This website uses cookies
Read our Privacy policy and Terms of use for more information.
Overview
This agent confirms whether your facility handover package is complete enough to sign off before an asset enters operations. It gives you a completeness score per package and a gap-by-gap remediation list, with document-level evidence. It scores each package against the seven ISO 19650-4 Clause 7 information-review criteria (Common Data Environment compliance, conformance, continuity, communication, consistency, completeness, and domain-specific), maps documents to the CFIHOS V2.0 reference for asset-information completeness (over 6,500 document requirements, six contract-scenario templates spanning over 140 entities, six sequenced handover events), and checks build-to-operate continuity against the ISO 19650-3 Asset Information Model. It runs inside your IBM watsonx Orchestrate environment, and reads from and writes to your IBM Maximo MAS instance under your own authorisation.
What you provide
To connect the agent to your environment you provide three things, all entered through your per-tenant dashboard at https://dashboard.vistergy.ai/{instance_id} (the link is returned in the IBM watsonx Orchestrate provisioning response and never exposed publicly).
Maximo MAS instance URL. The base URL of your Maximo SaaS or on-prem deployment, for example https://yourcompany.manage.suite.maximo.com.
Authentication method. Choose one:
Service account (recommended for production). A dedicated Maximo user account with the read and write entitlements scoped to the affordances the agent surfaces. See “Maximo entitlements” below.
Maximo API key. If your Maximo deployment supports long-lived API keys, paste the key value into the dashboard.
SaaS browser session. Suitable for evaluation only. The dashboard accepts the cookie set from a recent Maximo browser session.
Credential rotation cadence. Vistergy recommends 90-day rotation for service accounts and Maximo API keys. Rotation is performed by your Maximo administrator, then refreshed in the same dashboard field; no Vistergy intervention required.
Maximo entitlements (service account or API key). The Maximo administrator grants the connecting account these scopes:
Read on handover packages, asset records, document attachments, and document-status records needed for completeness scoring.
Write on service request creation and work order updates (the affordances the agent surfaces). Writes never auto-fire; the agent surfaces the proposed Maximo action for your operator to confirm.
What Vistergy never asks for. The agent does not require, and Vistergy will never request:
Your IBM Cloud account credentials.
Credentials for any system outside your Maximo deployment.
Any data for model training. Customer queries are not used to train any model, ever. See the End User Licence Agreement and Data Processing Addendum for the full data handling boundary.
Supported queries
→ "Validate handover package [ID] against ISO 19650-4 Clause 7."
→ "What is the completeness score for handover event [ID]?"
→ "Which Clause 7 criteria does handover package [ID] fail?"
→ "List the missing CFIHOS V2.0 document requirements for asset [ID]."
→ "Create a service request to close handover gap [GAP-ID] on package [ID]."
The agent supports natural-language questions about facility-handover completeness and compliance. It does not advise on radiation, plant operations, or any regulated technical discipline.
Audit-bundle format
Every query produces a session-linked audit record capturing:
The audit bundle is exportable on request, scoped per handover event (the unit at which findings carry regulator meaning) in addition to per-tenant scoping. Format aligns with ISO 19650-4 Clause 7 evidence requirements.
Integration steps
Recommended setup in watsonx Orchestrate
There are two ways to use the agent.
Chat with it directly. It works as supplied. Every answer is grounded in the published standards, carries source citations, and an audit bundle is available on request.
Add it as a collaborator inside your own Orchestrate agent. If you compose it into an agent you build, apply the tested configuration below so your agent behaves the way this one was designed to.
Agent style: select ReAct. Compliance is a high-stakes application, and ReAct is the style IBM recommends for high-stakes applications.
Paste the tested instructions below into your agent's instructions, keeping the collaborator name as it appears in your collaborator list.
Starter prompts: use the supported queries above as your starter prompts. They are written to return grounded, sourced answers first time.
Leave "Chat with documents" off for governed use. Uploaded-document answers bypass the agent's source-traced retrieval.
Tested instructions:
Route every question in this agent's domain to the Vistergy Facility Handover Compliance collaborator, including questions about items you do not recognise, without asking a clarifying question first. Present its responses exactly as given, with their source attributions and any audit-bundle content. When it reports that something is not found or has no result, that is the answer: never turn a not-found into a status, score, or verdict, and never add a conclusion it did not state. It reports the system-of-record connection state honestly; surface that response verbatim and never imply a connection that does not exist. Never answer from your own knowledge. Keep responses concise.
For teams building with the watsonx Orchestrate Agent Development Kit, the same configuration as an importable agent definition:
spec_version: v1 kind: native name: my_handover_assistant display_name: My Handover Assistant llm: watsonx/ibm/granite-4-h-small # or your preferred watsonx model style: react instructions: | Route every question in this agent's domain to the Vistergy Facility Handover Compliance collaborator, including questions about items you do not recognise, without asking a clarifying question first. Present its responses exactly as given, with their source attributions and any audit-bundle content. When it reports that something is not found or has no result, that is the answer: never turn a not-found into a status, score, or verdict, and never add a conclusion it did not state. It reports the system-of-record connection state honestly; surface that response verbatim and never imply a connection that does not exist. Never answer from your own knowledge. Keep responses concise. collaborators: - vistergy_facility_handover_compliance # the agent as named in your instance
Behaviour and limits
The agent never auto-fires writes to Maximo. Where a query contains a write intent ("create an SR for..."), the agent surfaces the exact Maximo affordance for the customer to confirm.
The agent emits a structured completeness score per handover package (scored against ISO 19650-4 Clause 7 and CFIHOS V2.0, with document status interpreted across the Draft, Review, Approved, and As-Built hierarchy) and natural-language responses for explanatory queries.
Where retrieval cannot find a confident answer, the agent returns a guarded boilerplate rather than fabricating a response. Queries outside the standards-grounded scope return a deterministic refusal that does not invoke the language model.
Service Level: see the .
Status: see the .
Terms: see the .
Quality management: Vistergy is certified to ISO 9001:2015 (certificate 526732026). Details at .
Last updated: 4th July 2026