- McKinsey Global Institute, “The economic potential of generative AI: The next productivity frontier”, June 2023.
