AI Engine
AI drafts. A human decides.
AI features require provider credentials and a verified connection before use. Once configured, PowerOps can send an authorized photo to Anthropic Claude for a diagnostic-assistance draft and can draft a customer compliance report from stored equipment and recent work-order data. Model output can be wrong, incomplete, or unsafe; a qualified human must verify it before action or approval.
Key features
Photo-analysis draft
An authorized staff member uploads a photo and receives model-generated identification, likely-cause, corrective-action, possible-parts, and urgency text for human review.
Permission boundary
The photo endpoint checks Field Execution access before it calls the model, so an unauthorized request spends no model tokens.
Interaction log
The tenant, named account, model, token count, latency, media type, and returned draft are recorded for the interaction.
Compliance-report draft
The reporting workflow supplies the selected customer, equipment, and recent work orders. It instructs the model to flag missing data; the reviewer must check the draft against the source records.
Human approval
A compliance report remains a draft until an authorized person reviews, edits, and explicitly approves it.
Honest roadmap boundary
Predictive PM, quote assistance, automated anomaly detection, and weekly AI digests are not represented here as live capabilities.
What it gets you
- Give a qualified technician a structured model draft to evaluate
- Keep model calls behind the same section-permission boundary as the work
- Record who requested each interaction and which model produced it
- Draft only from stored customer, equipment, and recent-work-order data
- Call missing compliance evidence out as a gap to collect
- Require a human review and approval before a draft becomes an artifact
Example workflows
Review these steps against the job you want to run during your demo.
Photo-assistance review
A technician uploads a control-panel or damage photo, reads the model draft, and verifies every identification and action against the equipment, manuals, measurements, and safe work practices.
Missing-evidence review
A report draft identifies missing classifications, dates, or test results so an authorized reviewer can collect the evidence instead of accepting invented content.
Report approval
An authorized reviewer edits the draft, confirms the source records and governing requirements, then explicitly approves or rejects it.
Questions before your first job
Which AI model powers this?
The current photo-analysis and compliance-report routes are configured to use Anthropic Claude Sonnet (claude-sonnet-4-6). Verify the active route configuration and model availability when evaluating the feature.
Do I need my own API key?
The feature requires PowerOps to be provisioned with an Anthropic API key. Confirm plan access, limits, data handling, and availability before rollout.
How accurate is photo analysis?
PowerOps does not publish an accuracy rate. Image quality, visible context, model behavior, and equipment variation all affect the result. Treat every response as an unverified draft for a qualified technician.
Is my data used to train models?
Data handling depends on the current PowerOps and Anthropic agreements and configuration. Review the applicable privacy terms and customer obligations before submitting sensitive images or records.
Related reading
See the workflow and the company plan.
One flat company plan with every available module, unlimited office and field accounts, and guided onboarding included. We share the price on your demo or when a person reviews your access request. Your 14-day trial starts after access approval. No card required.