In development, early access open
Governed AI for your most sensitive client data
The Brand Research Vault combines a private data lake, automated screening and AI report drafting in an isolated environment on AWS, with logging and human review at every step. The data never leaves the boundary.
Why research data needs its own platform
General-purpose AI tools weren't designed for data that clients have contractually forbidden you to expose.
-
Data of every size
Files range from single-kilobyte spreadsheets to multi-terabyte, impression-level campaign datasets.
-
Federal and state privacy
Clients and audiences span all US states, so the environment is designed with federal privacy laws and the strictest combination of state requirements in mind, not a single baseline.
-
Sensitive goes beyond personal data
Client names and first-party brand data are treated as sensitive, alongside traditional personal information.
-
Assurance clients can verify
Clients expect documented, verifiable guarantees that their data stays isolated and is never used to train third-party models, not informal promises.
What the platform does
-
Secure ingestion
An encrypted web upload with scoped, time-limited access replaces shared-drive links, email attachments and desktop transfers.
-
Brand Research Vault
A two-stage data lake, raw and curated, organized so each client's data is kept logically separate. Storage tiering keeps very large files affordable.
-
Automated screening
Finds personal information, client names and first-party brand data using custom identifiers. Unscreened files never reach curated storage.
-
AI on your data
Claude on Amazon Bedrock works on curated data over private endpoints. Prompts and responses stay inside the network and are never used for training.
-
AI report drafting
Claude drafts a DOCX or PDF from a consistent template, based on the client, campaign and survey type an analyst selects. An analyst reviews and approves every report before release.
-
Audit and retention
Every step is logged and monitored, and lifecycle rules expire data on schedule, with deletion on request.
How data moves through the platform
The DFS-AI environment runs in an isolated private network on AWS, with each client's data kept logically separate. Files go in, screened insight comes out, and the raw data stays inside.
- Secure uploadEncrypted web upload, TLS 1.2+
- Raw storageEncrypted object storage
- Automated screeningSensitive-data discovery
- Curated data lakeSeparated by client
- Claude on Amazon BedrockPrivate AI endpoint
- AI report draftingDOCX or PDF, for analyst review
- Data boundaries and lifecyclePer-client storage, automatic expiry, deletion on request
- Model data handlingPrompts and responses stay in the private network and never train models
- Identity and accessSingle sign-on, MFA, separate admin and analyst roles
- Audit and monitoringAudit logs, monitoring, threat detection, security checks
Where the platform is going
We're building in stages: crawl, walk, run. Each stage ships on top of the same governed foundation.
-
Now
In development
- Secure environment and the Brand Research Vault (per-client data lake)
- Automated screening for sensitive data
- AI report drafting with analyst review
- Audit logging and retention controls
-
Next
Planned
- Harmonization, analysis and evidence agents across studies and waves
- Governor agent: policy checks on every agent action
- Client insight portal with plain-language questions over approved data
-
Later
Exploring
- Scorecard and benchmark agents
- Presentation-ready slide output
- Partner integrations, starting with data retention
Roadmap items describe current plans and may change.
Join the early access program
We're onboarding a small number of research and media teams as the platform comes online. Tell us about your data and your reporting needs.