Everyone wants to disrupt an industry. Fewer people want to learn how the business currently answers the phone. That second group may be missing the better opportunity.
Codie Sanchez’s public teaching draws attention to established, cash-flowing businesses that are not fashionable startup stories. My application here is narrower: look for useful service and workflow improvements inside ordinary businesses. This is not advice to buy a company or a claim that acquisition is easy. See Contrarian Thinking’s own introduction.
The opportunity is often embarrassingly normal
Missed enquiries. Repeated data entry. Quotes that sit unfinished. Staff searching for the current version of a document. You do not need to describe these as an intelligence revolution. The owner already knows they are annoying.
The commercial question is whether solving one of them creates enough value to justify the work and ongoing cost. Start there. Not with a logo that looks like a chrome octopus.
My framework: Follow the repeat
Find a task that happens often. Observe how people actually do it. Measure the time and failure points. Then decide whether a process change, ordinary software or AI is the right intervention.
Yes, ordinary software is allowed. So is a clearer form. If a simple rule solves the problem reliably, adding a language model may just create a subscription with opinions.
Use AI when the task benefits from interpreting variable language, drafting from context or handling information that does not fit a neat rule. Keep a person involved where mistakes have meaningful consequences.
A hypothetical quoting workflow
Imagine a maintenance company receiving enquiries in inconsistent formats. An assistant could organise the information and prepare a draft summary. A person checks the scope, missing details and price before anything goes to the customer.
The pilot should compare the old process with the new one using similar enquiries. Count review time and corrections. Include awkward requests. If the system only works on perfect examples, you have built a presentation.
Define who maintains it when services, prices or staff change. A useful automation has an owner. An abandoned one becomes another mysterious thing nobody wants to touch.
Sell the improvement, not the technology costume
A clear offer describes the workflow, the people involved, the acceptance criteria and the support arrangement. Explain how the client can stop using it or export their information. Dependence is not the same thing as loyalty.
Do not promise savings you have not measured. Estimate where appropriate, label the assumptions and revise after a pilot. If the economics are weak, say so. You can lose a small sale and gain a reputation for not talking nonsense.
What to do this week
Speak to three operators in one type of business. Ask which recurring administrative task causes disproportionate frustration. Watch it happen if they are willing. Compare the answers before designing anything.
Choose one repeated problem you understand well enough to demonstrate a possible improvement. Build the smallest safe version. Test it with real-world variety and agreed handling of data. Then decide whether there is a service worth offering.
The attraction of ordinary businesses is not that their owners are unsophisticated. It is that useful work is often hiding behind unglamorous processes. Respect the operator. Understand the process. Make something better. The robot logo can wait forever.