Execution AI: Writing Custom Solvers

Last updated: July 28, 2026

Use CLEAR Prompting Framework:

C - Context: Set the scene. What is the primary action? Define the deliverable and be clear what you want the AI to do.

L - Limit. Set boundaries. Give it length limits, tone preferences, and things to avoid. (2-3 sentences or within the last week)

E - Example. Show the format or style you want. (Format in a bulleted list.)

A - Audience. Who is this for? A 5 year old? A CEO?

R - Role. Give the AI a persona. (Act as a maintenance technician..)

Iteration is Important

AI prompting is a conversation, not a one-shot command. Use your first result as a baseline, evaluate what worked, and ask the AI to adjust.

Writing Effective Prompts (The "Better" Way)

When writing a custom solver, specificity is your best friend. Use this table as a guide:

Goal

Avoid This (Too Vague)

Use This (Specific)

Root Cause

"Why is this machine breaking?"

"Review the 'Action Taken' notes for the last 10 breakdowns and identify the most frequent mechanical failure."

Team Performance

"How is my team doing?"

"Analyze the dispatches from last week. Identify which technicians have the fastest repair times for electrical issues."

Inventory

"Do we need more parts?"

"Based on the Spares History, flag any parts that have been used more than 3 times in the last 30 days on Asset [X]."


Pro-Tips for Power Users

  • Talk, Don't Type: If you're on the shop floor, use the Microphone icon in the chat input. You can dictate a custom solver while standing right at the machine.

  • The Feedback Loop: If the result isn't quite right, don't give up! Use a follow-up question like, "That's good, but now present that data in a table format with a column for 'Recommended Action'."

  • Save & Share: If you're a System Manager and you've written a "perfect" custom solver, you can save and share it with your entire team via the Setup menu. This ensures everyone is looking at the data through the same lens.