The bounded pilot
The bounded pilot, made explicit.
A first document workflow can be scoped and put into operation when the case fits a pilot. It works best for intake, extraction, drafting support, status replies and deadline tracking, with human review where needed.
- Scope, exclusions and success criteria agreed in writing
- Human review on exceptions and sensitive outputs
- No forced phase two if the pilot does not justify a next step
What the pilot covers
Built to make clear what is in scope, what is out of scope and under which conditions the first workflow operates.
Best fit for
- document intake from email, scans or client portals
- extraction of names, dates, DNI fields, references and obligations
- template-based first drafts for review
- status replies or document requests from approved sources
- deadline and missing-document tracking
Included in the pilot
- scope and success criteria agreed in writing
- isolated EU-hosted environment and DPA before live data
- setup around the tools already used in that workflow
- review queue for exceptions and low-confidence outputs
- team training, operating documentation and the pilot report
Not included in the pilot
- full replacement of the firm management system
- multi-workflow rollout across the whole organisation
- historical backfile cleanup
- custom DMS build
- client-facing automatic decisions without human review, unless explicitly agreed
Pilot, step by step
From the scoping call to a first operational document workflow, when the case fits.
Scope, exclusions and environment locked
Scope, exclusions, success criteria, DPA and isolated EU-hosted environment are agreed before any live data enters the pilot.
We map the current workflow and bottlenecks
We walk the workflow with the people who run it today. Inputs, approvals, exceptions, handoffs and failure points are all written down before build starts.
We configure the workflow around your current tools
Rules, folders, templates, review states and integrations are configured around the tools already used in that workflow.
We test on real files with your team
The workflow is exercised on representative files, with low-confidence cases and edge cases sent to review rather than forced through.
Live. This is what you leave with.
We put it into production, train your team, deliver operating documentation, a results report against the agreed success criteria and a 90-day roadmap with a recommendation on whether to maintain, expand or stop.
Risk controls
How risk is handled in practice on the pilot.
DPA before live data
The DPA is signed before any live client data enters the system.
Isolated environment per firm
Each pilot runs in an isolated environment rather than shared client infrastructure.
Regional EU deployment only
We use regional EU deployment for the agreed pilot scope. No optional global or data-zone modes unless explicitly agreed.
No model training on your data
The provider terms we use exclude training foundation models on your data.
Audit trail
Inputs, outputs and reviewer actions can be logged for traceability within the agreed workflow.
Low-confidence goes to review
Low-confidence extractions or classifications are flagged for review, not pushed through automatically.
Sensitive outputs stay human-approved
Client-facing or legally sensitive outputs remain under human approval unless explicit written rules say otherwise.
No forced phase two
If the pilot misses the success criteria agreed in writing, there is no obligation to continue.
Before you start
Common questions from firms considering the pilot.
What exactly is included?
One workflow, one fixed scope and one first operational version. The pilot covers the agreed logic, review, training, documentation and results report.
What is explicitly not included?
Firm-wide rollout, full practice-management replacement, historical cleanup, or a custom DMS build. Those are separate decisions, not hidden assumptions.
What happens if outputs are wrong?
Low-confidence cases are routed to review. Sensitive outputs do not go out automatically unless explicit written rules say they can.
What happens after the pilot?
You keep the operational workflow, documentation, training and pilot report. From there you decide whether to run it internally, keep us on for support or stop if a next step is not justified.
What do we own afterwards?
You keep the documented workflow, operating rules and the live setup agreed in the pilot. Any ongoing hosting or support arrangement is scoped separately.
Tell us where the work gets stuck today
We will see whether there is a first case worth doing.
Tell us where the workflow slows down today. We will tell you whether there is enough operational friction to start with a first case and which part should stay under review.
Prefer email? Request scope and conditions.