Research behind the advice · AI Workflows
Start with the work, not the tool.
This is one of the lenses I use in advisory work. Fourteen Workflows organise 85 public Cases into a practical view of where AI is being used, how strong the evidence is and what still needs testing.
Of the 85 Cases, 73 sit in core built-environment work. The other 12 are benchmarks, adjacent-sector comparisons and business-support uses such as bid and training tools — useful context, but not evidence about designing, building or operating assets.
A starting point, not a recommendation
Three Workflows to look at first.
This sorts the 14 Workflows against your perspective and objective. It cannot see your data, your baseline or your people, so treat it as a way into the argument rather than an answer.
Consulting work, not job replacement
AI changes the shape of professional work.
These are familiar built-environment consulting roles, viewed through the Workflows in this library. The likely shift is from producing a first answer towards framing the question, checking the output, resolving exceptions and owning the advice.
Project & programme management
See risk earlier. Change action sooner.
Schedule, progress and risk signals can reduce reporting lag. The role still owns the trade-off, the intervention and the client conversation.
Relevant Workflows: Construction Progress; Schedule & Project Controls.
Cost management & quantity surveying
Test options faster. Defend the number.
AI can accelerate extraction, comparison and early optioneering. Commercial judgement, source checking, procurement and negotiation remain accountable work.
Relevant Workflows: Technical Reports; Design Automation; CRM & Commercial Intelligence.
Architecture, engineering & technical leadership
Generate more options. Accept fewer on trust.
Constrained design and calculation loops may move quickly where outputs can be verified. Assumptions, coordination, professional sign-off and liability do not disappear.
Relevant Workflows: Design Automation; BIM Optimisation; Digital Twins & Simulation.
Planning, environment & stakeholder advice
Handle more evidence. Keep the judgement visible.
Classification, validation and consultation analysis are promising bounded tasks. Statutory decisions, contested interpretation and public accountability remain human responsibilities.
Relevant Workflows: Planning, Permitting & Consultation; Sustainability Management.
Digital delivery & information management
Make information usable, governed and reusable.
As more work is assisted, this role becomes less about file administration and more about reliable context, access, quality and traceability.
Relevant Workflows: BIM Optimisation; Enterprise Knowledge & Workflow Assistants.
Asset & operations advice
Turn signals into better interventions.
Telemetry-rich work can support detection, triage and optimisation. The operating owner still decides what to inspect, change, defer or escalate.
Relevant Workflows: Asset Operation; Asset Inspection; Digital Twins & Simulation.
Survey, inspection & asset information
Cover more ground. Keep the evidence trail.
Vision systems and reality capture can make condition and progress information more available. A professional still validates the finding and decides the intervention.
Relevant Workflows: Asset Inspection; Construction Progress; Digital Twins & Simulation.
Bid, commercial & client development
Reuse knowledge. Keep the client judgement.
AI can support relationship visibility, proposals and repeated commercial research. It cannot decide what is credible, differentiating or right for a client.
Relevant Workflows: CRM & Commercial Intelligence; Enterprise Knowledge; Technical Reports.
These are role hypotheses, not workforce forecasts. The public evidence is much stronger for bounded task and Workflow change than for whole-firm headcount, utilisation or career-path outcomes.
Signals beyond direct built-environment cases
Adjacent evidence shows the direction of travel.
These cards are deliberately separate from direct built-environment performance evidence. They help explain how professional work may be reorganised, but they do not prove outcomes in a building, project or asset context.
Professional services analogue · confirmed
Delivery modelAI can become part of how expert work is trained, delivered and billed.
A&O Shearman's public Harvey relationship is a signal that professional firms may redesign service delivery, not simply give individuals a drafting tool. It does not establish an AEC business-model result.
Engineering consultancy · confirmed
Governed knowledgeUseful assistance needs usable, controlled context.
Mott MacDonald's internal assistant is a direct consulting signal: access paired with governed knowledge can scale. Public material confirms use, not answer quality or financial return.
AEC benchmark · measured
Acceptance gapCapability is not the same as accepted professional output.
Drawing-literacy benchmarks show why every role needs an explicit checking and exception route. A benchmark is not a live deployment or a productivity result.
Adjacent-sector signal: these cases help indicate an operating mechanism or direction of travel. They are not built-environment deployment evidence and are not included in the built-environment case count, productivity reading or economic-confidence assessment.
What the Workflow view reveals
The useful pattern is narrower than the hype.
The strongest evidence clusters around defined work with existing data, an observable baseline and a person who remains accountable for acceptance. These are examples, not universal productivity rates.
Progress & reality capture
2,964 hours / $342kA direct money claim, with an important boundary.
Track3D reports the labour saving on one Hensel Phelps airport project. Hensel-owned sources support repeated use, but do not independently audit the figures.
Planning & consultation
18.5 hours to 16 minutesA dramatic result on a tightly bounded task.
The gain applies to one consultation-summary task. It is not a productivity claim for the whole planning service or the statutory decision.
Design & technical reports
Capability is ahead of proofGenerating an output is not the same as accepting it.
Benchmarks and live tools are improving, while checking effort, professional acceptance, liability and realised economics remain the critical questions.
The Workflow directory
Browse the 14 Workflows.
Each card describes the work and how strong the evidence behind it currently is. The confidence readings are deliberately conservative: they reflect what the public sources actually support today, not what the technology may eventually do. Only a handful currently have enough evidence to justify writing up in depth, and Construction Progress is the first of those.
Sustainability Management
Carbon, climate, biodiversity, environmental impact and sustainability due diligence.
Confidence: low-medium for quantified productivity; medium for strategic relevance.
BIM Optimisation
Structured project data, model enrichment, content reuse and information standardisation.
Confidence: low-medium; evidence often proves readiness rather than accepted-output productivity.
Digital Twins & Simulation
Asset twins, system models, scenario testing, forecasting and earlier operational decisions.
Confidence: medium for process acceleration; lower for realised outcomes.
Asset Operation & Optimisation
Energy, water, maintenance, networks and live control systems connected to mature telemetry.
Confidence: medium-high for bounded telemetry-rich optimisation.
Asset Inspection & Defect Detection
Roads, rail, bridges, utilities and built assets: detection, review, triage and intervention.
Confidence: medium-high for bounded detection and triage with human review; weaker for decisions where mistakes could affect safety.
Site Safety & Compliance
Near-miss detection, access control, monitoring and stronger evidence for intervention.
Confidence: medium for relevance; low-medium for measurable productivity.
Construction Progress & Reality Capture
Site capture, progress tracking, BIM comparison, QA, reporting and rework visibility.
Confidence: medium; repeatable, but much of the metric trail is vendor-owned.
Schedule & Project Controls
Risk forecasting, sequencing, optioneering, delay recovery and programme trade-offs.
Confidence: medium; strategically relevant with buyer-side corroboration gaps.
Design Automation & Engineering Apps
Option generation, calculations, CAD/BIM assistance and reusable engineering tools.
Confidence: medium for capability and bounded use; low-medium for professionally accepted outputs and financial value at scale.
Planning, Permitting & Consultation
Application triage, validation, consultation analysis and officer decision support.
Confidence: medium for bounded support; lower for statutory decisions.
Technical Reports & Due Diligence
First drafts, document review, evidence extraction and expert workflow triage.
Confidence: medium for bounded work with human sign-off.
Enterprise Knowledge & Workflow Assistants
Internal search, project memory, governed assistance and reusable workflow tools.
Confidence: medium for adoption; low-medium for economic value.
CRM & Commercial Intelligence
Relationship visibility, bids, proposals, client intelligence and pipeline support.
Confidence: medium for direction; low for quantified productivity.
Robotics & Autonomous Field Execution
Autonomous layout, intelligent plant and narrow physical construction tasks.
Confidence: medium for narrow tasks; lower for broad construction economics.
A worked example
What a Workflow looks like up close.
Construction progress and reality capture has some of the clearest evidence of repeated use, so it is the first Workflow written up in full: the named firms using it, the figures they report, who published each figure, and what the evidence still cannot tell you.
It is also the clearest illustration of the honest limit that runs through all fourteen. The evidence can show you that a Workflow is being used elsewhere. It cannot tell you what it will be worth on your projects.
How to use this
Choose, baseline, test and measure.
- Choose a Workflow with material friction or value.
- Name the baseline, accepted output and accountable owner.
- Measure review effort, quality, downstream use and exceptions.
- Scale only when the evidence survives into operational or economic value.
Running that loop well is a discipline in its own right, and it is the part most organisations skip.