Inside rahd·AI, the company betting decommissioning is a data problem first

Somewhere on the UK Continental Shelf sits a well drilled in the 1970s. The company that drilled it no longer exists, and the licence has changed hands several times since. Its casing specifications are recorded in a typed report that was later scanned and filed alongside thousands of others.

Before that well can be permanently plugged, an engineer has to establish what is inside it, what happened to it across 40 years of intervention, and where the barriers sit. The information exists somewhere in the archive. Finding it, cross-checking it and then standing behind it in a regulatory submission can absorb weeks of senior engineering time, and that time is billed by the day.

Around 1,000 wells fall due for plugging and abandonment on the UK shelf between 2026 and 2030. That arithmetic is the reason a company founded in Perth, Western Australia, moved its innovation hub to Aberdeen, and it explains why its chief executive talks about balance sheets more readily than about technology.

“rahd·AI is an industrial AI company transforming complex engineering information into trusted operational intelligence for the energy industry,” said Jake Stride, chief executive and co-founder. The company commercialised first in decommissioning, he explained, because it is one of the industry’s largest financial challenges, with more than $200 billion expected to be spent globally over the coming decades, and because those liabilities “sit on the balance sheet for years, tying up capital that could otherwise be invested into new projects and the energy transition.”

The bill has grown faster than the industry has cut it

The numbers behind that argument have moved in the wrong direction for several years. The North Sea Transition Authority’s Decommissioning Cost and Performance Update, published in July 2025, put the remaining UK Continental Shelf decommissioning bill at £44 billion, with well plugging and abandonment accounting for around half of the total. Operators expect to commit £27 billion of that between 2023 and 2032, more than half the remaining forecast cost.

Delivery has slipped behind schedule. A backlog of more than 500 wells has already missed original decommissioning deadlines. In late 2025, the regulator published its first table naming 13 operators who had missed consent deadlines across 153 inactive wells, a public escalation from an authority that had previously preferred private correspondence with licensees.

Because decommissioning spend is deductible against tax, the delay costs the state as well as the industry. The regulator’s February 2026 estimate puts the remaining Exchequer cost at £11.7 Billion, made up of £5.8 Billion in tax repayments and a £5.9 Billion reduction in offshore corporation tax. The UK government target of a 10% cost reduction by 2026 and 35% by 2035 has so far lacked an obvious delivery mechanism. rahd·AI reports validated savings of 15% during 2025 and has said publicly it believes 35% is reachable by 2027.

Stride is unusually direct about where the difficulty actually sat. “The biggest challenge wasn’t building AI, it was building trust,” he said.

Engineers who approve a well abandonment programme carry personal and corporate liability for the decision. A system that produces a confident answer with no visible workings is unusable in that setting, however accurate it happens to be. The company made provenance a structural property of the platform, so every answer stays traceable to the source document while engineers retain control of validation and approval. “Engineering and regulatory decisions cannot rely on ‘black box’ AI, so we developed the platform alongside operators,” Stride said.

That constraint explains a pattern seen across industrial software generally, where adoption stalls at the point where somebody has to sign. Buyers in energy, aviation and pharmaceuticals have funded trials readily and moved into production rarely, because approval workflows require a named individual to defend a decision in front of a regulator. Traceability turns a system’s answer into evidence that the individual can actually defend.

Charging for the pilot changed who turned up

The company built through paid Rapid Implementation Programmes, deploying into live campaigns alongside operators. “This has allowed us to validate customer use cases in live environments, prove measurable outcomes and ensure our roadmap is driven by genuine operational challenges rather than assumptions,” Stride said.

The approach produced named industrial validation quickly. A programme run with the Net Zero Technology Centre involved six energy majors including Equinor, Harbour Energy, TotalEnergies, Repsol, Petrobras and ConocoPhillips. Ben Foreman of TotalEnergies, quoted by the company, said early evidence showed the operator tracking towards savings “in the region of 10% across decommissioning campaigns”, adding that the work strengthened cost management and reduced balance sheet liabilities.

Money acts as the filter here. Unpaid pilots accumulate reference logos while consuming engineering attention on the vendor side, and industrial software is littered with companies that mistook a free trial for demand. An operator who pays for a programme also allocates internal engineering time to it, and engineering time is the scarcer of those two currencies.

rahd·AI’s platform draws on what the company describes as the largest validated dataset of its kind, covering more than 15,000 wells, 4,500 platforms and 2,000 pipelines, assembled from environmental statements, field plans and government and agency records. That material is cleaned, curated and structured through a proprietary management layer before the platform can use it.

Stuart McLeod, managing partner at Ventex, the Aberdeen climate tech venture studio behind the company, has described a sector burdened by “decades of unstructured data that complicate the safe and efficient shutdown” of infrastructure. His observation identifies why general-purpose tools have made so little headway in the category. The scarce ingredient is verified engineering history, and no amount of computing power substitutes for records that were never digitised in a usable form.

Pricing per well ties revenue to the size of the liability

The platform is licensed as software as a service on a per-well basis, with tiered pricing for larger portfolios. Customers typically enter through a paid programme before scaling across a campaign. Early deployments delivered an 85% reduction in engineering effort, a result Stride treats as the foundation for a much larger commercial argument.

“By helping operators make better decisions earlier in the lifecycle, rahd·AI has the potential to remove hundreds of millions of dollars of future liabilities from company balance sheets while improving regulatory confidence and operational outcomes,” he said. Per-well pricing scales the company’s revenue alongside the customer’s own exposure, which keeps the return calculation legible to a finance director who has no interest in software productivity claims.

Australia, where the company started, shows the same picture in different waters. A 2025 estimate commissioned by the Australian government from Xodus Group found industry will spend $43.6 Billion on decommissioning to 2070, rising to $66.8 Billion once inflation is included, with around 55% of the work falling before 2040. In the United States, the Government Accountability Office found more than 2,700 wells and 500 platforms in the Gulf overdue under regulatory deadlines as of June 2023.

The Middle East became the fourth market in June 2026, when rahd·AI was selected as the only UK company among 12 admitted to Cohort II of the Presight AI Accelerator in Abu Dhabi, chosen from 376 applications across 62 countries. Presight, majority-owned by G42 and listed in Abu Dhabi, gives participants access to enterprise computing infrastructure, technical support from Presight, G42, MBZUAI and Microsoft, and the opportunity to pitch to the $100 Million Presight–Shorooq Fund I. The Gulf offers mature offshore assets, well-capitalised national operators and state-backed computing infrastructure in a combination few regions currently match.

What remains unresolved is the pace - reaching a 35% reduction by 2027 depends on adoption across operators who have collectively struggled to award plugging and abandonment contracts on time. Software can compress planning cycles and surface risk earlier, though it cannot manufacture rig availability or repair the supply chain capacity the regulator has already flagged as a concern. The company answers that most of the saving is created long before a rig is ever booked, in the weeks an engineer currently spends looking for a well that nobody can describe.