Approach
The repetitive work that still matters is exactly the work to automate. Not the judgment sitting on top of it.
That is a narrower claim than most of this industry makes, and it is the reason the engagements look the way they do. A system that takes the mechanical first pass can be validated against historical data, measured against the process it replaces, and handed to the client's own team. A system that makes the determination cannot — not in work where someone has to answer for the result to a regulator, an auditor, or a board.
How engagements run
Three stages. Each one ends with something you can evaluate before committing to the next.
Discovery
Two to four weeks. We work against your historical data and establish what the current process actually costs — not what it is assumed to cost. On the compliance engagement that meant measuring review at up to 180 minutes per package and finding that inspectors dropped checks when time ran short. On the audit engagement it meant establishing that randomly selected encounters carried a 4.21% error rate, which became the baseline every later claim was measured against.
If the data does not support a system that beats the current process, this is where we say so.
Pilot
Four to eight weeks, with a target agreed in advance. The pilot runs on your data, in a form you can inspect. The utility engagement ran a first test cycle before anything was deployed permanently; it found theft 7.5 times more often than unprioritized inspection, and that number came from the test cycle rather than from a projection.
Production and handover
Deployment into the workflow your team already uses, then training so they can run and retrain it without us. Typical engagements run four to six months end to end — the audit system started September 2018 and first deployed February 2019; the collection-tagging system started June 2022 and deployed that October.
What you get
Not a model file. A model on its own is a liability: it decays, nobody remembers how it was trained, and the person who built it has left.
- The running system, deployed in your environment against your data.
- The transformation and training pipeline, documented, so the model can be rebuilt from scratch rather than only re-run.
- Your team trained to retrain it. On the utility engagement the client's own data science team took over running and retraining the model, which was the point of the engagement rather than a concession at the end of it.
- The previous process left intact as a fallback until the new one has earned the trust to replace it.
How we price
Fixed scope, quoted before the work starts, with the stages above as natural decision points. Ongoing maintenance and retraining runs as a separate retainer where the client wants it — one audit system has been in continuous production since 2019 on quarterly retrains, and one client relationship is in its eighth year.
The comparison worth making is not against other vendors. It is against hiring. A senior data scientist in Canada runs well north of $140k fully loaded, takes months to find, and leaves you with one person, no production deployment history, and a single point of failure the day they resign. A fixed-scope engagement costs a fraction of one salary, has a defined end, and finishes with your team trained.
What we don't do
- Systems that make the determination. The system ranks, flags, and drafts. A person decides. This is in every case study as a standing section, not as a disclaimer.
- Work we can't measure. If there is no historical record to validate against and no agreed baseline, there is no way to tell whether the result is real.
- Staff augmentation. We are not a source of contract data scientists to sit inside your team.
- Anything that requires us permanently. If the system cannot be handed over, it was built wrong.