A one-company internal pilot built from 15+ months of real operational data, with expansion to two–three additional carriers next.
Invoices, payments, bank activity, factoring records, email, and operating data do not agree. Operators discover revenue leakage and payment risk too late.
Owner-operators, small-fleet owners, dispatchers, and back-office staff at carriers with two–ten trucks.
Twenty interviews and five prototype tests disproved the dashboard-first hypothesis. The MVP now detects problems, links each alert to source evidence, and recommends an action.
Detect a narrow set of operational problems and recommend the next action. Testing showed that operators wanted problems surfaced, not another place to inspect data.
Alternative considered: Build a general dashboard or broad chatbot.
Trade-off: The MVP covers fewer workflows, but each alert has a clear usefulness measure.
Link each alert to invoice, payment, bank, factoring, and email evidence. Operators needed to understand why an alert appeared before acting on it.
Alternative considered: Generate autonomous recommendations without source traceability.
Trade-off: Entity resolution is harder than summarization, but creates a more trustworthy product.
20 interviews showed operators needed problems surfaced—not another dashboard.
90%+ invoice and payment-status detection accuracy.
3+ actionable insights per carrier each week.
70%+ weekly action or usefulness rate20 carrier interviews and five prototype tests.
90%+ invoice/payment-status accuracy.
3+ actionable insights per carrier each week.
70%+ weekly action or usefulness rate.
The internal pilot reached at least 90% invoice/payment-status detection accuracy.
The product generated at least three genuinely actionable insights per carrier per week.
At least 70% of pilot users acted on or explicitly confirmed one recommendation each week.
Definitions and sources are shown so every result can be examined in an interview.
Measured invoice and payment-status classification in the one-company internal pilot.
Genuinely actionable insights generated per carrier each week during the pilot.
Pilot users who acted on or explicitly confirmed at least one useful recommendation each week.
Discovery across owner-operators, small-fleet owners, dispatch, and back-office roles.
The pilot replaced a dashboard-first assumption with a narrower product thesis: find the problem, explain the evidence, and make the next action obvious.