Permanent plug & abandonment · Offshore wells

You only abandon
a well once.

PP&A is the last operation a well will ever see — and the one that has to hold for geological time. endwell.ai is the coordination backbone that makes it provable: one live model of the well, shared by every engineer, robot, and AI agent, from first diagnostic to final verified barrier.

Wellbore schematic of NSEA-17A with verified primary barrier, secondary barrier in review, and planned environmental plug cap rock reservoir 0 800 1 600 2 400 3 000 PRIMARY · VERIFIED ✓ SECONDARY · IN REVIEW ENV · PLANNED NSEA-17A
Live wellbore schematic
2,000+ North Sea wells to plug this decade — with ~2,900 more queued on the Norwegian shelf alone
$100B+ forecast to decommission the North Sea — the UK and Norwegian shelves combined
1 attempt. A permanent barrier is set once, then must hold for geological time
0 revenue upside. PP&A is a pure cost center — efficiency and certainty are the whole game

Figures: Norwegian Offshore Directorate · North Sea Transition Authority (UK) · Wood Mackenzie

— 340 m · below seabed

Getting it right is not optional.
It's the entire job.

Plugging a well is a regulatory obligation with no product at the end of it — only the proof that it was done properly. That proof is scrutinized for decades, and the consequences of getting it wrong outlive every team that worked on it.

New to PP&A? Watch the regulator's primer on permanent well abandonment (Norwegian Ocean Industry Authority) ↗
COMPLIANCE

Audit-ready, or it didn't happen

NORSOK D-010 and its equivalents require every permanent barrier to be defined, verified against acceptance criteria, and documented. That documentation is an unbroken chain from requirement to evidence to approval — and when it lives in PDFs, inboxes, and spreadsheets, assembling it becomes a project of its own while campaigns burn real rig time by the day.

COST

Small errors, priced in vessel days

Wells are roughly half of every decommissioning bill, and campaigns burn rig and vessel day rates while teams reconcile conflicting datasets and redesign barriers late. Even single-digit gains in planning and execution translate into billions basin-wide — and every avoided re-entry is worth an order of magnitude more.

ENVIRONMENT

A seal that must outlast everyone

A failed barrier is a leak path with consequences measured in decades — environmental, legal, and reputational. There is no cheap second visit to an abandoned well. The only economical time to be certain is before the final plug is set, which is precisely when the data is most fragmented.

— 900 m · surface casing shoe

No one can see the whole well.

The data needed to abandon a well correctly — subsurface, casing, cement, integrity history, inspection reports, operational logs — exists. It's just spread across systems that were never designed to agree with each other.

today Document-driven PP&A

  • XLSXBarrier schedules in versioned spreadsheets — final_v7_REV.xlsx
  • PDFVendor inspection reports, findings locked in static pages
  • SPThe "official" record in SharePoint folders, three renames deep
  • APPDiscipline suites — strong in one silo, blind to the rest
  • LAKEIn-house data lakes that never reach engineers as a daily tool
  • HEADSInstitutional knowledge that leaves with every crew rotation

Decisions anchored on partial or stale information; expert hours spent reconciling versions instead of improving the abandonment design.

with endwell One live well model

● synced Graph connecting well, barrier, requirement, task, evidence, and finding nodes WELL BARRIER REQ TASK EVIDENCE FINDING
  • Barriers, requirements, tasks, evidence, findings — linked, not filed
  • Continuously synced across disciplines, vendors, vessels, and the field
  • Every change carries who, on what evidence, pending which approval
  • Uncertainty stays explicit and measurable instead of buried in a PDF
— 1 600 m · intermediate casing

One well. One model.
Every participant.

Humans, robots, and AI agents read from — and write back to — the same continuously updated model of the well, through the same transactional, auditable interface. That's architecture, not configuration.

01

A live well model, provenance built in

A continuously synced graph connects well state, barriers, inspections, evidence, tasks, and decisions to the requirements they must satisfy. The audit trail for NORSOK D-010 is a property of the data — not a report you assemble in a panic before the regulator asks.

02

Edge AI, where the data lives

Local-first by design: full function on vessels and rigs with intermittent connectivity and strict data residency. On-device models give operators sub-second feedback, keep sensitive asset data private, and cost a fraction of cloud-only inference at fleet scale.

03

Robots & AI agents, first-class participants

An engineer's open question becomes a task for a field system; the finding posts back as a structured, signed observation on the exact barrier it concerns. Nothing autonomous enters the official record without a human approval gate. Ever.

REQUIREMENT EVIDENCE NEEDED TASK FIELD CAPTURE FINDING HUMAN REVIEW OFFICIAL RECORD ✓

Traceability-first compliance: every barrier decision defensible, for decades.

— 2 200 m · above cap rock

What it looks like in the field.

A subsea North Sea well, days from its next PP&A operation. The operation is blocked until wellhead access is verified — so the verification becomes a task, the task goes to a robot, and the evidence comes back into the record. Watch the chain close:

NSEA-17A · #well-ops
09:41

system · compliance check CC-114 subsea access verificationEVIDENCE NEEDED
Operation OP-07 set surface plug is blocked pending verified access.

09:42

M. Reyes well integrity
@InspectionBot inspect the subsea wellhead guide base, valve panel, connector, and ROV access point for corrosion, debris, marine growth, obstruction, or leak indicators before PP&A operations continue.

09:42

InspectionBot field system
Task accepted → TASK-231. Beginning external scan. ● video feed live

09:47

InspectionBot
Finding F-089ROV access point partly obstructed by debris.
1 image captured · on-device detection · confidence 0.93 · signed sha256:9f3c…

09:47

system · CC-114NEEDS REVIEW · evidence linked: F-089, IMG-1142, TASK-231

09:58

M. Reyes
Reviewed. Debris removal scheduled as OP-06A. Approving finding into the record.

09:58

system · CC-114VERIFIED ✓
chain closed: requirement → task → evidence → finding → review · recorded with full provenance

The robot contributed evidence into the same graph as the PP&A plan — and a human stayed the decision gate. Every step above is now part of the auditable record.

— 2 600 m · cap rock

Why now.

  1. 01

    The wave has arrived

    Thousands of North Sea and mature-basin wells enter PP&A this decade. The work can't be deferred — it's a legal obligation on assets whose production is declining on the same balance sheet.

  2. 02

    Scrutiny is tightening

    Regulators keep raising the bar on barrier verification and environmental risk — the UK's NSTA is now investigating missed plugging deadlines, publicly naming operators that fall behind, and can fine up to £1m per breach. Evidence standards that were survivable with documents are becoming untenable without traceability.

  3. 03

    Robots are ready — their data layer isn't

    Autonomous inspection and robotics can now do real work downhole and on deck. But they need a structured, continuously updated picture of the well to read from and write back to. Document stacks can't provide one. We start there.

— 3 048 m · total depth · barrier envelope verified ✓

See the whole well.

Try endwell, or let us walk you through it on a well shaped like one of yours. And if you're an operator — or a specialist PP&A / well-integrity service provider — with an active or upcoming campaign, ask about our design-partner program: a scoped pilot on one concrete workflow, such as evidence assembly and review for a barrier decision.

Design-partner enquiries and everything else: hello@endwell.ai

endwell.ai

All's well that ends well.