A useful AI decision begins with one recurring piece of work. The goal is to decide whether to configure or create what is missing, adopt a product that already fits, or leave the opportunity alone for now.
Three directions
Build. Configure, connect, adapt or create the missing capability. A Build can use software you already own. It may connect systems, add a bounded AI-assisted step or create a focused internal tool. It does not automatically mean custom software from scratch.
Buy. Adopt an existing product when it handles the important work sufficiently, with ordinary setup and team adoption. Test it against representative examples before choosing.
Skip. Preserve time and money when the likely value, readiness or risk does not support action. Skip can mean “not yet” or “keep this consequential decision with a person.” Name what would need to change before reconsidering it.
Start with the work
Describe the trigger, the main steps, the people involved and the intended result. Consider both frequency and consequence: a rare task with a serious consequence can matter more than a daily inconvenience.
Then ask whether the workflow has a stable core. Exceptions are workable when someone knows how to handle them. If the steps and desired result change substantially every time, map the process before automating it.
Check products before commissioning a Build
A tested product that handles the workflow end to end points toward Buy. If a product covers the core but material work remains, name that gap:
- Configure and extend: adapt the tool to the process.
- Connect systems: move information or actions across tools.
- Focused AI workflow: interpret documents or messages, apply bounded judgment or prepare a draft for review.
These are descriptions of a Build, not separate verdicts. If you have not compared relevant products, research and test them before treating a custom Build as the answer.
Keep a credible review point
Can someone reliably check the result before it affects a customer, payment, legal obligation or safety?
If an error may be hard to spot before it causes harm, skip that form of automation. Keep the consequential decision human-led, or redesign the workflow so AI prepares information without taking the action.
Before proceeding, establish three basics: a person responsible, real examples including unusual cases, and a clear description of an acceptable result. Missing basics are preparation work, not a reason to buy more tools.
An illustrative workflow
After a customer meeting, an account manager prepares follow-ups and updates the CRM. The core is consistent, but information must move between the notes and the customer record.
If the existing product handles that end to end with ordinary setup, Buy may be sufficient. If an important connection or reviewed drafting step remains after product testing, Build may be justified. If nobody can own or check the result, prepare those basics before implementing it.
This is an example of the decision process, not a client case study or a claim that a particular product has been evaluated.
Make the next step small and concrete
For a Build, gather representative examples, write pass/fail tests and map the trigger, main steps, exceptions and final result. Scope the lightest implementation that can pass those tests.
For Buy, write three must-have requirements and test two or three products against the same real example. Compare setup effort, data access, review controls, operating cost and ownership.
For Skip, record the reason and what would change it: higher consequence or volume, a stable process, a reliable approval step, or an accountable owner.
Apply the method
Run the Build, Buy or Skip worksheet with one workflow. The result includes the direction, the reasons, the remaining checks, a next step and what could change the answer. It is a first-pass decision, not product research or an implementation scope.
If you need help comparing opportunities, explore the AI Opportunity Review. If you already have a defined process, you can go directly to an AI Workflow Build. The review is not a required first purchase.