All projects
AI & ML
Logistics
Nova Freight Copilot
An LLM copilot that answers operations questions in seconds.
Client
Nova Freight
Delivered
2025 · 4 months
Team
6 specialists
Category
AI & ML
Overview
What we built
Built a retrieval-augmented copilot on top of Nova's TMS, WMS and finance systems that answers complex ops questions with cited sources.
Business impact
Outcomes that moved the P&L
0 min
Time saved / query
0%
Auto-resolved
0
Active users
0.0M
Docs indexed
The challenge
- Ops leads copy-pasted between six systems to answer routine questions.
- Tribal knowledge lived in SharePoint, email, and one senior analyst's head.
- Leadership wanted an AI experience — without hallucinations on customer data.
How we approached it
- Indexed 1.4M documents and 42 structured systems into a governed retrieval layer.
- Built an evaluation harness with 900 golden questions before shipping to users.
- Rolled out with human-in-the-loop feedback and quarterly model reviews.
What changed
- 12 minutes saved per operations query, on average.
- 84% of queries fully resolved by the copilot with cited sources.
- Copilot now used by 780 employees across ops, finance and support.
Project gallery
Inside the delivery
Grounded retrieval layer
Every answer cites its sources.
Evaluation harness
900 golden questions gate every release.
Feedback loop
Human review continuously improves quality.
Technology
Built with
The tools and platforms we chose for this engagement.
OpenAI
LangChain
Pinecone
Azure
TypeScript
PostgreSQL
"It is the single most impactful piece of software we've rolled out in five years. Adoption was near-universal in a month."
Renata Kowalski — COO, Nova Freight
Talk to our team
Have a similar challenge?
Tell us about your goals. We'll put together a scoped conversation with the right specialists.
Ready to amplify your data?
Book a 30-minute strategy call. We'll map your data landscape and identify the three highest-leverage moves.