Platform · AI Controls

AI you control. Fully.

Disable it completely. Use our constrained managed AI. Or bring your own Anthropic, OpenAI, or Azure OpenAI API key. You decide how AI touches your maintenance data — not us.

See it in a demo

Why skepticism is rational

The practitioners said it. We listened.

These are not edge cases. They are the majority view across the field research we built this product on.

“They ended up making business decisions based upon an AI looking at data generated by an AI and not any reality.”

— Dr. Howard Penrose, MotorDoc

“It only takes one or two of those hallucination cases to get people to never use the tool again.”

— Roshan Satish, MaintainX AI Lead — candid self-assessment

“If you don't have high quality data, you can run the most sophisticated AI approach on it — the results will be disappointing.”

— Peter Jurcso, CERN — 30+ years in CMMS

“Do not trust and absolutely verify.”

— Dr. Howard Penrose — the governing rule practitioners endorse

Three modes. One toggle.

Set at the organization level. Override per site or per role.

Off

Zero AI. No models. No exceptions.

AI processing is completely disabled. No language models, no embeddings, no external API calls. Your data never touches an AI system. Appropriate for highly regulated industries, air-gapped environments, or teams that have simply decided AI is not for them right now.

Includes

  • No model API calls — ever
  • No embeddings or vector storage
  • No data sent to third-party AI providers
  • Audit log confirms zero AI events

Good for

  • · Regulated industries (pharma, defense, healthcare)
  • · Air-gapped or classified environments
  • · Teams that failed with AI before and want a clean slate

Managed

Constrained, honest, human-in-the-loop.

Our AI features — constrained entirely to your CMMS data. No external web knowledge. No hallucination from training data. Human review required before any AI output becomes an action. We took 12+ months to build this deliberately, following the FlowPath model of constrained-AI-only.

Includes

  • Scoped to your hub data only
  • Human confirms before any AI action is taken
  • Anomaly detection on meter readings
  • Close-out note summarization (voice → structured)
  • Asset Q&A on work order history and documents
  • Shows "not enough data" rather than guessing

Good for

  • · Teams that want AI assistance without the trust risk
  • · Mid-market manufacturing and facilities
  • · Organizations building their data foundation first

BYO Provider

Your key. Your billing. Your data residency.

Bring your own Anthropic, OpenAI, or Azure OpenAI API key. You control the model version, the rate limits, the billing account, and the data residency region. Sympl CMMS passes your key through to the provider — we never store credentials or intermediate outputs.

Includes

  • Anthropic Claude, OpenAI GPT, or Azure OpenAI
  • Your API key — never stored by us
  • Data residency in your chosen provider region
  • Model version pinning (don't auto-upgrade)
  • Same features as Managed mode, your infrastructure

Good for

  • · Enterprise teams with existing AI contracts
  • · Organizations with data residency requirements
  • · Teams that want to control AI spend and model versions

Honest about what AI can and can't do here

We don't lead with AI. These features exist because practitioners asked for them specifically.

We build (narrow, honest)

Meter reading anomaly detection

"This usually reads 23.5 — you entered 235. Did you forget the decimal?" Constrained output, no hallucination risk.

Close-out note summarization

Technician records 30-second voice note → AI generates structured summary (symptom, diagnosis, resolution, recommendation) → tech edits before saving.

Asset history Q&A

"What was done last time this motor tripped?" Returns relevant work order history and uploaded documentation. Scoped to your data. Says "not enough history" when data is thin.

Nameplate OCR on asset creation

Photo of data plate → AI parses make, model, serial, specs → tech confirms before saving. Solves the 29-character alphanumeric serial number problem.

Job plan generation from uploaded manuals

Upload a PDF maintenance manual → AI extracts relevant tasks → proposes PM procedure outline → human reviews and edits. Zero hallucination risk when grounded in your documents.

We never build

Auto-create work orders from AI

Every AI-suggested action requires human review and explicit approval before it becomes a work order.

Predictive failure claims without data prerequisites

We don't claim "this will fail in 30 days" until an asset has 12+ months of clean, audited meter data.

Train on synthetic data

The "AI-on-AI data fabrication" failure mode is real. We never fill data gaps with synthetic data or train models on AI-generated outputs.

External web knowledge injection

Our AI Q&A is scoped to your hub data only. No external web scraping, no training data bleed-through.

See how AI controls work in your context

We'll walk you through all three modes with your vertical's data as the example.

Request a demo