Purpose
This policy explains how artificial intelligence (AI) features in the Caliber platform process, store, and protect client data. It states what Caliber’s AI does, what data it touches, and the commitments Avèro Advisors makes to every client whose data enters the platform.
This policy supplements the engagement contract between Avèro Advisors and the client. It does not replace that contract. Where the two differ, the engagement contract controls.
Scope
This policy covers every AI capability in the Caliber platform, across all modules and pricing tiers. It applies to all data a client provides to Caliber and to all output the platform generates from that data. It applies to Avèro Advisors staff, to the platform itself, and to the third-party AI providers described below.
What Caliber’s AI does
Caliber uses AI to accelerate the work of qualified consultants. Current AI capabilities include:
- Document analysis. Parsing client documents, including PDFs, Word files, and scanned images, and extracting structured content with source citations.
- Transcript synthesis. Analyzing interview transcripts to identify process steps, decisions, and exceptions, with speaker attribution.
- System data analysis. Analyzing event logs from client systems to identify process patterns, variants, and bottlenecks.
- Requirements and RFP support. Generating draft requirements and procurement content from evidence gathered during the engagement.
- Vendor evaluation support. Supporting the scoring of vendor capabilities against the client’s documented needs.
- Reporting. Producing summaries, diagrams, and exports for steering committees and project teams.
AI in Caliber assists people. It does not replace them, and it does not make decisions on its own. Human oversight applies to all AI output, described below.
Data Caliber processes
Clients decide what enters the platform. Depending on engagement scope, Caliber may process:
- Engagement documents such as policies, procedures, reports, forms, and correspondence provided by the client.
- Interview recordings and transcripts from engagement sessions.
- Event logs and operational data exported from client systems.
- Screen recordings and walkthrough captures made with client knowledge and consent.
- User account information for client staff who access the platform, limited to name, work email, and role.
Avèro Advisors asks clients to limit uploads to what the engagement requires and to flag any dataset containing sensitive personal information before providing it.
How Caliber protects client data
Tenant isolation
Each client’s data lives in an isolated tenant. No client can see, query, or infer another client’s data. AI features operate inside the tenant boundary and never combine data across clients.
Encryption
Client data is encrypted in transit and at rest, including all artifacts held in object storage.
Access control
Access is role-based and scoped to engagement participants. Only the Avèro team members and client users assigned to an engagement can access that engagement’s data.
Audit logging
The platform logs access to client data and the processing actions performed on it. Audit records are available to the client on request.
No training on client data
We train on the public record. Never on your data.
Caliber does not use client data to train AI models. This commitment has no exceptions:
- Client documents, transcripts, logs, recordings, and other artifacts are never added to any model training set.
- AI output generated from client data is never used to train models.
- Third-party AI providers used by Caliber are contractually prohibited from training on data submitted through the platform.
What “client data” means
For the purpose of this commitment, client data means the data a client provides to Caliber through an engagement. This is engagement-channel data: the documents, transcripts, recordings, logs, and account information that flow into the platform because a client placed them there for their engagement.
Independently sourced public information
Caliber maintains a knowledge base of government vendor capabilities that informs vendor evaluation. This knowledge base is built and kept current from independently sourced public information: publicly available records, vendor documentation, and information obtained through Avèro Advisors’ direct relationships with vendors. It is never built from client engagement data. Information a client provides through an engagement is never added to this knowledge base, and the knowledge base is never trained on one client’s data for use in another client’s engagement. The line is firm: the vendor knowledge base draws only on the public record and Avèro’s own vendor relationships, and client engagement data stays inside the client’s isolated tenant.
Third-party AI providers
Caliber’s AI features run on commercial large language model services accessed through their enterprise APIs, currently including Anthropic. When the platform sends data to a provider:
- Only the data required for the specific task is transmitted, and always over encrypted connections.
- The provider processes the data under terms that prohibit training on it and restrict retention.
- Providers do not receive client data for any purpose other than returning the requested output.
A current list of AI subprocessors is available to any client on request. Avèro Advisors notifies clients before adding a new AI provider that will process their data.
Accuracy, transparency, and human oversight
AI systems can make mistakes. Caliber is built around that fact:
- Evidence citations. AI-generated findings link to their sources: the document excerpt, transcript quote, or data query behind the claim. Clients can inspect the evidence for any conclusion.
- Confidence scoring. AI-generated content carries a confidence rating so reviewers know how strongly the evidence supports it.
- Consultant review. Qualified consultants review AI output before it informs any finding, recommendation, or client deliverable.
- Client validation. Client subject matter experts validate reconstructed content before sign-off.
No procurement decision, vendor score, or engagement finding rests on unreviewed AI output.
Retention and deletion
Client data is retained on the schedule set in the engagement contract. Unless the contract states otherwise, Avèro Advisors deletes client artifacts from the platform after engagement close, following any transition period agreed with the client.
A client may request deletion of its data in writing at any time. Avèro Advisors confirms completion of the deletion, subject to legal hold or records obligations that require otherwise.
Client responsibilities
Clients help keep the platform private by doing three things:
- Provide only data you have the authority to share, and limit uploads to what the engagement requires.
- Flag sensitive personal information, such as health, financial, or personnel records, before providing it so handling can be agreed in advance.
- Manage your own users. Tell Avèro Advisors promptly when a member of your team no longer needs access.
Legal and public records compliance
Avèro Advisors serves state and local government clients and understands that engagement records may be subject to public records laws. Avèro Advisors cooperates with clients to meet public records obligations that apply to engagement content.
Caliber’s data handling is designed to comply with the privacy laws applicable to the engagement. Where an engagement involves regulated data categories, specific handling terms are set in the engagement contract before that data enters the platform.
Changes and contact
Avèro Advisors reviews this policy at least annually and whenever Caliber’s AI capabilities change materially. Clients are notified of material changes before they take effect.
Questions about this policy, subprocessor lists, audit records, or deletion requests: [email protected]