The earned value problem
Every construction project manager faces the same question at every progress meeting: what percentage of the work is actually complete? The answer is usually a combination of site walk observations, subcontractor self-reporting, and educated guessing. It is subjective, inconsistent, and frequently wrong.
The consequences of inaccurate progress reporting are real. Overstating progress leads to over-certification of interim payments — the contractor is paid for work not yet done, which becomes a cash flow problem if the project falters. Understating progress delays payment to subcontractors, damages relationships, and masks genuine achievement. Both erode trust between client, contractor, and design team.
Earned value management (EVM) provides a structured framework for measuring progress: planned value (what should have been done), earned value (what has been done), and actual cost (what it cost). But the "earned value" component — what has actually been built — requires objective assessment of physical completion, and that is exactly what AI can provide.
The traditional approach relies on site walks where the project manager or QS estimates percentage completion for each activity: "Blockwork to first floor — looks about 75% complete." These estimates are subjective. Two people walking the same floor will give different numbers. And on a large project with hundreds of activities, no one can assess everything in a single walk.
A subcontractor reports 80% completion on first-floor mechanical services installation. The project manager's site walk assessment is 70%. The monthly valuation depends on this number. Who is right?
Photo-based progress tracking
The simplest form of AI progress monitoring uses ordinary site photographs — the ones your team is already taking — to assess what work has been completed.
How it works:
-
Baseline definition. Before monitoring begins, each programme activity is linked to visual completion criteria. For example:
- "Blockwork to first floor" is complete when all walls on the first-floor plan are built to the underside of the floor slab, with all openings formed and lintels installed.
- "Suspended ceiling to Meeting Room 3" is complete when the grid is installed, all tiles are in place, and service penetrations are sealed.
-
Systematic capture. Photographs are taken at regular intervals (daily or weekly) following a defined route through the building. Consistency matters — the same locations, same angles, similar times of day for consistent lighting.
-
AI analysis. The photographs are processed by AI, which compares the current state against:
- The previous capture (what has changed since last week?)
- The completion criteria (does the current state match "complete"?)
- The design intent (if a BIM model or design drawings are available, does the built work match the design?)
-
Progress classification. Each activity is classified as:
- Not started — no evidence of the activity in the photographs
- In progress — partial evidence, some elements visible but not all
- Complete — visual evidence matches the completion criteria
- Defective — work appears complete but with visible defects that require rectification
-
Reporting. AI generates a progress report mapping classification results against the programme, showing which activities are ahead, on track, behind, or blocked.
The limitation: photograph-based tracking can only assess work that is visible at the time of capture. Concealed work (services in ceiling voids, below-ground drainage, cavity insulation) must be photographed before it is covered up, or it cannot be assessed later.
360-degree capture and point cloud comparison
Standard photographs have a limitation: they capture one view at a time, and coverage depends on the photographer's choices. Two emerging capture methods provide more comprehensive site data.
360-degree cameras. Cameras like the Ricoh Theta and Insta360 capture a complete spherical image from a single position. Mounted on a tripod or hardhat, they document everything visible from that location.
The structured capture workflow:
- Define capture positions on the floor plan (typically at 3-5m intervals along corridors and in the centre of rooms)
- Capture at each position at every progress interval
- The result is a navigable virtual tour of the site at each point in time
- AI processes the captures to assess completion status for each visible element
This approach is used by Disperse, whose platform processes 360-degree site captures to map work completion against programme activities. The systematic capture ensures no area is missed — unlike ad hoc photography where the photographer might skip an area.
LiDAR and point cloud scanning. Laser scanning produces a 3D point cloud — millions of measured points representing the as-built geometry of the site. This data is precise (typically +/- 2-5mm) and comprehensive.
Point cloud to design comparison. The most powerful application is overlaying the as-built point cloud against the design model (BIM or 2D drawings converted to 3D). The comparison reveals:
- Work that is built and matches the design (complete)
- Work that is built but deviates from the design (potentially defective)
- Work that is designed but not yet built (not complete)
- Work that is built but not in the design (potentially unauthorised or a design change)
This is the approach used by Buildots, which combines indoor positioning with image capture to automate this comparison. Their system mounts cameras on hardhat clips and uses the building's ceiling grid as a positioning reference.
Your project uses 360-degree captures at 4m intervals through corridors and room centres. The AI reports the suspended ceiling in a meeting room as 'complete.' However, the captures were taken from the corridor, and the meeting room has opaque full-height partitions with a closed door. Can you trust this assessment?
The programme link — from physical progress to CPM schedule
Progress monitoring is only useful if it connects to the construction programme. Knowing that "the blockwork looks about 75% done" is less valuable than knowing that "Activity 3.2.1 — First Floor Internal Blockwork (planned duration 12 days, Day 8 of 12) is assessed at 65% physical completion, indicating it is approximately 1 day behind programme."
Linking AI observations to programme activities:
-
Activity mapping. Each programme activity is linked to physical locations and elements on the drawings. Activity "First Floor Internal Blockwork" maps to specific walls shown on the first-floor plan.
-
Physical completion assessment. AI processes site captures and assesses the completion percentage of the mapped elements. This might be based on:
- Area of completed blockwork versus total area on the plan
- Number of rooms with complete blockwork versus total rooms
- Linear metres of wall built versus total wall length
-
Programme status calculation. Physical completion is compared against planned completion at this point in the programme. If the activity is on Day 8 of 12, planned completion is approximately 67%. Assessed completion of 65% suggests the activity is marginally behind programme.
-
Earned value integration. Physical completion translates to earned value: the planned budget for the activity multiplied by the assessed completion percentage gives the earned value. Comparing this to actual cost gives cost performance index (CPI), and comparing to planned value gives schedule performance index (SPI).
Practical limitations:
- Not all activities are equally amenable to photographic assessment. Mechanical rough-ins, electrical containment in ceiling voids, and below-ground works are difficult to assess after they are concealed.
- Percentage completion is not always linear with physical completion. The last 10% of an activity (snagging, commissioning, cleaning) often takes disproportionately long.
- AI assessment of percentage completion is an estimate, not a measurement. It is more objective than a human estimate, but it is still an approximation.
Automating progress reports
One of the most immediate practical benefits of AI progress monitoring is automating the weekly or monthly progress report — the document that consumes hours of project management time and is often outdated by the time it is issued.
AI-generated progress reports can include:
- Floor-by-floor heat maps showing completion status (colour-coded by percentage)
- Activity-by-activity status summary linked to the programme
- Comparison photographs: last period versus this period for each major area
- Flagged concerns: activities that are behind programme, areas with visible defects, locations where work appears to have stopped
- Automated metrics: overall completion percentage, earned value calculations, predicted completion date based on current progress rate
The human review step:
AI generates the draft report. The project manager reviews it, adds context that photographs cannot capture (design changes in progress, known subcontractor mobilisation delays, weather impacts), and adjusts any assessments that the AI has got wrong. The project manager's value is in the interpretation and the narrative — explaining why progress is behind and what the recovery plan is — not in compiling photographs and counting completed rooms.
What stakeholders actually want:
Client-side project managers and construction directors want concise, evidence-based progress reports that answer three questions: Are we on programme? Are we on budget? What are the risks? AI-generated reports, properly reviewed and contextualised by the project manager, answer these questions with photographic evidence rather than subjective assessment.
Your AI progress system generates a report showing overall project completion at 62%. The programme says you should be at 65% at this point. The project director asks: 'Are we in trouble?' How do you use the AI data to answer?
Key takeaways
- Subjective progress assessment is the core problem — different people assess different percentages for the same work, and subcontractors have a financial incentive to overstate completion.
- AI-based progress tracking uses photographs (standard or 360-degree) compared against completion criteria to provide objective, repeatable assessments.
- 360-degree capture ensures comprehensive coverage but still requires line-of-sight to every space — walls and doors are opaque to all cameras.
- Progress must link to the programme — overall completion percentages are headline numbers. Activity-level analysis against the critical path is what project directors need.
- Automated reporting saves project management time and provides evidence-based stakeholder communication.
Next up: AI for Safety and Compliance.
Module 8 — Final Assessment
Why is subjective progress assessment problematic for interim valuations under a construction contract?
What is the key advantage of 360-degree site capture over standard photography for progress monitoring?
A project programme shows a critical path activity 'Roof Steelwork Erection' planned at 50% completion. AI assesses 45% physical completion. Non-critical activity 'Ground Floor Decoration' is planned at 30% but assessed at 50%. What is the project status?
What information can AI NOT reliably assess from post-completion photographs of an area where services are now concealed above a suspended ceiling?
