How AI can help close the capital delivery gap

Energy investment is at record levels, but most major projects still overrun. AI can boost delivery confidence across the capital portfolio and help leaders turn ambitious plans into performing assets that deliver sustained returns.

Across energy and infrastructure, investment is surging. The International Energy Agency expects global energy investment to reach $3.4 trillion in 2026, with around $2.2 trillion going to renewables, grids, storage, nuclear, low-emissions fuels and electrification. But can organisations turn that money into working assets, on time and on budget?

The track record should give pause. Research on megaprojects suggests nine in ten overrun on cost, schedule or benefits. The UK Government's own major projects portfolio (189 projects worth £924 billion) rates only 15% of projects ‘green’ for delivery confidence.

There is a similar picture emerging globally. A Boston University study of 662 energy infrastructure projects across 83 countries (worth $1.36 trillion) found that more than three in five projects ran over budget. This resulted in projects costing 40% more than planned and finishing almost two years late.

Used well, AI gives leaders what they need most: delivery confidence. That means knowing, with evidence rather than optimism, that they can turn the investment plan from strategic ambition into a credible delivery plan.

Deliverability is the new strategic prize

When capital is abundant and deliverability is scarce, deliverability is what attracts capital and certainty becomes worth more than speed.

For owners whose projects run late, higher financing costs and tighter regulatory settlements make every delay expensive. Revenue is postponed, penalties are triggered and confidence in the wider investment plan is shaken.  

Owners with a demonstrable track record of predictable delivery, by contrast, secure more favourable regulatory settlements, better supply chain terms, cheaper financing and more shareholder patience. They also become the client that suppliers and skilled people choose to work with. 

Take the UK Infrastructure Pipeline. The 734 planned projects and £718 billion of investment need a workforce of up to 706,000 people a year over the next five years. With everyone drawing on the same pool of engineers, contractors, equipment and manufacturing slots, those scarce resources will flow to the owners who are easiest to deliver for.

Why delivery risk is rising

Despite the shifting investment landscape, the familiar causes of failure still apply: optimistic business cases, weak early definition and risk passed down the supply chain rather than managed. But two newer pressures mean delivery risk can rise fastest in organisations with the strongest track records.

First, portfolio complexity is outgrowing delivery capability. Network and water companies built for high volumes of repeatable work are now taking on major programmes. Oil and gas businesses are combining new builds with renewal and life extension. Many are moving into asset classes they’ve never delivered before, with governance, controls and skills designed for a steadier, more traditional portfolio. 

Second, data is frequently fragmented across the lifecycle. Most owners already hold the data that would warn them of delivery risk: schedule progress, cost and change data, supplier lead times, lessons from past projects. But this data is typically spread across multiple systems, with varying levels of accuracy. That means warnings are often pieced together after the point where leaders could have acted.

The delivery confidence gap is not caused by a lack of project information, but by the difficulty of connecting it and acting on it in time.

How leading owners are using AI

Leading owners are using AI to make earlier, better decisions across the whole portfolio, not just within individual projects. They treat projects as repeatable investment products rather than one-off engineering endeavours. One global energy company we work with has moved to highly repeatable, modular projects and is encouraging its project directors to ‘think more like investment bankers’.

To get the most from AI, owners are connecting their project and portfolio data, and the most advanced are putting a digital thread in place first. This is a continuous link between optioneering, engineering, procurement, cost, schedule, risk and supply chain data, running from project initiation through to design, final investment decision, delivery and into operations. It sits over existing systems and gives leaders a clear line of sight from trusted data to decisions and outcomes.

Where AI makes the biggest difference

• With that foundation, owners can match the right type of AI to each task. Machine learning and predictive models are best suited to classifying and forecasting using past performance data. Generative AI models streamline drafting of contracts, consents, regulatory reporting and other repeatable deliverables. Agentic AI coordinates actions across systems, while people stay accountable and have the final say. Three uses stand out:
• Identifying which projects will slip, and why. Models compare live projects with how similar ones performed and flag the patterns that come before schedule slips occur, protecting key energisation and commercial operation milestones.
• Matching the pipeline to scarce capacity. Scenario testing shows where projects will compete for the same engineers, suppliers, equipment or outage windows, so leaders can optimise the sequencing of work before committing.
• Taking repeatable work off project teams. AI undertakes drafting and first-line assurance checks across repeatable tasks, from consent applications and land rights agreements to engineering drawings, change requests and payment applications, freeing project managers and engineers to focus on critical work.

Above all, AI enables leaders to make better decisions at a point when the outcome can still be changed. Take a portfolio of grid connection projects. If models show that a long-lead transformer order is following the same pattern as orders that delayed past projects, the re-sequencing of work or securing a second supplier can be undertaken before the delay shows up in the schedule. 

The same principles apply beyond delivery. National Grid Electricity Transmission (NGET) is using AI to understand individual asset risk and drive resilience across the network. We’re working in partnership with them to develop this capability.  

Technology is only part of delivery confidence - AI won't fix weak delivery discipline. To get real value from AI, firms still need key delivery enablers in place: fit-for-purpose governance and assurance, demand and capacity planning, effective procurement, local skills and design standardisation. 

Where to start with AI for delivery confidence

If you’re a board or portfolio leader looking to build delivery confidence, the starting question is not ‘where can we deploy AI?’ but ‘which decisions need better evidence?’. From there, focus on:
•    Assessing how existing systems are limiting the adoption of digital tools and AI
•    Pinpointing past successes with data or AI and identifying what made them work
•    Examining AI pilots that failed to change delivery outcomes and understanding why 
•    Tackling the risks that can be addressed today, rather than leaving them to surface later

Whatever your organisation’s level of AI maturity, the principle stays the same: start with the decision that needs to improve, then build the data, governance and technology around it.

Join the conversation

Over the coming months, we’ll be talking to capital leaders across sectors and regions about how they’re using AI in delivery: what’s working, what isn’t and what they’ve learned along the way. You may have used AI to spot a cost or schedule risk before it materialised. Or perhaps a promising pilot never made it into day-to-day delivery, or legacy systems and patchy data which held you back. Whatever your experience with AI, we’d like to hear from you. Get in touch via the link below.

Sources

• IEA, World Energy Investment 2026 
• UK Government, Infrastructure Pipeline update (2026)
• NISTA Major Projects Annual Report 2025-26 
• Reltio, 2024 Data Leader Survey
• Science Direct, Beyond economies of scale: Learning from construction cost overrun risks and time delays in global energy infrastructure projects

Key takeaways

1. Delivery, not funding, is now the constraint. Global energy investment is expected to reach $3.4 trillion in 2026, yet nine in ten megaprojects overrun. Only 15% of the UK Government’s major projects are rated ‘green’ for delivery confidence.
2. Deliverability is a competitive advantage. Owners with a track record of predictable delivery secure better regulatory settlements, cheaper financing and better supply-chain terms. They also become the client that suppliers and skilled people, both in short supply, choose.
3. AI can make better use of the warning signals owners already hold. Most owners have the data that could flag delivery risk early, but it’s scattered across multiple systems, so warnings are often pieced together too late. With connected, well-governed data, AI can identify those signals while there’s still time to act.
4. The biggest AI opportunity is across the whole portfolio. Leading owners treat projects as repeatable investment products, using AI to compare plans with past performance, spot early signs of overruns, resolve competing demands for resources and test sequencing options before committing.
5. AI is not a substitute for delivery discipline. Delivery still requires core enablers such as capacity planning, effective procurement, skills and design standardisation. AI strengthens these foundations rather than replacing them.

Frequently asked questions

What is delivery confidence in capital projects?
Delivery confidence means knowing, with evidence, that an investment plan will turn into working assets on time and on budget. It applies to the whole portfolio, not just individual projects. In the UK, only 15% of major government projects are rated ‘green’ for delivery confidence.

How can AI improve capital project delivery?
AI helps capital project owners spot delivery risk earlier and make better decisions while there’s still time to act. Applied to connected, well-governed project data, it enables owners to:

• compare plans with the delivery performance of similar projects;
• identify patterns that come before cost or schedule overruns;
• see where projects compete for the same resources; and
• test sequencing options before committing to one.

Why do major energy and infrastructure projects overrun?
Major projects overrun because of optimistic business cases, weak early definition and risk passed down the supply chain rather than managed. Newer pressures add to this. Portfolios are growing more complex than delivery capability can handle, and project data is fragmented across many systems, so warnings often surface too late to act on.

What do organisations need in place before using AI for project delivery?
Organisations need clean, connected data and clear accountability for the decisions AI informs. They also need core delivery enablers, including demand and capacity planning, effective procurement, local skills and design standardisation. AI won’t fix weak delivery discipline. Without these foundations, it simply surfaces the same problems faster.

Where should capital leaders start with AI?
Capital leaders should start by asking which delivery decisions need better evidence, rather than where they can deploy AI. From there, they can assess how existing systems limit AI adoption, learn from past successes and failed pilots, and build the data, governance and technology around the decisions that matter most.