The most expensive subcontractor problems in residential construction are the ones that develop slowly. A billing rate that creeps up between invoices. A scope that expands through incremental verbal approvals. A productivity pace that drops off midway through a phase. None of these announce themselves — they accumulate quietly until the final invoice arrives and you're looking at a number that doesn't match the contract.
The human brain is not well-suited to detecting slow drift across multiple simultaneous data streams. AI is.
Billing Pattern Detection
Every subcontractor invoice logged in BLT updates the sub's running financial profile: estimate total, invoiced to date, paid to date, remaining scope. But the AI looks at more than the current totals — it looks at the pattern of how invoices have been arriving.
A sub who invoices $8,000 in month one and $8,000 in month two and then $14,000 in month three has changed their billing pattern. That change might be justified — maybe month three covers more scope. But it's worth noting. BLT's AI flags billing rate changes that exceed a configurable variance threshold, so the manager can ask about the change before paying the invoice rather than after.
Scope-Estimate Mismatch Alerts
When a sub's cumulative invoices first cross 80% of their original estimate, BLT surfaces an alert with context: how much has been invoiced, how much remains, and a note on what scope appears to still be outstanding. The manager can then review the sub's scope documentation and assess whether the remaining estimate is sufficient for the remaining scope.
At 100% of estimate with scope remaining, the alert escalates. At this point, any additional invoice from the sub represents a change order situation — and should be treated as one, with formal documentation, regardless of how casual the original scope change conversation was.
Cross-Project Comparison
If you're running three builds and the same sub is working on two of them, BLT's AI compares their billing pace across projects. If they're billing 30% faster on Project 2 than on Project 1 for equivalent phases, that's signal. Either the projects are genuinely different (different complexity, different spec), or something changed in how the sub is working that's worth understanding.
Photo Evidence Integration
Progress photos in BLT are tagged to phases and tasks. The AI can flag a situation where hours are being logged against a phase but progress photos haven't been updated in several days. This isn't an accusation — it's a prompt. "Rough-in plumbing has 40 hours logged this week but no new photos since Monday. Is everything on track?"
The combination of financial data, time data, and photo documentation creates a multi-dimensional picture of subcontractor progress that no single data stream could provide. The AI reads all three simultaneously and surfaces when they're inconsistent.