Wrench Scribe

Wrench Scribe is an AI workflow product for maintenance teams. It converts unstructured service notes into structured work order language that technicians can review, edit, and submit quickly.

The workflow was designed for the people who create, review, and depend on maintenance records across daily plant operations.

• Maintenance technicians closing corrective and preventive work
• Supervisors responsible for record quality and repeatability
• Planners and reliability teams that depend on clean maintenance history

In many plants, maintenance notes are written under time pressure and vary by person, shift, and site. That inconsistency makes records hard to trust and hard to reuse. Technicians needed speed at close-out, while supervisors and planners needed clean records they could rely on.

Documentation quality directly affects planning, repeat-failure analysis, and preventive maintenance decisions. If the workflow adds friction, adoption drops. If quality is weak, downstream teams lose signal. The product had to solve both.

The product pairs structured input capture with AI drafting so outputs fit work order workflows instead of free-form chat. Human-in-the-loop review is built in so technicians keep final control. Access was designed to be anonymous-first and low-friction, with privacy-first handling for operational context.

• Optimized for structured outputs compatible with work order workflows, not open-ended chat responses
• Invested in reliability hardening and error handling before broad rollout
• Balanced latency and cost through iterative model, prompt, and orchestration tuning
• Protected editability so users could correct edge cases without losing speed gains

I owned product direction from scope definition through launch readiness, while working closely with engineering on system behavior, reliability, and evaluation quality.

• Defined product scope, success criteria, and release priorities
• Led user and domain framing based on maintenance operations constraints
• Partnered with engineering on prompt design, schema strategy, evaluation, and reliability
• Drove launch readiness, instrumentation, and post-launch iteration planning

• Fast output only matters when users trust the draft, so editability and consistency were essential adoption levers
• Strict structure improved downstream usability but needed careful handling of uncommon edge-case language
• Operational AI products require product judgment on reliability and workflow fit, not only model quality

This public page focuses on product decisions and measurable results. Interface walkthroughs are shared directly in interviews because the workflow includes customer-sensitive operational detail.

• Technician input capture and draft generation flow
• Human-in-the-loop review and edit path before submission
• Structured output handoff for work order and planning workflows