May 30, 2026 / 4 min read

A simple framework for deciding how to handle any AI task

Deciding when to automate or hire is hard. This framework walks a small business through five options for any task so you pick the one that fits.

For any problem in your small business, you have five real options: automate it with AI, build something custom, buy a tool off the shelf, hire a person, or wait. Deciding when to automate or hire, and when to do neither, is the whole skill here. Most people fixate on one option, usually "automate it with AI," and never weigh it seriously against the rest. That is how you end up with expensive, fragile systems doing work a person could have done better. The right answer depends on the problem. Here is how to think through each path without a chart.

Automate when the task is stable and checkable

Automation earns its keep when a task repeats often, follows a fairly consistent shape, and produces output you can verify. Sorting inbound messages, drafting first versions of routine documents, pulling data out of forms: good candidates, because the cost of a mistake is low and a human can review the result.

The trap is automating something that changes constantly, or where errors are expensive and hard to catch. If you cannot easily tell whether the automation got it right, you have not removed work. You have moved it somewhere harder and added risk on top.

Build only when the thing is core to you

Building something custom makes sense when the capability is genuinely central to your business and nothing off the shelf fits. Building gives you control and a solution shaped exactly to your needs. It also gives you a maintenance burden that never ends, a thing that breaks when models change, and a cost that keeps running long after launch.

Most small businesses should build far less than they think. The honest question is whether this capability is a real differentiator or just a process you want handled. If it is the latter, building is almost always the wrong call.

Buy when someone already solved it well

Buying an existing tool is underrated by people who like building. If a vendor already solved your problem, and a lot of others share that problem, buying is usually faster, cheaper over time, and less fragile than rolling your own. You get maintenance, updates, and support you would otherwise own forever.

The cost of buying is dependence and a recurring bill, and you should watch the usage-based fees that increasingly ride on top of subscriptions. But for a common problem, buying beats building most of the time, and it is not close.

Hire when judgment and relationships matter

The option people skip in an AI conversation is hiring an actual person. Some work is about judgment, accountability, trust, and dealing with other humans in ways a model handles poorly. A person owns outcomes, adapts to situations nobody scripted, and builds relationships. Try to automate that away and you often get something worse that merely looks cheaper, which is not the same as cheaper.

Often the best answer is a person supported by AI, not one or the other. The human does the judgment and the relationship. The tools handle the repetitive surface work. The combination beats either alone.

Wait when the technology is about to make this trivial

The most overlooked option is waiting. AI is moving fast enough that a problem needing a complicated custom solution today might be a simple feature in a few months. Spending heavily now to solve something the next release will solve for free is a real and common mistake.

Waiting is not always available. Sometimes the problem is urgent. But for non-urgent problems where you can see the capability improving quickly, the disciplined move is a cheap stopgap now and a revisit later, rather than over-investing in a solution that will be obsolete before it pays back.

The actual skill

The point is not which option to pick. It is the habit of considering all five before committing, and being honest about how stable the task is, how core it is to your business, how expensive mistakes are, and how fast the underlying technology is moving. Most bad AI decisions come from skipping that pause and reaching for the option that felt exciting rather than the one that fit. The five-way frame exists to slow that reflex down.