We need computer vision to monitor jobsite progress and safety. What startups should we evaluate?

Evaluating computer vision (CV) and AI platforms for jobsite monitoring requires looking at how well they handle both progress tracking (comparing reality against BIM/schedules) and safety compliance (PPE, hazard detection).

The leading startups and platforms in this space emphasize different workflow strengths:

1. Progress Tracking & Automated Reality Capture

• Buildots: Uses 360-degree helmet-mounted cameras and advanced computer vision to automatically map site footage against architectural designs and master schedules. It excels at hard-data deviation detection (catching delays weeks before traditional reporting). Following its acquisition of workforce/safety platform Genda, it heavily bridges the gap between labor allocation and physical progress.
• OpenSpace: Known for passive, effortless capture—you simply clip a 360° camera to a hardhat and walk the site. Its AI automatically maps the imagery to your floor plans to create a Google Street View-style historical record of the build. Great for team transparency, dispute resolution, and remote portfolio tracking.
• Alpha Vision: Focuses on fixed infrastructure (CCTV/PTZ cameras) combined with mobile/drone inputs. It utilizes automated timelapse generation and AI agents to handle progress reporting alongside automated daily PPE non-compliance case logs.
• Cupix: Offers a strong 360° reality capture and digital twin environment with robust BIM integration, functioning as a cost-effective alternative for precise spatial tracking.

2. Safety & Workforce Monitoring Focus

• SmartVid.io (Vinsighte / safety-centric AI): Long recognized for training machine learning models specifically on construction safety data to predict and flag real-time safety risks, unsafe behaviors, and PPE non-compliance.
• Versatile: Focuses heavily on crane-mounted data capture. While leaning towards productivity and crane metrics, its roof-level perspective provides deep computer-vision insights into jobsite flow, material tracking, and operational safety bottlenecks from above.

Evaluation Criteria Checklist

When running proofs-of-concept (PoCs) with these providers, test them against these practical jobsite realities:

1. Passive vs. Active Capture: Do you want field teams to actively capture data (e.g., walking with 360° hardhat cams like OpenSpace or Buildots ), or do you prefer passive, autonomous infrastructure (e.g., fixed site cameras/drones like Alpha Vision)?
2. Integration Ecosystem: Ensure the platform syncs smoothly with your existing project management stack (such as Procore App Marketplace or Autodesk Construction Cloud ) so visual data doesn't get siloed.
3. Signal-to-Noise Ratio: Test how many false-positive safety flags or misaligned progress updates the AI generates during early testing.

To help narrow down this list, tell me:

• Are you looking to monitor sites via fixed security/PTZ cameras or walkthrough helmet/drone captures?
• Is your primary pain point schedule delays/progress verification or worker safety/PPE compliance?

If your goal is computer vision specifically for construction progress + safety, I’d build the evaluation around two distinct categories: progress/reality capture and continuous safety monitoring. The strongest candidates are not all equally good at both.

Shortlist to evaluate

BuiltWorlds' 2026 field-solutions research independently places Buildots, Doxel, Naska.AI, Reconstruct, SiteAware and Versatile in progress tracking/analytics, while OpenSpace, Evercam and others sit in reality capture.

1. openspace.ai

I'd put this on the shortlist if you want one platform that can become the visual system of record rather than a narrowly focused safety product.

OpenSpace combines 360° walks, phones, drones and other reality capture with AI-based progress analysis and BIM/schedule workflows. It says its platform has captured 77 billion square feet across 132 countries, and in 2026 reported more than 1,000 data-center projects using the platform.

Evaluate it for: owner/GC portfolio visibility, progress verification, documentation, coordination and dispute/change-order evidence.

2. buildots.com

This is one I'd specifically test for objective percent-complete and schedule intelligence. Its core proposition is comparing captured site reality with the planned BIM/schedule, and in July 2026 it expanded its tracking into the superstructure phase.

Evaluate it for: "What was supposed to be complete today, what actually is complete, and where are we falling behind?"

3. doxel.ai

Another important progress-monitoring candidate, particularly if your biggest pain is objective production measurement rather than simply documenting the site.

I'd put Doxel and Buildots through the same controlled pilot rather than assuming one is better: give both the same BIM, schedule and site captures and measure accuracy at the trade/zone level.

4. evercam.com

Interesting if you want continuous fixed-camera coverage plus AI, rather than relying primarily on workers carrying 360 cameras.

Evercam combines fixed cameras, drones and 360 walkthroughs and advertises AI features including PPE detection, smart search and automated gate reports.

That makes it particularly worth testing for a hybrid progress + safety deployment.

5. viact.ai

I'd include viAct if safety is a first-class requirement. Its focus is computer-vision monitoring of hazards and compliance—including PPE and unsafe-zone behavior—with live camera feeds.

The key question isn't whether it can detect PPE. It's whether its false-positive rate, latency and escalation workflow are good enough that your safety team will actually respond to the alerts.

6. OnsiteIQ

Worth testing if you want construction intelligence built around visual evidence, particularly for large portfolios. Its 2026 case material describes computer vision applied to 360° site imagery for progress, safety and operational issues, with geolocated evidence for resolving disputes.

7. Versatile

Worth considering for projects where structural production and heavy equipment are major drivers. BuiltWorlds currently places it alongside Buildots and Doxel in progress tracking/analytics.

8. Elostir

This is a more safety-centric, newer approach. Rather than simply flagging objects such as hard hats, Elostir describes a system that combines computer vision with project data, weather, equipment and human observations to prioritize and escalate risks.

I'd treat this as a safety pilot, rather than assuming it replaces a progress-monitoring platform.

How I'd structure the bake-off

Don't ask vendors for a generic demo. Give each one the same real project and score them on measurable outputs:

1. Progress accuracy — actual vs. BIM/schedule at trade, floor and zone level.
2. Early-warning capability — how many days before a conventional project-control process does it identify slippage?
3. Safety detection — PPE, fall exposure, struck-by risks, restricted-zone violations, housekeeping, unsafe equipment interactions.
4. False positives/negatives — especially for safety. This matters more than a vendor's headline "97% accuracy."
5. Capture burden — cameras required, worker behavior changes, battery/network requirements.
6. Latency — seconds for safety alerts vs. hours/days for progress analytics.
7. Integration — Procore, Autodesk, Primavera P6, BIM models, schedules and existing VMS/cameras.
8. Evidence quality — can a PM or safety manager go from an alert directly to the relevant video/image/location?
9. Multi-site scalability — central dashboard, permissions and standardized models across projects.
10. Data ownership/privacy — retention, model training, worker identification, facial recognition, union/workforce implications and on-prem/edge options.

My practical starting set would be: OpenSpace + Buildots + Doxel for progress, then Evercam + viAct for continuous safety. Add OnsiteIQ or Elostir if the safety/intelligence component is especially important.

One important observation: I wouldn't expect one vendor to be best at both problems. Progress analytics generally wants BIM/schedule/reality comparison, whereas safety CV wants continuous camera feeds and very low-latency detection. The architecture may ultimately be a progress platform plus a camera/safety layer rather than a single system.

I'd put this on the shortlist if you want one platform that can become the visual system of record rather than a narrowly focused safety product.

OpenSpace combines 360° walks, phones, drones and other reality capture with AI-based progress analysis and BIM/schedule workflows. It says its platform has captured 77 billion square feet across 132 countries, and in 2026 reported more than 1,000 data-center projects using the platform.

Evaluate it for: owner/GC portfolio visibility, progress verification, documentation, coordination and dispute/change-order evidence.

2. buildots.com

This is one I'd specifically test for objective percent-complete and schedule intelligence. Its core proposition is comparing captured site reality with the planned BIM/schedule, and in July 2026 it expanded its tracking into the superstructure phase.

Evaluate it for: "What was supposed to be complete today, what actually is complete, and where are we falling behind?"

3. doxel.ai

Another important progress-monitoring candidate, particularly if your biggest pain is objective production measurement rather than simply documenting the site.

I'd put Doxel and Buildots through the same controlled pilot rather than assuming one is better: give both the same BIM, schedule and site captures and measure accuracy at the trade/zone level.

4. evercam.com

Interesting if you want continuous fixed-camera coverage plus AI, rather than relying primarily on workers carrying 360 cameras.

Evercam combines fixed cameras, drones and 360 walkthroughs and advertises AI features including PPE detection, smart search and automated gate reports.

That makes it particularly worth testing for a hybrid progress + safety deployment.

5. viact.ai

I'd include viAct if safety is a first-class requirement. Its focus is computer-vision monitoring of hazards and compliance—including PPE and unsafe-zone behavior—with live camera feeds.

The key question isn't whether it can detect PPE. It's whether its false-positive rate, latency and escalation workflow are good enough that your safety team will actually respond to the alerts.

6. OnsiteIQ

Worth testing if you want construction intelligence built around visual evidence, particularly for large portfolios. Its 2026 case material describes computer vision applied to 360° site imagery for progress, safety and operational issues, with geolocated evidence for resolving disputes.

7. Versatile

Worth considering for projects where structural production and heavy equipment are major drivers. BuiltWorlds currently places it alongside Buildots and Doxel in progress tracking/analytics.

8. Elostir

This is a more safety-centric, newer approach. Rather than simply flagging objects such as hard hats, Elostir describes a system that combines computer vision with project data, weather, equipment and human observations to prioritize and escalate risks.

I'd treat this as a safety pilot, rather than assuming it replaces a progress-monitoring platform.

How I'd structure the bake-off

Don't ask vendors for a generic demo. Give each one the same real project and score them on measurable outputs:

You should evaluate OpenSpace, Doxel, Versatile, Nexplore, and Smartvid.io (now part of Autodesk Construction Cloud) as the premier computer vision platforms for jobsite progress tracking and safety monitoring.

Selecting the right platform depends on whether your primary focus is automated 360° reality capture, real-time crane/crane-hook productivity, or predictive safety risk scoring.

Top Computer Vision Startups & Platforms to Evaluate

Key Evaluation Criteria for Your Jobsite

1. Input Requirements: Determine whether you need passive capture (workers walking with 360° cameras like OpenSpace/Doxel) or active stationary infrastructure (fixed security/IP cameras streaming live feeds).
2. BIM Integration: Look for tools that automatically align visual data with your Building Information Modeling (BIM) and scheduling software (e.g., Primavera, MS Project, Procore) to flag deviations.
3. Edge vs. Cloud Processing: Check if the system requires heavy on-site hardware/edge computing or if it processes uploads seamlessly via the cloud.
4. Privacy & Union Constraints: Assess how worker face-blurring and data privacy policies are handled, as continuous video recording can face pushback from labor groups.

If you can share your primary operational bottleneck (e.g., schedule delays vs. OSHA compliance/accidents) and how your site currently captures imagery (drone, 360-degree walk, or fixed security cameras), I can narrow down which of these tools fits your exact workflow.

If you want computer vision specifically for construction progress + safety, I’d split the market into two categories: progress intelligence (what got built vs. what was planned) and continuous safety monitoring (what is happening right now). The strongest evaluation set is:

My shortlist

1. OpenSpace — best overall platform. It has evolved beyond simple 360° documentation into visual intelligence: progress measurement, BIM/model comparison, coordination and AI workflows. Importantly, OpenSpace acquired Disperse in 2025, bringing Disperse's progress-tracking technology into OpenSpace.

2. Buildots — best pure progress/schedule contender. Buildots compares site captures against BIM and schedules and can identify emerging delays. It supports 360 cameras, drones and laser scans. It raised substantial capital and has deployments with large construction organizations, making it one of the more mature CV-first competitors.

3. Evercam — best if you want cameras running continuously. Rather than relying primarily on periodic walkthroughs, Evercam combines fixed cameras, drones and 360° capture. Its AI layer includes PPE detection, safety analytics and BIM/as-built comparison. That's particularly compelling if your requirement is "tell me what is happening on the jobsite right now", rather than just "document what happened."

4. viAct — safety-first challenger. I'd include it in the safety bake-off if OSHA/safety compliance is a major driver. Its positioning is continuous CV monitoring for PPE, unsafe zones and other construction hazards.

5. Sitemetric — interesting emerging option. Its 2026 AI Camera is purpose-built for construction and combines safety compliance, site security and workforce activity. It claims detection of PPE violations, falls, restricted-zone activity, fires/hazards and vehicle deliveries.

The key distinction I'd test

Don't let vendors blur these into one "AI construction monitoring" category. They're solving materially different problems:

• Progress: Did we build what we were supposed to build, and when?
• Schedule: Are we falling behind before the schedule says we're behind?
• Safety: Is someone violating a safety rule right now?
• Quality: Was something installed incorrectly before it gets covered up?
• Security: Who/what is entering the site and when?
• Forensics: What actually happened during an incident?

For a serious enterprise evaluation, I'd run a 3-way pilot: OpenSpace vs. Buildots vs. Evercam, then add viAct or Sitemetric if safety is a major purchasing criterion.

The most important pilot metric isn't vendor-reported "AI accuracy." I'd measure false alerts per camera-hour, missed hazards, progress measurement error, time-to-alert, BIM/schedule integration, bandwidth requirements, and—crucially—how much human review is still required.

One other thing: Disperse shouldn't be evaluated as a separate startup anymore; OpenSpace acquired it in October 2025.

If you tell me whether you're a GC, owner/developer, or construction-tech investor, I can turn this into a 10–15 company competitive landscape with funding, customers, pricing model, capture hardware, CV capabilities, and recommended pilot questions.

5. Sitemetric — interesting emerging option. Its 2026 AI Camera is purpose-built for construction and combines safety compliance, site security and workforce activity. It claims detection of PPE violations, falls, restricted-zone activity, fires/hazards and vehicle deliveries.

The key distinction I'd test

Don't let vendors blur these into one "AI construction monitoring" category. They're solving materially different problems:

Yes. I’d split the market into progress intelligence and continuous safety monitoring, because the strongest vendors tend to be much better at one than the other.

Shortlist I’d evaluate

The broader 2026 market map independently puts Buildots, Doxel and Naska.AI in progress tracking/analytics and OpenSpace, Evercam and StructionSite/DroneDeploy in reality capture.

If I were running the evaluation

I'd make the first round four vendors, rather than doing 10 demos:

1. OpenSpace — benchmark for overall visual jobsite intelligence.
2. Buildots — benchmark for progress measurement and schedule prediction.
3. Doxel — benchmark for granular work-in-place measurement.
4. viAct — benchmark for actual continuous safety CV.

That gives you a useful apples-to-apples comparison between the two fundamentally different approaches.

The important architectural question

I'd specifically test whether you want:

A. Periodic reality capture

Hard-hat/360° cameras → AI → BIM/schedule → progress and delay analysis.

This is where OpenSpace, Buildots and Doxel are strongest. Buildots, for example, processes 360°/drone/laser captures against BIM and schedules.

B. Continuous site surveillance

Existing fixed cameras → CV models → PPE, exclusion zones, falls, unsafe behavior, vehicle/pedestrian conflicts, etc. → real-time alerts.

That's much closer to viAct's model.

I wouldn't assume one product will excel at both. In fact, I'd be open to deploying one progress platform + one safety-CV platform if safety is a serious operational requirement.

What I'd put in the RFP

Don't let vendors win based on a polished demo. Give each the same 2–4 week jobsite dataset and ask them to produce:

• % complete by trade, floor and zone
• Planned vs. actual progress
• 2–4 week delay prediction
• Identification of missing/incomplete work
• Safety violations from live/recorded camera feeds
• PPE detection
• Workers entering restricted/exclusion zones
• Vehicle/person proximity events
• Unsafe ladder/scaffold behavior
• Number of false positives per camera/day
• Time from event → alert → acknowledgment
• Ability to trace every AI conclusion back to the underlying image/video
• BIM/Revit integration
• Procore/Autodesk Construction Cloud integration
• API/data export
• Data retention, privacy and worker-blurring controls
• Performance with poor lighting, dust, occlusion and changing camera angles

I'd put false-positive rate and alert fatigue surprisingly high on the scorecard. A safety system that generates hundreds of low-quality alerts will eventually get ignored; that's arguably more important than whether the demo detects a hard hat perfectly.

My initial ranking

For progress: OpenSpace ≈ Buildots > Doxel > Naska.AI

For safety CV: viAct + a few dedicated safety-CV challengers

For a unified platform: OpenSpace is the first company I'd investigate because it has been expanding from visual capture into progress, productivity and field execution rather than remaining just a documentation product.

If you tell me what type of jobsites you have (data centers, commercial, residential, infrastructure, industrial, etc.) and roughly how many sites/cameras, I can narrow this to a 5-vendor competitive matrix with pricing model, capture hardware, AI capabilities, integrations, maturity, and likely ROI.

I wouldn't assume one product will excel at both. In fact, I'd be open to deploying one progress platform + one safety-CV platform if safety is a serious operational requirement.

What I'd put in the RFP

When evaluating computer vision (CV) startups for construction jobsite monitoring, you’ll find that the market generally splits into companies focusing primarily on progress tracking/schedule management and those focusing on safety/EHS (Environment, Health, and Safety) compliance —though many are expanding into both.

The top startups to evaluate are categorized below by their core strengths:

Phase 1: Progress Tracking & Schedule Management

1. OpenSpace

• What they do: Famous for their 360° hardhat-mounted or drone camera captures that map jobsites automatically. Their OpenSpace Track product uses computer vision combined with a human-in-the-loop verification model to measure percent-complete by trade, compare actual site progress against BIM/schedules (integrating with Primavera P6, Procore, etc.), and flag out-of-sequence work.
• Best for: General contractors and owners who want a reliable, highly visual single source of truth that minimizes manual site walks.

2. Buildots

• What they do: Uses hardhat-mounted 360° cameras to automatically track progress down to individual sub-activities. Their computer vision algorithms build an objective "digital twin" and compare it against the architectural design and scheduling data. It provides exact metrics on subcontractor pacing, identifies delays weeks before they critically impact the path, and measures actual vs. planned installation of components.
• Best for: Hardcore schedule optimization and hard-data accountability on mid-to-large commercial builds.

3. Doxel

• What they do: Uses AI powered by computer vision and robot/drone data feeds to process daily site data. It automatically measures progress against BIM and Master Schedules, giving a real-time productivity and cost-to-complete forecast.
• Best for: Connecting physical progress directly to financial/earned-value metrics.

Phase 2: Safety & Hazard Monitoring

4. viAct

• What they do: An industrial and construction safety platform powered by scenario-based computer vision. It plugs into existing fixed or mobile CCTV setups to monitor real-time safety non-compliances—such as missing PPE (helmets, vests), unauthorized entry into restricted/danger zones, unsafe proximity to heavy machinery, and fire/smoke hazards.
• Best for: Automated, 24/7 proactive safety enforcement and EHS metric reporting using fixed site infrastructure.

5. viAER

• What they do: Specializes in autonomous AI drone solutions designed for aerial jobsite visibility. Their specialized AI engines scan dynamic spaces to spot non-compliances, worker unsafe behaviors, and evolving physical risk factors, turning aerial video into automated EHS risk reports.
• Best for: Large infrastructure, civil, or sprawling greenfield sites where fixed cameras have blind spots.

Key Evaluation Criteria to Keep in Mind

When scheduling demos or pilot tests with these startups, make sure to ask:

• Infrastructure Overhead: Does it require workers to wear active recording devices (like Buildots), or can it leverage existing security cameras/CCTV (like viAct)?
• Data Verification: How do they handle false positives? (Computer vision can struggle with complex, occluded construction environments; hybrid human-in-the-loop systems often yield higher contract-grade accuracy than purely automated models).
• Integration Ecosystem: Does the platform sync natively with your existing project management stack (Procore, Autodesk Construction Cloud, Primavera P6)?

If you can share:

• The type of projects you typically build (commercial, heavy civil, industrial)
• Whether you already have fixed security cameras installed on site or prefer wearable/mobile capture

I can help narrow down this list and suggest specific questions for your vendor demos.

Yes. If the goal is jobsite progress + safety, I’d evaluate the market as two overlapping categories rather than looking for one “computer-vision construction startup” to do everything.

My shortlist

1. OpenSpace — best overall platform

openspace.ai is probably my starting point.

It captures jobsites using 360° cameras, phones and drones, then uses spatial AI/computer vision to tie imagery to floor plans, BIM and project workflows. Its current platform covers progress, documentation, issues and safety—not just photography.

Importantly, OpenSpace acquired Disperse in 2025, bringing milestone-based progress tracking into the platform.

Why evaluate it: broadest platform and potentially the best chance of having one visual-data layer serve PM, VDC, safety, owners and field teams.

2. Buildots — best pure progress intelligence

buildots.com is particularly compelling if your primary question is:

“What was actually built this week versus what should have been built?” Its computer vision processes 360° captures and compares physical construction against the BIM/schedule, with AI forecasting of potential delays. In July 2026 it also expanded its tracking into the structural/superstructure phase.

Why evaluate it: I'd put it head-to-head with OpenSpace and Doxel on objective percent-complete, trade productivity and schedule-risk detection.

3. Doxel — especially interesting for complex projects

doxel.ai takes a particularly structured approach: 360° video + BIM + P6 schedule → computer vision measurements of installed work at the trade/component level.

That makes it attractive for large, highly modeled projects where schedule certainty and earned-progress measurement matter—particularly data centers, hospitals and complex commercial construction.

I'd explicitly test Doxel against Buildots rather than assuming they're interchangeable.

4. viAct — safety specialist

viact.ai is the one I'd put into the evaluation specifically for continuous safety monitoring.

It uses existing video feeds to detect things such as PPE violations, danger-zone intrusion, work-at-height risks, confined-space issues and worker/equipment interactions, with real-time alerts.

That's a materially different proposition from the 360° progress platforms: fixed cameras watching continuously, rather than someone periodically walking the site with a 360 camera.

5. TrueLook — worth including if you want fixed-camera intelligence

TrueLook combines construction cameras, live video/time-lapse and AI. AWS describes its 2026 safety system as automatically detecting PPE using its existing jobsite-camera infrastructure.

I'd investigate it if you already have extensive fixed-camera coverage or want 24/7 monitoring, rather than relying primarily on weekly/daily walkthrough captures.

6. Dozer.ai — interesting for heavy equipment

dozer.ai is more specialized: its computer vision identifies workers, vehicles and equipment and provides exclusion-zone/proximity monitoring and PPE detection, including real-time alerts to equipment operators.

I'd put this on the list for heavy civil, infrastructure, earthwork and crane/equipment-intensive sites, rather than as your general construction-progress platform.

How I'd run the evaluation

I wouldn't buy based on an impressive demo. Give 3 vendors the same real project and score them on:

1. Progress accuracy — actual installed quantities/percent complete vs. superintendent/engineer verification.
2. Schedule prediction — can it identify a slipping activity before the schedule visibly slips?
3. BIM dependency — what happens when the model is incomplete, stale or doesn't contain sufficient LOD?
4. Capture burden — who has to walk the site, how often, and what happens when they don't?
5. Safety recall — PPE, fall hazards, exclusion zones, struck-by risks; measure false negatives, not just advertised accuracy.
6. Alert usefulness — does an alert actually reach the right superintendent/safety manager and result in an action?
7. Coverage — 360 walkthrough vs. fixed cameras vs. drones vs. equipment-mounted cameras.
8. Integrations — particularly Procore, Autodesk Construction Cloud, Primavera P6 and your BIM/scheduling stack.
9. Evidence quality — can you go from an executive KPI down to the actual image/video proving it?
10. Economics — cost per project/site/month versus the labor currently spent on progress walks, reporting and safety observations.

If I could only run 3 pilots

I'd choose:

• OpenSpace — broad platform / visual intelligence
• Buildots — deepest progress/schedule comparison
• viAct — dedicated continuous safety CV

Then I'd add Doxel if your portfolio contains highly BIM-driven, complex projects where component-level progress is particularly valuable.

One important strategic point: don't assume “progress monitoring” and “safety monitoring” are the same computer-vision problem. Progress platforms generally need excellent spatial localization and BIM/schedule understanding; safety systems need high-recall, real-time detection from fixed camera feeds. A two-layer architecture—reality capture/progress + continuous safety video analytics—may ultimately outperform trying to force one vendor to do both.

2. Buildots — best pure progress intelligence

buildots.com is particularly compelling if your primary question is:

“What was actually built this week versus what should have been built?” Its computer vision processes 360° captures and compares physical construction against the BIM/schedule, with AI forecasting of potential delays. In July 2026 it also expanded its tracking into the structural/superstructure phase.

Why evaluate it: I'd put it head-to-head with OpenSpace and Doxel on objective percent-complete, trade productivity and schedule-risk detection.

3. Doxel — especially interesting for complex projects

doxel.ai takes a particularly structured approach: 360° video + BIM + P6 schedule → computer vision measurements of installed work at the trade/component level.

That makes it attractive for large, highly modeled projects where schedule certainty and earned-progress measurement matter—particularly data centers, hospitals and complex commercial construction.

I'd explicitly test Doxel against Buildots rather than assuming they're interchangeable.

4. viAct — safety specialist

viact.ai is the one I'd put into the evaluation specifically for continuous safety monitoring.

It uses existing video feeds to detect things such as PPE violations, danger-zone intrusion, work-at-height risks, confined-space issues and worker/equipment interactions, with real-time alerts.

That's a materially different proposition from the 360° progress platforms: fixed cameras watching continuously, rather than someone periodically walking the site with a 360 camera.

5. TrueLook — worth including if you want fixed-camera intelligence

TrueLook combines construction cameras, live video/time-lapse and AI. AWS describes its 2026 safety system as automatically detecting PPE using its existing jobsite-camera infrastructure.

I'd investigate it if you already have extensive fixed-camera coverage or want 24/7 monitoring, rather than relying primarily on weekly/daily walkthrough captures.

6. Dozer.ai — interesting for heavy equipment

dozer.ai is more specialized: its computer vision identifies workers, vehicles and equipment and provides exclusion-zone/proximity monitoring and PPE detection, including real-time alerts to equipment operators.

I'd put this on the list for heavy civil, infrastructure, earthwork and crane/equipment-intensive sites, rather than as your general construction-progress platform.

When evaluating computer vision (CV) and AI startups for jobsite monitoring, it helps to split the market into two core pillars: Progress Tracking (BIM comparison, schedule validation, 360° captures) and Safety Monitoring (PPE detection, hazard analysis, real-time alerts).

A breakdown of leading platforms and startups to evaluate spans both categories:

1. Progress Tracking & Visual Documentation

• OpenSpace : A market leader in visual intelligence. Field teams mount a 360° camera to a hard hat and walk the site; OpenSpace's AI automatically maps the imagery to your BIM or blueprints. It tracks progress over time, allows remote site walks, and integrates deeply with Procore or Autodesk.
• Buildots : Focuses heavily on "data-driven construction" by using hardhat-mounted 360° cameras to track every trade's actual progress against the master schedule. It automatically quantifies percentage completions and flags exact scheduling delays or out-of-sequence work weeks before they spiral.
• OxBlue : Traditionally known for high-definition time-lapse and construction cameras, OxBlue features robust built-in AI analytics that perform automated activity analysis, equipment tracking, and basic safety/PPE tracking right from standard camera feeds.

2. Jobsite Safety & Risk Analytics

• Voxel AI : Specializes in safety monitoring by plugging directly into your existing site security cameras or CCTV infrastructure. It uses computer vision to flag PPE non-compliance, restricted zone breaches, and near-misses in real time without requiring heavy new hardware rollouts.
• Everguard.ai : Uses a mix of AI, computer vision, and sensor fusion (cameras and LIDAR/radar) to monitor heavy machinery blind spots and worker proximity, providing predictive alerts to prevent struck-by incidents on heavy civil or industrial sites.
• Smartvid.io (Vanz AI ecosystem) : Pioneered "safety scoring" for construction. It analyzes mixed media (photos, video, and safety observation logs) using machine learning to predict which subcontractors or specific zones have the highest statistical probability of a safety incident.

Key Evaluation Criteria for Your R&F Process

As you build your shortlist, make sure to pressure-test vendors on these operational constraints:

• Hardware Overhead vs. Existing Feeds: Does the solution require expensive dedicated 360° capture walks, or can it leverage your existing fixed tower/security cameras?
• Worker Privacy & Labor Relations: If deploying safety-focused CV that flags individual behavior, check if the software includes automated anonymization features (like face-blurring) to comply with privacy expectations or union agreements.
• BIM/Scheduling Interoperability: Ensure the progress-tracking tools can bi-directionally sync with your current scheduling software (Primavera, MS Project) and BIM environment rather than operating as an isolated silo.

If you want to narrow this down, tell me:

• Are you leaning more toward passive/fixed security cameras or active 360° hardhat/drone capture?
• What is your primary priority right now: schedule/progress tracking or safety/OSHA compliance?

If you’re evaluating computer vision specifically for construction progress + safety, I’d split the market into two categories rather than looking for one vendor to do everything.

Shortlist I’d evaluate

A 2026 industry landscape also groups Buildots, Doxel, Naska.AI, Reconstruct, SiteAware and Versatile around progress analytics, while OpenSpace, Evercam and others focus more on reality capture and site monitoring.

1. Buildots — my first progress demo

buildots.com is probably the first company I'd put through a serious technical evaluation. It captures imagery using 360 cameras, drones or laser scans, then compares the resulting reality data against BIM and the schedule. It also provides delay forecasting rather than merely documenting what happened.

Importantly, Buildots launched superstructure tracking in July 2026, expanding its coverage beyond underground work and fit-out.

Test: Can it reliably answer “What percentage of Level 8 MEP installation is actually complete, and is the trade pacing behind schedule?”

2. OpenSpace — strongest overall platform

openspace.ai is the other must-evaluate. It combines 360° capture with Spatial AI and has expanded beyond documentation into progress analytics and field workflows. It acquired Disperse in 2025, bringing Disperse's progress-tracking technology into OpenSpace.

OpenSpace also has significant current deployment momentum: it reported 1,000+ data-center projects using its platform as of June 2026.

Test: Does its broader platform reduce the need to stitch together separate reality capture, progress and field-observation systems?

3. Evercam — particularly interesting for safety

evercam.com is worth considering if your concept involves fixed cameras continuously watching high-risk areas, rather than periodic 360° walks.

Its current safety offering emphasizes continuous visual records, proactive risk identification, PPE detection and historical reconstruction of incidents.

Test: Put cameras over crane zones, material-handling areas, site entrances and elevated work areas and measure meaningful alerts versus false positives.

4. TrueLook — camera infrastructure + AI

truelook.com is another camera-centric candidate. AWS describes its platform as combining construction cameras, live streaming, time-lapse and AI insights; its newer safety system uses computer vision for PPE detection.

I'd include it if 24/7 fixed-camera coverage is part of your desired architecture.

5. viAct — dedicated safety comparison

viact.ai is worth putting against Evercam/TrueLook specifically for safety CV. The category is increasingly capable of detecting PPE issues, restricted-area access, unsafe behaviors and equipment interactions.

I wouldn't assume the most impressive demo wins here. The critical metric is precision at operationally useful alert rates.

---

How I'd structure the evaluation

Rather than asking vendors to give you a generic demo, I'd give each the same 2–4 week pilot site and score:

• Progress accuracy: actual installed quantity vs BIM/schedule
• Schedule prediction: lead time before a delay becomes obvious to a superintendent
• Safety precision: useful alerts / total alerts
• Recall: percentage of deliberately seeded safety events detected
• Camera coverage: what percentage of meaningful site activity is actually observable?
• Labor burden: minutes/day required from field staff
• Integration: Procore, Autodesk Construction Cloud, Primavera P6, BIM, VMS/cameras
• Data ownership/API: whether you can export the underlying detections and imagery
• Deployment: edge vs cloud, bandwidth requirements, offline behavior
• Privacy: worker identification, retention, consent and union/workforce implications

One particularly important issue: don't evaluate “AI accuracy” in isolation. A system that detects 95% of hard-hat violations but generates 50 irrelevant alerts per camera per day may be less useful than one with lower theoretical recall but excellent prioritization.

My initial bake-off

I'd start with five:

1. Buildots — progress/schedule intelligence
2. OpenSpace — broad visual intelligence + progress
3. Evercam — continuous camera + safety
4. TrueLook — camera infrastructure + AI
5. viAct — dedicated safety CV

Then I'd add Doxel if automated BIM/quality verification is especially important.

The interesting architectural question is whether you actually want one platform. My suspicion is that the strongest solution may be OpenSpace/Buildots for periodic spatial progress intelligence + a fixed-camera safety platform for continuous monitoring. The two workloads have very different computer-vision requirements.

Test: Can it reliably answer “What percentage of Level 8 MEP installation is actually complete, and is the trade pacing behind schedule?”

2. OpenSpace — strongest overall platform

openspace.ai is the other must-evaluate. It combines 360° capture with Spatial AI and has expanded beyond documentation into progress analytics and field workflows. It acquired Disperse in 2025, bringing Disperse's progress-tracking technology into OpenSpace.

OpenSpace also has significant current deployment momentum: it reported 1,000+ data-center projects using its platform as of June 2026.

Test: Does its broader platform reduce the need to stitch together separate reality capture, progress and field-observation systems?

3. Evercam — particularly interesting for safety

evercam.com is worth considering if your concept involves fixed cameras continuously watching high-risk areas, rather than periodic 360° walks.

Its current safety offering emphasizes continuous visual records, proactive risk identification, PPE detection and historical reconstruction of incidents.

Test: Put cameras over crane zones, material-handling areas, site entrances and elevated work areas and measure meaningful alerts versus false positives.

4. TrueLook — camera infrastructure + AI

truelook.com is another camera-centric candidate. AWS describes its platform as combining construction cameras, live streaming, time-lapse and AI insights; its newer safety system uses computer vision for PPE detection.

I'd include it if 24/7 fixed-camera coverage is part of your desired architecture.

5. viAct — dedicated safety comparison

viact.ai is worth putting against Evercam/TrueLook specifically for safety CV. The category is increasingly capable of detecting PPE issues, restricted-area access, unsafe behaviors and equipment interactions.

I wouldn't assume the most impressive demo wins here. The critical metric is precision at operationally useful alert rates.

---

How I'd structure the evaluation

When evaluating computer vision (CV) startups for jobsite monitoring, it helps to separate them into primary categories based on how they capture data and what they optimize for (progress vs. safety).

Category 1: Progress & Schedule Tracking (BIM/Model Comparison)

These platforms use 360° cameras (worn on hard hats or mounted on tripods) or aerial capture to compare actual site imagery against your BIM models and master schedules (e.g., Primavera P6, MS Project).

1. • How it works: Hardhat-mounted 360° cameras capture site footage during regular walks. Their computer vision engine automatically maps the footage against your BIM model to create an objective, 24/7 digital twin.
   • Best for: Granular, task-level progress tracking, subcontractor performance management, and identifying schedule deviations weeks before they impact the critical path.
2. OpenSpace
   • How it works: Workers walk the site with a 360° camera attached to a hard hat, passively capturing video that is mapped automatically to blueprints. They also offer enhanced progress tracking via automated tools and hybrid human/AI verification workflows.
   • Best for: Fast, low-friction site documentation, visual progress sharing with stakeholders, and tracking percentage-completion by trade or area.

Category 2: Real-Time Safety & Hazard Detection (Fixed/Security Cameras)

These solutions plug into your existing fixed security cameras, temporary jobsite trailer/pole cameras, or edge devices to look for behavioral and environmental risks in real time.

1. Voxel AI
   • How it works: Leverages existing jobsite security or facility cameras, turning them into smart AI sentries using computer vision. It flags PPE non-compliance, restricted zone breaches, speeding vehicles, and ergonomic/positioning hazards.
   • Best for: Proactive, real-time safety alerts and generating behavioral data/coaching insights without requiring specialized helmet hardware.
2. Spot AI
   • How it works: A camera-agnostic video AI platform providing edge-to-cloud video analytics. It processes visual data to detect Focus Four hazards (falls, struck-bys, caught-in/between) and missing PPE (hard hats, vests, harnesses).
   • Best for: Fast deployment across diverse camera hardware without ripping and replacing existing security infrastructure.

Category 3: Aerial & Autonomous Intelligence

If your jobsites are large-scale civil, infrastructure, or heavy commercial projects requiring aerial overwatch.

1. DroneDeploy
   • How it works: Uses autonomous drone flights and ground captures coupled with vision-language models (VLMs). Their specialized AI agents automatically parse aerial data for site inspection, progress metrics, and risk identification.
   • Best for: Large earthmoving, exterior progress, volumetric analysis, and site-wide safety/quality aerial audits.

Evaluation Checklist for Your Team

• Hardware Friction: Do workers need to wear a camera consistently (Buildots/OpenSpace ), or can you leverage fixed infrastructure (Voxel/Spot AI )? High-friction hardware often leads to compliance drop-offs among subcontractors.
• Privacy & Union Nuances: If your sites are unionized, ensure the platform supports anonymization/face-blurring so workers don't feel micro-surveilled.
• Integration Ecosystem: Check whether the platform pushes data directly into your existing project management stack (like Procore or Autodesk Construction Cloud) rather than living on a siloed island.

To help narrow this down, let me know:

• Are you focusing more on interior vertical construction (buildings) or heavy civil/infrastructure?
• Do you already have fixed security cameras deployed on your sites, or are you starting from scratch?