Method and technologyMethod, security and AI
Per project
A clear method. Controls defined for each matter.
Written scope, evidence preservation, human review and traceable deliverables guide our work. Infrastructure, AI use and security controls are assessed and documented for each project.
01 · Status reports
What every status report shows.
Status of each stage, volumes processed, review metrics, delivered documents and costs to date, sent on the agreed calendar to in-house counsel and the law firm, with the custody log attached on request.
Illustrative report · sample data
Example- Summary
- Volumes and metrics
- Deliverables
- Costs
- Custody log
Project · Arbitral production
02 · Security
Know where your data is held and how it will be handled.
Data stored and processed in Brazil, with access controls and security requirements documented for each engagement.
- 01
Data in Brazil
Data is stored and processed in Brazil, including processing by AI providers. The proposal documents the infrastructure and access controls for the engagement.
- 02
Encryption
Encryption and key management checked against the selected infrastructure. Evidence integrity documented through hashes and custody records.
- 03
Role-based access
Each person reaches only the project and functions they need. Access granted, reviewed and revoked with a record.
- 04
Audit trail
Relevant operations are logged with the responsible person, date and item, according to the tools and scope. The proposal defines which records accompany delivery.
- 05
Retention and disposal
Retention periods, preservation obligations and disposal procedures are agreed in the contract, with execution records suited to the environment.
- 06
Incident response
Our own response plan, with client notification within a contractual deadline, we apply to ourselves what we recommend.
03 · AI governance
AI with human review and transparent terms of use.
Models are selected for the task and assessed against the document collection. Specialists check results; counsel defines legal criteria. Data use, retention and training terms are checked before deployment.
“AI helps find and organise information. The analysis must show its sources, criteria and limitations.”
- 01
Data use in AI
Each provider's data-use, retention and training terms are checked and documented before any data is submitted.
- 02
Multi-model per task
Models are selected per task and tested against project documents. Selection considers quality, data handling and cost.
- 03
Sampling and metrics
Recall, precision and elusion calculated by statistical sampling in every round and recorded in the methodology memorandum.
- 04
Traceable analysis
Model version, instructions, parameters and checks are recorded to explain the process and support reassessment, within each tool's limitations.
- 05
Human in the loop
Relevance and sensitive-data calls go through human quality review before any production; privilege calls belong to the client's counsel.
04 · Method
From scope to delivery in four steps.
The same steps in every practice, whatever the engagement model.
| Step | Stage | What happens | Output |
|---|---|---|---|
| 01 | Assessment and scope | Initial assessment of the request, confirmation of response timing and a written proposal covering scope, deliverables and fees. | Proposal |
| 02 | Preservation and collection | Evidence preservation and collection with integrity and custody records, using the infrastructure agreed for the project. | Custody record |
| 03 | Analysis under human control | AI prioritises, specialists decide, and sampling metrics are recorded in every round. | Metrics memo |
| 04 | Report and support | A deliverable that separates fact, inference and conclusion, with clarifications when needed. | Final report |
Contact
How would this approach work for your matter?
Share the context, estimated volume and deadline. We assess the request and confirm the next steps towards a defined scope and engagement.