What problem does it solve?
A general contractor knows what should be built from the schedule and the BIM model, but not reliably what is actually built. The standard method is manual: a superintendent or project engineer walks the site, estimates the percentage complete of each trade by eye, and writes it up for the weekly meeting. That estimate is subjective, it takes site staff away from managing the work itself, and it usually surfaces a schedule slip or a quality issue only after it has already cost time to fix.
The effect compounds on a multi trade, multi floor project. Different superintendents judge "80 percent complete" differently, subcontractors have an incentive to round their own progress up, and nobody has a single, dated record of what a wall or a duct run looked like on a given day if a dispute over payment or defects comes up later.
How does it work?
- Capture the site. Workers walk their normal routes wearing a hardhat mounted camera, so the site is photographed at a set cadence without anyone doing an extra task.
- Build a digital twin. The images are stitched and located in the building, then matched automatically to the BIM model, element by element: which duct, partition, or fixture is visible, and whether it is there yet.
- Compare to the plan. The detected state of each element is compared with the schedule and the design, producing an objective percent complete per trade, per area and per task, plus a list of items that were meant to be installed and are not, or that do not match the design.
- Surface deviations. The system flags trades falling behind pace, sequencing conflicts (for example ceiling work starting before the ducts behind it are in), and installed work that differs from the BIM model.
- Act on it. Superintendents and project managers use the dashboard in the weekly trade meeting to compare each trade's measured pace of installed work against the schedule, and owners use it to verify progress claims behind payment applications instead of relying on self reported percentages. Keep the measurement at the level of the trade's output: using the same pace data to rank or discipline a named individual worker moves the system toward the Annex III employment category described under risk below.
- Audience
- Employee facing
- Autonomy
- Assist
- Adoption
- Early adopters
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Productivity gain | Too few to pool | 70% to 95% | 2 | 2 vendor |
| Error reduction | Too few to pool | 10% | 1 | 1 vendor |
Value drivers: Employee productivity, Speed and cycle time, Risk and loss reduction.
Indicative value
A general contractor running 10 active projects averaging 40,000 square feet each
USD 103,500 to USD 1.6 million
Superintendent and engineer time cost avoided on manual progress tracking per year
How this is calculated
Formula: projects * hoursPerProjectPerWeek * reductionShare * costPerHour * weeksPerYear. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Active projects tracked projects, projects | 5 | 15 | Editorial assumption for a mid size general contractor, replace with your own project count. |
| Superintendent and engineer hours spent on manual progress tracking, per project per week hoursPerProjectPerWeek, hours per project per week | 15 | 30 | Doxel reports six superintendents and project engineers on one 82,000 square foot healthcare project spent a combined 60 hours a week on manual progress tracking before automation; scaled down to this reference project's average 40,000 square feet (60 times 40,000 divided by 82,000, about 29 hours, rounded to 30). Source |
| Share of manual tracking time removed reductionShare, fraction of tracking hours | 0.5 | 0.7 | Capped at NCC's reported 70% reduction in manual reporting time, the lower of the two reductions the evidence on this page reports; Doxel's higher reported 95% reduction at Layton Construction is not used, to keep the range conservative. Source |
| Fully loaded cost of a superintendent or project engineer hour costPerHour, USD per hour | 60 | 100 | Editorial assumption, replace with your own fully loaded labor cost. |
| Weeks worked per year weeksPerYear, weeks | 46 | 50 | Editorial assumption for an active construction calendar. |
What it leaves out: Gross tracking labor cost avoided only. It leaves out the cost of the platform, the camera hardware, the one time BIM and schedule integration work, and any separate gains from fewer overbilling disputes or less rework, which the evidence on this page reports separately.
Market estimates (analyst estimates, not deployments)
- Doxel, citing McKinsey, states that using digital tools could improve overall construction productivity by 14 to 15 percent and reduce project costs by 4 to 6 percent. How Owners Can Drive Construction Forward (2024)
Who already uses it?
2 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Layton Construction
United States · Real estate · 2024
Layton Construction, a US general contractor, used Doxel's AI progress tracking platform on an 82,000 square foot healthcare facility project. Before Doxel, six superintendents and project engineers spent a combined 60 hours a week manually tracking progress; Doxel cut that to 3 hours and reduced overbilling by 10 percent through precise progress tracking that eliminated disputes over the percent complete.
- Productivity gain: 95%, On an 82,000 square foot healthcare facility project
"On a recent 82,000 SqFt healthcare facility project, six superintendents and project engineers were initially spending a combined 60 hours per week manually tracking progress. With Doxel's technology, this time was slashed by 95%, reducing the task to just 3 hours total."
Claimed by: vendor - Error reduction: 10%
"Doxel enabled a 10% reduction in overbilling by providing precise progress tracking, simplifying billing processes, and eliminating disputes over the percent complete."
Claimed by: vendor
NCC
Finland · Real estate · 2022
NCC, one of the largest construction companies in the Nordic region, piloted Buildots' hardhat mounted camera and AI platform on Helsinki building projects, in a pilot announced in February 2022, to replace manual, subjective percent complete tracking with a measured comparison against its BIM model. The partnership expanded to four projects totalling 68,500 square metres, and Buildots reports large cuts in NCC's manual reporting time and more tasks completed on plan.
- Productivity gain: 70%
"70% reduction in manual reporting."
Claimed by: vendor
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- A BIM model broken into elements the AI can track against, kept reasonably current
- A construction schedule with tasks the detected progress can be matched to
- A fixed cadence of site photo capture, by staff wearing a hardhat mounted camera
Systems to integrate
- BIM authoring or viewing platform (for example Autodesk Revit or Navisworks)
- Scheduling tool (for example Primavera P6, Microsoft Project or the contractor's own tool)
- Site capture hardware and its upload pipeline (hardhat mounted camera)
- Document control or project management platform for routing flagged issues and RFIs
Complexity: Medium
Capturing images is simple once site staff have the hardhat mounted camera. The work is in getting a BIM model that is actually kept current, mapping it cleanly to the schedule's tasks, and agreeing with subcontractors that the automated percent complete is the number the weekly meeting and payment applications will use.
- 1
Start with one project and a clean BIM model
Pick a project where the BIM model is genuinely kept up to date and the trades are willing to be measured, and agree up front which trades and floors are in scope for the pilot.
- 2
Map the schedule to trackable elements
Break the schedule's tasks down to the same level as the BIM elements the vision system detects (a duct run, a partition, a fixture), so a detected element maps to one task.
- 3
Set the capture cadence and route
Fix how often and by what route the site is captured, so every area gets covered on a predictable schedule and gaps in coverage do not look like missing progress.
- 4
Put the dashboard in the weekly trade meeting
Replace the subjective percent complete slide with the measured one, and agree with subcontractors in advance how disagreements between measured and self reported progress get resolved.
- 5
Route deviations to an owner
Send sequencing conflicts and design mismatches to the trade or engineer who owns the area, with the photo evidence attached, rather than only surfacing them on a dashboard nobody acts on.
- 6
Extend to more projects once the workflow sticks
Add projects once one team has changed how it runs its trade meetings around the data, not before, so the second project benefits from a working playbook rather than a repeated pilot.
Guardrails
- Treat the automated percent complete as the primary number only after a defined period of running it alongside the manual estimate and reconciling differences
- Keep a photo dated record of the evidence behind every reported percent complete, for payment application and dispute support
- Mask or restrict access to imagery that incidentally captures workers, beyond what is needed to verify installed work
- Require a human review before a flagged design mismatch is escalated to a stop work or rework instruction
KPIs to instrument
- Time site staff spend on manual progress tracking, before and after
- Share of tasks completed on the plan (Percent Plan Complete or equivalent), before and after
- Overbilling or billing disputes tied to disagreements over percent complete
- Coverage, meaning the share of the site captured on the agreed cadence, so gaps are visible
Human in the loop
Superintendents and project engineers decide what to do with a flagged deviation: the system measures and flags it, it does not instruct a trade to redo work or change the schedule. Owners and their quantity surveyors use the record as evidence, not as an automatic approval or rejection of a payment application.
Common failure modes
- Stale BIM model
- When the model is not kept current, the AI reports a mismatch that is really a documentation gap, not a site problem. Require model updates as part of the change process, not as an afterthought.
- Capture gaps read as missing work
- A missed walk or a blocked camera route can look like work has not started. Track and display capture coverage alongside progress, not just the progress number.
- Dashboard nobody acts on
- Data with no owner and no place in the weekly meeting gets ignored. Assign an owner for every flagged deviation and review coverage and open flags every week.
- Disputed percent complete undermines adoption
- If subcontractors are not told in advance how the measured number will be used, a lower automated figure than their own estimate becomes a fight instead of a fix. Agree the reconciliation process before go live.
What are the risks and rules?
EU AI Act
Depends on design
Tracking built quantities against a BIM model and schedule is not itself an Annex III use. Annex III point 4 covers employment, the management of workers and access to work as a self employed person, and brings a system into that category under point 4(b): AI used to make decisions affecting work related relationships, or, quoting the regulation, "to monitor and evaluate the performance and behaviour of persons in such relationships", meaning employees, subcontracted individuals and others in a work related relationship with the organization. Measuring the quantities and pace of the work itself, without naming or ranking a crew or an individual, stays out of that category. This page's own weekly meeting step (howItWorks, step 5) keeps the measurement at the level of the trade's output for exactly this reason: using the same per trade pace data to evaluate or discipline a named crew member or subcontracted individual would stretch the system into that high risk category. Treat any such individual level use as high risk and apply the controls below to it, rather than relying on element level scoping alone.
Controls to put in place
- Written scope statement that the system measures installed work, not individual worker performance
- Data retention and access limits on site imagery, since workers appear in it incidentally
- Change control on the BIM to schedule mapping, since it decides what "on plan" means for a trade
Frequently asked questions
- Does this replace the superintendent's judgment?
- No. It replaces the manual estimate of percent complete with a measured one, and flags deviations for a human to act on. NCC's own account is that it freed time from arguing with subcontractors so staff could spend it actually fixing problems.
- What has to be true before this works?
- A BIM model that is genuinely kept current, and a schedule broken down to the same level of detail as the elements the system tracks. Without both, the flagged mismatches mostly reflect a stale model, not a real site problem.
- Is this only for the general contractor, or does the owner get value too?
- The evidence on this page comes from contractor deployments: Doxel's Layton Construction case study reports a 10 percent reduction in overbilling, credited to precise progress tracking that eliminated disputes over the percent complete, and Buildots' NCC case study reports fewer conflicts that could have caused an overspend of tens of thousands of euros. Neither source says which party was overbilling or on which side an averted overspend would have landed. An owner still gets a related benefit from the same measured percent complete: this page's howItWorks step 5 describes owners using the dashboard to verify progress claims behind payment applications instead of relying on self reported percentages, though no evidence record on this page quantifies that owner side use.
- Does it also check safety and quality, or only schedule progress?
- Some platforms extend the same site imagery to installation quality checks against the BIM model as a byproduct, as NCC Finland reported. Treat that as a bonus of the same capture, not a substitute for a dedicated safety monitoring program.
How to cite this page
Blits.ai AI Use Case Library, "AI for construction schedule and progress tracking", last verified 29 September 2026, https://www.blits.ai/ai-use-cases/construction-schedule-and-progress-tracking. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 30 September 2026: First published