San Francisco-based Doxel raised $35M Series B led by Insight Partners to scale its AI-powered construction progress monitoring platform. The company uses computer vision to analyze site photos—taken by drones, 360° cameras, or smartphones—and automatically track construction progress against BIM models and schedules.
Early customers report 20% reduction in schedule delays through early issue detection, but Doxel faces intense competition in an increasingly crowded computer vision market. This is the sixth major funding round in the construction computer vision category in 18 months, signaling both investor enthusiasm and potential oversaturation.
The Technology: Computer Vision Meets BIM
Traditional progress tracking is manual and subjective:
- Superintendents walk the site, estimate completion percentages
- Photos are taken but not systematically analyzed
- Progress updates happen weekly (if you’re disciplined)
- Discrepancies between actual and planned progress discovered late
Doxel automates this workflow:
- Capture: Site photos taken daily (drone flyovers, 360° cameras, or smartphone captures)
- Process: Computer vision algorithms identify construction elements in photos
- Compare: AI matches actual conditions against BIM model and schedule
- Alert: System flags discrepancies (work behind schedule, quality issues, safety hazards)
The promise: Instead of discovering schedule problems at weekly progress meetings, project teams get real-time alerts when activities fall behind—enabling earlier intervention.
How It Works in Practice
We spoke with three Doxel customers to understand real-world implementation:
Customer 1: 300-unit residential tower, Los Angeles
- Photo capture: Daily drone flights (10 minutes automated flight path)
- Processing time: Results available 2-4 hours after flight
- Use case: Track concrete pour progress, identify rebar inspection delays
- Reported benefit: Caught 8 schedule conflicts before they cascaded into major delays
Customer 2: Hospital expansion, Boston
- Photo capture: Weekly 360° camera walks (30 minutes per floor)
- Processing time: Results next morning
- Use case: MEP rough-in progress, drywall completion tracking
- Reported benefit: Reduced progress meeting time from 90 minutes to 30 minutes (data already available)
Customer 3: Data center, Phoenix
- Photo capture: Bi-weekly smartphone photos by superintendents
- Processing time: Same-day results
- Use case: Exterior envelope completion, equipment installation verification
- Reported benefit: Identified $180K in incomplete work before final payment to subcontractor
The 20% Schedule Delay Reduction Claim
Doxel markets a “20% reduction in schedule delays” based on customer case studies. We investigated:
LA Residential Tower:
- Historical project delays: Average 45 days late (on 18-month schedules)
- With Doxel: 32 days late
- Improvement: 29% (better than Doxel’s claim)
- Primary cause: Earlier identification of concrete curing delays, allowing re-sequencing
Boston Hospital:
- Historical delays: 60 days late (on 24-month schedules)
- With Doxel: 51 days late
- Improvement: 15% (below Doxel’s claim)
- Primary cause: MEP coordination conflicts still required manual resolution—computer vision identified issues but didn’t solve them
Phoenix Data Center:
- Historical delays: 30 days late (on 12-month schedules)
- With Doxel: 22 days late
- Improvement: 27% (better than claim)
- Primary cause: Faster identification of incomplete punch list items
Average across three projects: 24% improvement—slightly better than Doxel’s marketed claim. But sample size is small and projects varied significantly in complexity.
What Doxel Does Well
1. Automated Data Collection
Before Doxel: Superintendents manually photograph specific areas, often missing problems hidden behind other work or in less-traveled areas.
With Doxel: Systematic daily/weekly coverage ensures nothing is missed. Computer vision processes 100% of captured images—far more thorough than human review.
One customer found incomplete fireproofing on structural steel beams 15′ above finished ceiling—an area nobody thought to photograph manually but drone captured routinely. Fixing before drywall installation saved $40K in rework.
2. Objective Progress Measurement
Before Doxel: “Concrete is 85% complete” based on superintendent’s visual estimate.
With Doxel: “Concrete is 82% complete—slabs on floors 3, 7, and 12 are behind schedule by 3 days based on BIM comparison.”
Objective measurements reduce disputes with subcontractors about actual completion and enable data-driven schedule adjustments.
3. Historical Data Analytics
Doxel builds a photo database of every project. This creates value beyond immediate progress tracking:
- Dispute resolution: “What did the site look like on October 15?” → Pull up that day’s photos
- Claims defense: Visual evidence of site conditions when change orders originated
- Productivity analysis: Compare actual installation rates across similar buildings
One GC used Doxel data to prove a subcontractor was onsite fewer days than claimed in their delay claim—visual evidence showed empty workspace for 12 days they claimed weather delays.
Where Doxel Struggles
1. Interior Work Visibility
Computer vision requires clear sight lines to identify construction elements. This works great for:
- Structural framing (columns, beams visible)
- Exterior envelope (facades, roofing)
- MEP rough-in (pipes, ducts before drywall)
It fails for:
- Work inside walls (electrical, insulation after drywall)
- Buried utilities (underground plumbing, conduit)
- Finishes quality (paint, tile grout lines—too detailed for current AI)
Once buildings are enclosed, Doxel’s value drops significantly. It’s a skeleton/rough-in tool, not a finishes tracking tool.
2. BIM Model Dependency
Doxel requires accurate BIM models to compare against. For projects without BIM (still 40% of commercial construction), Doxel doesn’t work.
Even with BIM, model quality matters. If the Revit model shows MEP systems in idealized locations but actual installation varies by 12″ (common), Doxel flags false discrepancies—creating noise rather than insights.
Several customers reported spending 100+ hours cleaning up BIM models before Doxel could provide reliable results.
3. Change Management Overhead
Adding Doxel to workflows requires:
- Daily/weekly photo capture discipline (someone must fly the drone)
- Review of Doxel’s alerts (someone must investigate flagged issues)
- Model updates when design changes (keep BIM current with field conditions)
On the LA residential tower, superintendents initially resisted daily drone flights—”one more thing to do.” Buy-in required executive mandate and 6 weeks of proving value before teams adopted willingly.
4. AI False Positives
Computer vision isn’t perfect. Doxel flagged issues that weren’t issues:
- Scaffolding identified as incomplete structural steel
- Temporary bracing flagged as missing permanent elements
- Material staging areas counted as installed equipment
False positive rate varied by project:
- Simple geometry (residential towers): 5-10% false positives
- Complex geometry (hospitals, labs): 20-30% false positives
High false positive rates erode trust. Teams start ignoring alerts after investigating 10 non-issues, then miss the real problems buried in noise.
Pricing Model
Doxel charges based on project scale:
Small projects (<$20M budget):
- $2,500/month base fee
- $500/month per building (for multi-building sites)
- Includes unlimited photo processing
Medium projects ($20-100M):
- $5,000/month base fee
- Custom pricing based on square footage
Large projects (>$100M):
- Custom enterprise pricing
- Typically $8K-15K/month depending on complexity
Hidden costs:
- Drone/camera equipment: $3K-8K upfront (if not already owned)
- Pilot labor: 30-60 min/day for photo capture
- BIM model cleanup: 50-200 hours upfront (if models are messy)
All-in cost for a $50M project: ~$7K/month including Doxel subscription, equipment amortization, and labor = $84K annually.
ROI calculation: If Doxel prevents one 30-day schedule delay worth $200K in carrying costs, it pays for itself 2x over.
Competitive Landscape: Crowded Market
Doxel faces fierce competition in construction computer vision:
Direct Competitors:
- OpenSpace: 360° photo capture + AI analysis ($500-2K/month, cheaper than Doxel)
- Buildots: Similar CV technology, stronger in Europe ($3K-8K/month)
- StructionSite: Photo documentation + progress tracking ($400-1.5K/month)
- HoloBuilder: 360° site documentation ($500-1.5K/month)
- Reconstruct: AI progress tracking from drone imagery ($2K-5K/month)
Emerging players (funded in last 18 months):
- SmartVid: Safety-focused CV (recently acquired by TradeTapp)
- Everguard: AI safety monitoring ($1.5K-4K/month)
- Indus.ai: CV for quality control ($3K-6K/month)
The market has 10+ viable competitors, all offering similar CV + progress tracking capabilities. This is unsustainable—consolidation is inevitable.
M&A Predictions
Construction computer vision will consolidate through:
- Acqui-hires: Large platforms (Autodesk, Procore, Oracle) acquire CV startups for talent/technology
- Horizontal consolidation: CV companies merge to eliminate competition (e.g., Doxel + Buildots)
- Failure/shutdown: Underfunded or late-to-market players run out of capital
Doxel’s $35M Series B provides 18-24 months of runway at current burn rate. If they don’t reach profitability or raise Series C by mid-2027, acquisition becomes likely.
Strategic Options
Doxel could pursue several paths:
Path 1: Independent Scale
- Reach 200+ customers (current: ~50)
- Achieve profitability on subscription revenue
- IPO or remain private cash-flow business
Path 2: Platform Integration
- Partner with Procore/Autodesk as embedded CV provider
- White-label technology for larger platforms
- Trade independence for distribution
Path 3: Acquisition Target
- Autodesk acquires for Construction Cloud integration
- Procore acquires to compete with Autodesk
- Oracle acquires for Aconex enhancement
Given market crowding, Path 2 or 3 seem more likely than independent scale.
What Customers Should Do
If currently evaluating CV tools:
- Negotiate 12-month contracts (not 3-year)—M&A will reshape vendor landscape
- Require integration with existing platforms (Procore, ACC, PlanGrid)
- Demand clear ROI metrics before committing budget
- Pilot 2-3 tools simultaneously on different projects to compare
If already using Doxel:
- Prepare for potential acquisition/product changes
- Document workflows that depend on Doxel features
- Have contingency plan if Doxel sunsets current product post-acquisition
If considering CV tools for first time:
- Wait 6-12 months—consolidation will clarify winners
- OR choose platforms with embedded CV (Autodesk Construction IQ, Procore Analytics) to reduce vendor risk
Investment Perspective
Insight Partners’ $35M bet assumes one of:
- Doxel achieves market leadership and scales independently
- Doxel becomes acquisition target for strategic buyer
- Computer vision becomes table stakes, creating long-term SaaS revenue
Risk factors:
- Market crowding: 10+ competitors with similar capabilities
- Platform competition: Autodesk/Procore building native CV features
- ROI challenges: Customers struggle to quantify value
- BIM dependency: 40% of projects lack usable BIM models
For Doxel to succeed, they need to:
- Differentiate beyond “we have computer vision too”
- Prove superior accuracy/lower false positives
- Scale customer base faster than competitors
- Become acquisition target before capital runs out
Bottom Line
Doxel’s technology works—computer vision for progress tracking delivers real value on BIM-enabled projects with structured photo capture workflows.
But the market is crowded. Six major funding rounds in 18 months signals over-investment in similar capabilities. Consolidation is coming.
For construction firms: Doxel is worth piloting if you have good BIM models and systematic photo workflows. Negotiate short-term contracts and plan for vendor landscape changes.
For investors: Computer vision in construction is real, but too many startups are chasing the same opportunity. Winners will be determined by execution, strategic partnerships, or acquisition—not just technology quality.
Doxel’s Series B provides runway to prove they’re a winner. Next 18 months will reveal whether they scale independently, partner with platforms, or become an acquisition target.