Sector
Water and utilities
Asset knowledge sits in two estates: the telemetry and work management systems, and the drawings, surveys and permits no register was going to hold.

Where it sits
The corpora that matter here.
As-built drawings and schematics
Meaning carried in title blocks and annotations rather than prose, frequently scanned, frequently superseded without a clear marker.
Condition surveys and inspection reports
Heavily tabular. The row-to-asset relationship is the finding, and it is the first thing naive text extraction destroys.
Permits, consents and regulatory correspondence
Conditions attached to specific discharge points and abstraction sites, amended by letters filed separately from the permit.
Contractor handover packs
Often commercially restricted, which makes permission inheritance a deployment blocker rather than a refinement.
Challenges
What actually goes wrong.
Not the sector's problems in general, the specific failure modes that decide whether a system in this sector survives contact with production.
Superseded revisions
An answer drawn from a superseded drawing is worse than no answer. Revision lineage has to be an explicit pipeline stage, not a metadata field somebody hopes was filled in.
Asset identity across systems
One asset carries a site code, a legacy reference and two aliases. Retrieval that cannot resolve them returns a quarter of what it should.
Escalation as a bottleneck
When engineers cannot find something quickly, they raise a request. The specialist team becomes a queue, and the queue becomes the constraint on planning decisions.
Regulatory reporting cycles
Evidence assembly for periodic review is a scramble across the same corpora, done by hand, under a deadline that does not move.
Meter reads that arrive late, twice or never
A consumption figure built without a stated policy for missing intervals is one nobody can defend at a price review. The policy belongs in the pipeline, declared, not in the head of whoever built the report.
Job counts that never reconcile
The maintenance system, the asset register and the finance ledger each count work differently. Reconciling them by hand every month is a standing cost that a modelled join and one agreed definition remove.
Constraints
What shapes the architecture before a line is written.
Safety-critical accuracy
Answers inform decisions about buried and pressurised assets. Abstention on a low-confidence answer is not a degraded experience; it is the required behaviour.
Commercially restricted material
Contractor packs and legal correspondence sit in the same repository as general asset documentation. Access must be resolved per user at query time.
Long-lived evidence
A retrieval decision may need to be explained years later. Lineage from answer to source revision is a design requirement.
Engagement
How this looks in practice.

Water & utilities
Making thirty years of asset documentation answerable at a UK water utility
A retrieval pipeline over 380,000 asset documents: drawings, condition reports, permits and handover packs, with permission inheritance preserved from the source document management system.
UK regional water utility, ~4,500 staff · 16 weeks
Read the engagementServices that apply here
- Intelligence extractionRetrieval, extraction and enrichment over the documents, tickets, contracts and transcripts that never reached the warehouse.
- Data & platform engineeringWarehouse and lakehouse build, migration, ingestion, modelling, and the contracts that keep the numbers trustworthy enough to act on.
Start here
Talk to us about water and utilities.
Describe the estate and the decision it should be supporting. We will tell you which parts are tractable now, which need remediation first, and which we would advise against.
- Prefer email
- hello@vectisflow.com
- Response time
- One working day, from a person who has read it.