Traversaal Labs

Production AI,
in the wild.

We're a forward-deployed engineering team. We take on two or three clients at a time, embed with their team, and ship to production. Below is what that has actually looked like - anonymised where we have to, specific where we can be.

Let's talk →
Engagement
6–12 wks typical
Deliverable
Working production system
Hand-off
Docs · training · code
Stack
Yours, not ours
Selected work

What shipping AI
actually looks like.

Real work, real clients, real results. Anonymised where we have to be - specific everywhere we can.

Operations · Delivery review
Decision layer live queue · today
At-risk · 12Auto · 47
#4821 · Whirlpool 27" Rangedamaged appliance · Atlanta GA
risk
#4819 · Vanity 48"contractor install risk
review
#4816 · LG Washercustomer confirmed window
auto
#4821Auto-reroute
Causehub damage
Actionreroute replacement
Ownerno human needed
~73% auto-resolved15× faster review daily scope
~73% automated2025
Retail
Closing 73% of delivery review cases automatically, and giving one retailer 8x the scope with no new headcount
Fortune 500 Retailer
Agentic AIRetailProduction
Learning · Expert archive
Cited answer workspace pilot · customer-ready
100% cited
Learner question
What does our research say about psychological safety on hybrid teams?
The guidance points to explicit norms, leader availability, and a visible trail back to source material.
source lockedfollow-up ready
Framework · team climate.94
Assessment guide.87
Facilitation guide.81
10M+ pages indexed100% cited3 sources matched
10M+ pagesEnterprise
Enterprise learning / L&D
Turning ten million pages of leadership expertise into a learning tool users can trust
Global leadership development organisation
RAGEnterpriseProduct discovery
Demand · LTO #312 · 14-day
Forecast before first receipt new LTO launch
Under target
MAPE
18.23%
▼ vs <20% target
Supply
8–12w
ahead
Coverage
~50%
of category
Day 0Day 4Day 8Day 14
18.23% MAPEQSR
Restaurants / QSR
Forecasting limited-time menu items before they have a single receipt
National restaurant chain
ForecastingProduction APIQSR
destination · research · Lisbon
Destination research assistant traveller-facing
live brief
Ask4 days in Lisbon · food + viewpoints · no tourist trapsrunning
38 reviews14 sources
Source scanmaps · guides · traveller notes
3.2s
Local signal filterAlfama · Belém · Príncipe Real
live
Draft answer
Start in Alfama, save Belém for early day 2.Avoids coach-tour peaks and keeps the food stops close to viewpoints.
30% sales liftTravel AI
Travel / Consumer AI
A destination research engine that put a travel company ~18 months ahead of the category, and lifted sales 30%
AI travel planning startup
Agentic AITravelHosted infra
How an engagement runs

Discovery → production,
in three phases.

Every engagement follows the same shape. Enough structure to be predictable, enough flexibility to fit the reality of what we find.

Phase 01
Discover
Audit goals, systems, data. Find where AI creates real value - and where it doesn't. No assumptions, no pre-sold solutions.
Phase 02
Design
Prioritise use cases. Architect the solution. Build the roadmap with the team who'll own it after we leave.
Phase 03
Deploy
Build, test, ship. Hand off with documentation, training, and code. The stack stays yours - we don't lock you in.
Your data is already there

If it should be driving
decisions, let's talk.

Two or three clients at a time. 30-minute call, no pitch deck, no slides about “the AI revolution”.