Each rectangle represents a single white-collar job found on a typical construction project or within a contractor, consultancy, or employer's organisation. The area of the rectangle is proportional to the estimated number of people working in that role across the UK construction sector. Larger tiles mean more people.

The colour runs from deep green (score 1, very low AI risk) through amber (score 5, moderate) to deep red (score 9, near-certain automation of most tasks within five years). Roles are grouped into 14 categories: Commercial and Cost, Project Management, Claims and Disputes, Planning and Programming, Site Technical, Digital and BIM, Compliance and Safety, QA/QC, Procurement, Finance and Accounts, HR and Administration, IT and Systems, Marketing and BD, and Legal and Governance.

Hover over any tile to see the role's full risk assessment, including a plain-language rationale explaining why that score was given. The breakdown panels beneath the treemap show risk distributions by tier, category, and individual role.

How the Scores Were Derived

Each risk score was generated through a structured heuristic evaluation using a composite weighted-factor methodology. The approach combines task decomposition analysis (breaking each role into its constituent daily activities), cognitive complexity indexing (assessing the level of ambiguity, novelty, and professional judgement each task requires), physical presence dependency modelling (measuring the degree to which the role requires on-site, in-person execution), and interpersonal information asymmetry assessment (evaluating how much of the role's required knowledge can only be obtained through human-to-human interaction). These four dimensions were evaluated against current AI capability benchmarks and near-term (3-7 year) technology readiness projections to produce a single displacement susceptibility index on a 1-10 scale.

In practical terms, the logic behind every rating follows a five-question framework applied consistently to each role:

1. What proportion of the role's daily task list is repetitive, structured, and rule-based? Tasks like processing payment certificates, updating document registers, or producing standard cost reports are straightforward for AI to handle. The higher this proportion, the higher the score.

2. Does the role require physical presence on site? Construction sites are noisy, unpredictable, and dangerous. AI can process data remotely but cannot walk a site, inspect a pour, or talk to a banksman. Roles that require regular site attendance score lower.

3. Does the role depend on gathering information from multiple people through conversation, persuasion, or professional authority? A claims specialist, for instance, must track down the right people, ask the right questions, chase records that may not exist in any system, and piece together a factual narrative from contradictory accounts. AI cannot do any of that. Roles with heavy interpersonal information-gathering score lower.

4. Does the role require professional judgement in ambiguous or novel situations? Interpreting a contract clause in the context of a particular dispute, or deciding whether a design is buildable, requires experience-based reasoning that current AI cannot reliably replicate.

5. How mature is the existing AI tooling for this role's core tasks? Where commercial AI products already exist and are being adopted (e.g. automated takeoff software, AI-driven scheduling tools), the score is higher. Where AI tooling remains experimental or non-existent, the score is lower.

What the Scores Mean

Scores 1 to 3: Difficult to displace

These roles are defined by physical site presence, leadership of people, and the need to gather information through face-to-face interaction. Project Directors, Construction Managers, Site Engineers, Claims Specialists, and Health and Safety Managers fall here. AI may speed up their reporting or give them better data, but the core of what they do (managing people and situations in real time on a live construction site) sits well beyond current AI capability.

Claims work and delay analysis both sit at a score of 2. The claims specialist must visit site, interview supervisors and foremen, request and chase contemporaneous records (daily diaries, allocation sheets, progress photographs, RFIs, site instructions), determine what records are missing, work out whose evidence conflicts with whose, and then construct a coherent narrative tying the facts to specific contractual provisions. None of this can be done remotely by a language model. The information does not sit in a single database; it is scattered across emails, filing cabinets, personal notebooks, and the memories of people who may have left the project. The claims specialist's job is to find it, organise it, and make sense of it.

The delay analyst's work is similarly resistant. Although the programme software can run a critical path calculation, the actual job of forensic delay analysis on a disputed project requires the analyst to visit site, interview the planning engineer, the construction manager, and the subcontractors to establish what actually happened and when. They must reconstruct the factual sequence from conflicting accounts, incomplete records, and the practical realities of the construction methodology. AI can run the arithmetic, but it cannot visit site, interview witnesses, or assess credibility. Both roles sit at the bottom of the displacement risk scale.

Scores 4 to 6: Partial automation, role reshaping

These roles contain a mix of automatable routine work and tasks that are genuinely human. The quantity surveyor position splits sharply between contractor and client side. The contractor's QS (scored 4) spends a large proportion of their time on site, walking the works, agreeing measurements face-to-face with engineers and subcontractors, and extracting information from the people doing the work. That physical, interpersonal requirement provides strong protection. The client-side or PQS quantity surveyor (scored 6) works more from the office, producing cost plans, benchmarking, and payment assessments from documented applications. More of their daily output is structured and data-driven, making them moderately more exposed.

The same logic applies to commercial managers. The contractor's commercial manager (scored 3) is regularly on site, dealing directly with subcontractors on problem packages and making real-time commercial decisions based on what they can see and hear. The client's commercial manager (scored 5) is more office-based, focused on cost reporting, change control, and reviewing contractor applications.

Contract Administrators score 5. The administrative portion (issuing notices, maintaining registers, tracking deadlines) can be automated, but the interpretive portion (advising on entitlement under the contract, managing interfaces between parties) cannot. Compliance Officers and BIM Managers sit at 6. In both cases, the monitoring and checking functions are well suited to AI, but the human oversight, standards management, and exercise of professional discretion keep the roles alive in some form.

Scores 7 to 9: Substantial displacement likely

These are roles where the bulk of daily work is structured, repetitive, and data-driven. Document Controllers score 9: almost everything they do (maintaining registers, tracking transmittals, enforcing naming conventions, chasing overdue submissions) can be handled by an automated system. CAD Technicians score 8: AI-generated drawings from 3D models and automated detailing are already reducing the volume of manual draughting required. Estimators score 7: AI can read drawings, produce quantities, and apply rates from historical databases faster than a human can.

In the finance and accounts category, Payroll Administrators and Accounts Clerks score 8. Payroll processing, invoice matching, and ledger posting are among the most straightforward tasks for AI. Project Accountants and Management Accountants score 7, because while the routine bookkeeping is automatable, cost reporting to project teams still requires some human interpretation and presentation.

No role scores 10

No role on this chart received a score of 10. Even the most automatable roles have some residual human element, whether that is handling an exception, making a phone call to resolve an ambiguity, or exercising a small amount of discretion. A score of 10 would imply total automation with no human involvement at all, and that is not realistic for any white-collar construction role within the next five to seven years.

Sector-Wide Observations

The weighted average risk across all 63 roles comes out at roughly 4.7 out of 10, suggesting that the construction sector's white-collar workforce faces moderate overall exposure. But this average conceals a sharp split. The pain is concentrated in specific functions: document control, draughting, estimating, scheduling, payroll, and junior administration. These are roles that employ large numbers of people at the lower end of the salary scale, and they are the ones most likely to see headcount reductions within the next few years.

At the other end, roles defined by site presence, stakeholder management, professional judgement, and information-gathering from people are well insulated. The construction industry's physical, fragmented, and adversarial nature is itself a form of protection. AI works best with clean data in structured environments; construction sites are neither clean nor structured.

A likely outcome is that mid-level professional roles (QS, contract admin, planning) will be reshaped rather than eliminated. Fewer people will be needed, but those who remain will be expected to handle the higher-value, judgement-intensive work while AI handles the volume processing. This is already happening in estimating departments, where AI takeoff tools are reducing the need for junior estimators while increasing demand for senior estimators who can review outputs and apply commercial judgement to tender strategy.

For individual professionals, the practical implication is straightforward: the more your daily work involves talking to people, being on site, exercising professional judgement in ambiguous situations, and piecing together information from multiple imperfect sources, the safer your position. The more your work involves processing structured data, producing standard documents, or maintaining registers, the more exposed you are.