AI and Simpro.

What Simpro’s own AI does (October 2026)

Simpro Lightning is Simpro’s “AI-First Operating Platform”. JustAsk, powered by Simpro’s AI Cooper, answers business questions in plain language: margins by technician, overdue invoices, customer history. Four agents are available: FieldReady, JobReady, JobScribe and JobBrief, covering technician training, job preparation, on-site documentation and customer summaries. The RAIN release adds an Intelligent AI Scheduler, which builds daily schedules from technician location, skills, licences, availability and job history, plus AI-Guided Forms; Simpro says these need the Lightning upgrade.

Simpro has an Auckland office and an NZ sales line, and publishes no prices. Its AI Pledge says it will not share customer data with third parties to train AI models without express permission. If you are on Lightning, use JobScribe and the scheduler before building anything for documentation or dispatch. The gap they leave is the front door: turning messy requests into clean, matched jobs.

Integration routes and when to use each

Automation 1: service request triage

Inputs

Emails and web forms reaching the service inbox, plus customer, site and recent job records read through the API.

Model step

Classify trade and urgency against written definitions, extract site address, asset and access notes, and match the customer and site. Flag maintenance-contract customers.

Repeat-visit check

Code looks for jobs at the same site and asset in the last 90 days and marks a possible callback, which is often unbillable.

Human approval

The dispatcher reviews a triage card (summary, match, urgency, history) and confirms or corrects before any job exists.

Write-back

On approval, the job is created through the API with a structured description and the original message attached.

Failure and recovery

If matching is uncertain or the API is unavailable, the message stays in the inbox tagged “triage failed”. The inbox remains the record, so nothing is lost.

Property managers face the same front-door problem; compare our maintenance request triage workflow.

Automation 2: after-hours voicemail to a morning queue

1. Transcribe. Each after-hours voicemail is transcribed with the audio kept.
2. Apply safety rules first. A fixed keyword list (gas smell, sparking, flooding, no power to medical equipment) sends an immediate alert to the on-call technician. This step does not depend on the model.
3. Extract. The model pulls caller, address, problem and urgency into a triage card.
4. Queue. Cards wait for the morning dispatcher with transcript and audio links.
5. Call back. A person phones the customer and books the job.

If you want calls answered rather than recorded, see AI voice agents.

Permissions and approval design

Register the integration as its own application under OAuth2, run it under a dedicated login with the least access your build allows, and log every write with the user and source message. With no official MCP server, build one with read tools and a single “create draft job” tool, following our MCP for business systems design. Nothing deletes, reschedules or invoices.

Job notes often contain gate codes, alarm codes and key-safe locations. Strip them before text reaches a model; they serve no purpose in triage and are a security risk if logged. Customer names and addresses are personal information under the Privacy Act 2020, and you remain responsible for them when a provider processes them for you.

Costs and limits

• Simpro: pricing on request; Lightning features may change your subscription.
• API limits: not published on the pages we could reach, and the docs sit behind a bot check. Queue requests and back off on errors.
• Model cost, worked example: classification suits a small model. 600 requests a month at 3,000 input and 400 output tokens on Claude Haiku 5.5 (US$0.10 and US$0.50 per million) is about US$0.30 a month.

Where it goes wrong

• Wrong build or company. The API works against a specific Simpro build, and Simpro asks for the build URL in every bug report. Pin and test the target build, and the company within it, before any write.
• Under-triaged emergencies. A calm email can describe a dangerous fault. Safety rules run before the model, and dispatchers see everything.
• Duplicate jobs. Reply-all threads create several requests for one fault. Group by email thread and site.
• Slow answers. Simpro’s API FAQ says its forum is monitored at least weekly, so allow lead time for questions.

Smaller trade businesses on ServiceM8 should read our ServiceM8 guide. For a scoped build, see AI integration services.

Written by the Levy Street team in Wellington. Last updated 19 October 2026.