From the Utility Management edition of November 3, 2026
Welsh Water's ORAI control-room AI moves from proof of concept to MVP, assurance and parallel running
Dwr Cymru Welsh Water (not-for-profit) with FOIL AI and Microsoft · Water and wastewater operations control · Wales, UK
- Publisher
- Microsoft
- Original publication
- Microsoft customer story, 22 Oct 2026
- Source retrieved
- 2026-11-04
What happened
Welsh Water, which can receive more than 1,800 telemetry alarms across about 14 systems on difficult days, built ORAI with FOIL AI on Microsoft Foundry and Fabric. It gives evidence-led recommendations in Teams while operators keep accountability; each event opens an auditable thread. After a proof of concept it is in MVP delivery, assurance and parallel running. The business case forecasts about 10% improvement against regulatory compliance over two years and a 15-30 minute event-time reduction.
Why it matters
A not-for-profit utility running control-room AI under explicit human accountability and parallel running, directly comparable with SAM in this edition.
Evidence and measured results
Vendor case study. Benefits are business-case forecasts, not measured results.
Limitations and uncertainty
Vendor-authored; forecasts only; UK context; cloud dependency.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-11-03; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Operators collate data from many systems before assessing an incident.
- Stakeholders
- Control centre manager
- OT security lead
- Head of operations
- Regulatory compliance
- Discovery questions
- How many systems are consulted per incident?
- Is OT data allowed in cloud services?
- How are decisions audited today?
- Value hypothesis
- Pre-assembled evidence may shorten event assessment; to be measured in parallel running.
- Potential engagement
- Parallel-running evaluation design and assurance plan.
- Unsupported claims
- 10% compliance improvement
- 15-30 minute savings achieved
- Autonomous incident response
Pre-sales engineering
Role takeaway
- Fit
- Utilities with telemetry spread across many systems and cloud approval.
- Architecture and integration
- One-way OT data into a cloud data platform, AI recommendations in collaboration tools with audit threads.
- Prerequisites
- Telemetry data integration
- Cloud policy for OT data
- Incident history
- Constraints
- Vendor forecasts only
- Cloud availability in incidents
- UK context
- Security
- One-way export from OT, no write-back, audit logging, tenant isolation.
- Proof of value
- Parallel-run agreement with operators
- Assessment time comparison
- Audit trail completeness
Delivery
Role takeaway
- Work
- Integrate telemetry, build recommendation layer, assure, run in parallel.
- Dependencies
- Data platform
- OT security approval
- Operator time
- Ownership
- Operations control centre manager.
- Skills and adoption
- Operators trained to challenge and document overrides.
- Governance checkpoints
- Assurance before parallel run
- Parallel-run review
- Go-live sign-off
- Acceptance criteria
- Proposed: assessment time reduced against baseline
- Proposed: recommendations traceable to source data
- Proposed: documented operator override process
- Risks
- Automation bias
- Cloud outage
- Stale data
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Cloud AI (Foundry) over a data platform (Fabric) aggregating ~14 telemetry systems, surfaced in Teams with auditable event threads.
Governance
Who approves, reviews and stays accountable for outcomes?
Assurance and parallel running before reliance; auditable threads support review.
Security and privacy
What data, permissions and controls need testing?
Telemetry from OT systems flows to a cloud platform; segmentation and data flows are not described.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Reduces manual collation for control-room staff; operators retain decisions.
Procurement
What should contracts, pricing and exit terms secure?
Structured vendor selection preceded build; US utilities would weigh cloud OT data policies.
Operating model
Which teams own the service once it runs?
Recommendations delivered inside existing collaboration tools change control-room workflow.
Publication history
- 2026-11-03Utility Management · Issue 066 resources
Stable resource ID: welsh-water-orai-control-room-ai-alarm-triage-2026