The Small-Business AI Opportunity Map

A consulting workbook for choosing what to improve first

By Chris Costa · Costa AI Labs

When several tasks look suitable for AI, the next question is priority. Which problem is worth addressing, what would success mean and what needs to be true before anyone builds a solution?

This workbook helps you prepare a practical project conversation. It is a planning aid, not a prediction of return on investment or a promise that every process should use AI.

1. Map one process

Choose one journey, such as a new inquiry becoming a scheduled planning call.

Trigger: ______

Who receives it: ______

Information needed: ______

Systems involved: ______

Current output: ______

Who approves the next step: ______

Mark where information gets copied, where someone waits and where questions repeat. Do not assume that the slowest step needs AI; it may need a clearer form, a better template or a simpler handoff.

2. Turn complaints into candidate projects

These are examples, not diagnoses. Describe the actual frequency, people and information involved in your business before choosing a solution.

3. Compare opportunities consistently

For each candidate, record:

• How often the task occurs.
• Whether the input is accessible and approved for the intended tool.
• Whether a good result can be checked.
• How costly a mistake would be.
• Whether the work can start as a reversible draft.
• Who will own review and ongoing operation.

Do not combine these into an impressive-looking savings number without evidence. A frequently repeated task can still be a poor first project if the data is restricted or errors are difficult to detect.

4. Separate three kinds of help

Coaching: you want to learn a tool or repeat a workflow yourself.

Consulting: you need to compare opportunities, clarify requirements and decide the next scope.

Implementation: you need software, system connections, dashboards or automation built and tested.

You may need more than one, in sequence. A useful proposal identifies which work is included, who provides access, which deliverables are expected and how completion will be reviewed. Ongoing operation is a separate responsibility that needs an owner.

5. Write a first-version brief

Problem we are solving: ______

People who will use the result: ______

Input and its approved source: ______

Output and where it goes: ______

Human approval point: ______

Explicitly out of scope: ______

Example: “Organize approved sample inquiries into a review table” is a smaller first version than “run our sales process.” A clear boundary makes it easier to test whether the proposed tool solves the problem you actually have.

6. Check access and dependencies

Before connecting a live system, identify the account owner, available permissions and the minimum access the workflow needs.

Questions for the project:

• Is the data available in a form we can use?
• Does the business have permission to use it here?
• Do the relevant accounts support the required integration?
• Who handles exceptions and failed requests?
• What is the plan for removing access or stopping the workflow?

Avoid sharing passwords in a brief. Use supported account roles and secret storage during implementation. A successful demonstration with sample data does not establish production readiness.

7. Include the whole cost

Record the costs you can identify, and mark the rest as unknown:

Ask whether quoted work includes testing, handover and ongoing changes. Compare total effort with the current process rather than treating the AI generation step as the entire job.

8. Define acceptance before building

Choose a small set of approved examples and describe the result you expect.

For an inquiry dashboard, you might check that each sample appears once, missing details stay marked as unknown and only authorized people can access the records. For a draft-writing workflow, review factual accuracy, commitments, voice and next step.

Test exceptions as well as the ordinary path. Agree on who signs off, how corrections are handled and what would stop the launch. Your real criteria depend on the project and should be written into its scope.

9. Bring a useful consulting brief

Our priority process: ______

Evidence of the current problem: ______

Three candidate improvements: ______

The smallest first version: ______

Data and access questions: ______

Quality checks and reviewer: ______

Budget or constraints to discuss: ______

The decision we need help making: ______

Use this brief to make the first conversation specific. Costa AI Labs can help compare priorities and define a practical next step; implementation is scoped separately.

Explore consulting: https://costaailabs.com/ai-coaching#consulting

Start a conversation: https://costaailabs.com/contact?service=ai-coaching