27% Embedded, the Rest Catching Up: APM's New Numbers on AI in UK Project Delivery

For most of the past decade, the conversation about AI in project management has run on adoption curves and breathless predictions. The Association for Project Management's latest research, conducted with Censuswide and based on 1,000 UK project professionals, shifts that conversation onto firmer ground.

The headline figure is 27%. That is the share of UK project professionals who say AI is now fully embedded into their workflows. The number itself is the smaller part of the story. What it signals is the larger part. The discipline has crossed the line from experimentation into structural integration, and the sectoral patterns underneath the headline tell a more useful story than the headline does.

For Project Flux readers, the figure to watch is the sectoral split. Construction and engineering both register above the national average, which contradicts the long-running narrative that AEC trails every other industry on enterprise technology. We have been tracking this shift for months. The APM data crystallises it.

The APM survey, run with research firm Censuswide, polled 1,000 project professionals across UK industries, including construction, engineering, financial services, manufacturing, technology, and transport and logistics. Adoption is no longer concentrated in pockets of innovation teams. It is showing up in scheduling, risk forecasting, document review, and decision support across every sector studied.

His commentary paired the adoption data with a call for deeper practitioner capability in prompt engineering, ethical decision-making, and data-driven leadership. APM's response was a new prompt engineering learning module aimed at bridging that capability gap, which signals where the organisation thinks the next year of practitioner development needs to land.

The headline number is a national average. The sectoral breakdown is where the strategy questions live.

Construction sitting at the top of that list will surprise the AEC professionals who have spent years hearing the industry described as a digital laggard. It should not surprise anyone tracking the operational data. The 38% of contractors reporting measurable business impact from AI in the ServiceTitan 2026 Commercial Specialty Contractor Industry Report (covered last week in Project Flux) is the same trend told from the contractor side. APM's data tells it from the project management side. Both numbers point in the same direction.

The APM research identifies three workflow areas where AI is delivering the strongest returns in project delivery: scheduling, risk analysis, and document review. None of these will surprise practitioners. All three are pattern-recognition heavy, data-rich, and relatively forgiving of model imperfection because human review remains in the loop.

Scheduling has become a competitive battleground. The recent McKinsey partnership with ALICE Technologies, focused on generative scheduling for infrastructure and construction, claims schedule acceleration of up to 20% across more than 35 client deployments spanning data centres, energy, and manufacturing. McKinsey's broader research on generative AI productivity in knowledge-heavy roles puts the gain at 20 to 40%, and project management roles fit squarely inside that envelope.

The pattern Nwadigo describes matches the APM data. Adoption is outpacing the industry's ability to narrate it.

The APM data also surfaces the areas where AI is least embedded: stakeholder management and benefits realisation. Both gaps deserve more attention than they tend to get.

Stakeholder management is the human-relational core of project delivery. It depends on judgement, political reading, trust calibration, and the ability to hold ambiguity, all of which sit well outside the current capability frontier of language models. AI assistance in stakeholder workflows tends to land on the periphery: drafting communications, summarising meetings, surfacing past commitments. The substance of the work remains stubbornly human.

Benefits realisation is the harder gap. It is the longitudinal discipline of proving that delivered outputs translated into the business outcomes the project was justified against. AI should, in principle, be useful here. The data is structured, the questions are well-defined, and the comparison is quantifiable. The gap likely reflects something more cultural. Many organisations still treat benefits realisation as a closing-out exercise, and the AI tooling industry has prioritised front-loaded use cases like estimating, scheduling, and document review where the business case writes itself.

For senior project leaders, both gaps create opportunities. Firms that build differentiated capability in AI-supported stakeholder analytics or longitudinal benefits tracking are working in spaces where the leaders have not yet been declared.

92% of APM respondents reported feeling confident their current skill set is aligned with the changing demands of an AI-enabled workplace. 45% described themselves as "very confident". Those numbers should be read with caution. Confidence in tool use frequently outruns capability in tool deployment, and this gap is one of the most consistently observed patterns across enterprise AI rollouts.

The skills the APM respondents identified as becoming most critical for project managers are revealing:

The pattern is that the skills practitioners now consider most critical are predominantly human skills augmented by data fluency, and the only one focused purely on technical tooling sits at the bottom of the list. This is itself a maturity signal. The conversation has moved past early worries about AI replacing project managers toward the more useful question of what excellent project management looks like once AI is in the workflow.

For senior leaders in AEC and adjacent sectors planning the next 12 months of project delivery investment, the APM data points to three concrete moves.

The first is to treat AI capability development as part of capability development generally. The APM prompt engineering module is one signal of where the discipline is heading. Internal training programmes that combine prompt design with critical evaluation, data literacy, and domain expertise will compound faster than tool licences alone. Tool capability is converging across vendors. Building organisational capability around it is now the harder part.

The second is to build cost discipline early. The economics of AI tooling in project delivery are still being written. Token-based pricing at scale, particularly for agentic workflows, can produce cost trajectories that bear no resemblance to per-seat estimates. The Uber CTO's recent admission that the company's full-year 2026 AI budget was exhausted by April should reach every CIO planning AI rollouts at scale.

The third is to design governance before deployment. The APM respondents flagged transparency, accountability, and reliability of AI-generated outputs as the top ethical concerns. These are governance problems before they are tooling problems. They are best solved with clear accountability frameworks, documented review protocols, and named owners, all of which need to be in place before the workflow goes live.

The story APM's data tells is not about whether AI will reshape project management. That question is now closed. The story is about what excellent project leadership looks like once the discipline crosses 27%, then 40%, then 60% adoption. The firms shaping that question will be defining the practice for the next decade.

Subscribe to Project Flux for weekly briefings on what AI is doing inside project delivery teams across AEC, infrastructure, and complex capital programmes. We track the operational reality, the spending patterns, and the leadership shifts you need to see before they reach the trade press.

All content reflects our personal views and is not intended as professional advice or to represent any organisation.