Where should an electrical business start with AI?

By Ivan Barnett · 9 September 2026

Start with the owner. The goal is to free up your time for work that moves the business forward.

Choose a repetitive task that keeps coming back to you. Then give AI the information and instructions you would give an employee taking it over.

1. Pick one repeated task

What do you keep doing yourself? Finding information, preparing a document, checking job notes or answering the same questions?

Choose something that takes meaningful time and has a result you can check. This applies to industrial, commercial and residential electrical businesses.

2. Define the job

Write down four things:

• Process: What work needs doing?
• Trigger: What starts it—an email, a completed job or your instruction?
• End goal: What should be finished?
• Good outcome: What must the result contain, and how will you know it is right?

For example: when job notes arrive, prepare a draft handover using your template. Include the work completed and outstanding items, flag missing details and leave it for review.

3. Brief it like an employee

If someone started this job tomorrow, what would you give them?

Gather the relevant instructions, business knowledge, templates and a good example. Explain where the information comes from, which version to use and what decisions need to come back to you. Start with sample or redacted records while testing.

4. Ask AI to help build the workflow

Use an AI tool suited to doing the work, such as ChatGPT Work, Claude Cowork, Codex or Claude Code. Which one fits depends on the task and the files, tools and access it needs.

Give it the brief and supporting information. Ask:

Help me turn this into a repeatable workflow. Ask me any questions you need answered before starting. Flag missing information instead of guessing. Produce a first draft for me to check.

Ask it to explain any missing software access or setup. Describing a trigger does not connect it automatically.

5. Test, correct and try again

Check the output against your example and definition of a good result. Be specific about what needs fixing: a missing detail, the wrong source, an unclear instruction or the format.

Don't give up after one poor answer. Explain the problem, update the instructions and test again on a different example. Include a case with missing information.

Think of briefing a new employee: feedback helps when it becomes part of how the job is done. Save corrections in the workflow's instructions; AI does not necessarily remember or learn them automatically.

If the same error keeps returning, narrow the task or change the approach. The aim is a dependable result, not endless retries.

6. Save it and hand over more gradually

Ask the AI to save the tested instructions, templates and checks as a reusable skill or workflow, where supported. Run it again yourself before connecting an automatic trigger.

Keep review in place while you build confidence. Measure the time spent checking and correcting as well as producing the output. Expand as the workflow proves useful.

Think beyond the obvious

Don't stop at emails and quotes. Look at research, tender preparation, project handovers, finding business knowledge and understanding job margins.

AI can help with a wide range of business processes. Ask what would be needed to make your idea work, then test a manageable first step. Your tools, information and the decisions involved determine how much you can hand over.

Want help getting started?

Flowstate helps identify where AI could make the biggest difference, then builds and tests workflows around your processes, knowledge and existing software.