ASPR Sales Analyst AI — agentic threat model

Not certain from the listing — The specific foundation models powering ASPR are not disclosed. Standard LLM risks apply, including prompt injection that could alter email drafts or CRM update payloads, and potential data leakage of training/fine-tuning data if proprietary models are used.

ASPR ingests highly sensitive data including CRM records, meeting transcripts, sales playbooks, and ex-employee deal data. This creates a high-value target for data exfiltration, embedding inversion, and knowledge-base poisoning, which could corrupt coaching insights or leak proprietary sales strategies.

The agent framework orchestrates CRM updates, document creation, and email drafting. Vulnerabilities here include tool misuse (e.g., unauthorized CRM writes or malicious email generation via prompt injection) and memory poisoning from malicious transcripts or external documents.

Not certain from the listing — The hosting infrastructure, network isolation, sandboxing of document generation processes, and secret management for CRM/email API keys are not detailed in the public directory.

Not certain from the listing — There is no mention of real-time guardrails, LLM firewalls, or observability logging to detect anomalous CRM writes or malicious email drafts before they are presented to the user.

Not certain from the listing — Compliance certifications (such as SOC 2, GDPR, or ISO 27001) and specific identity/access management controls governing which sales reps can access ex-employee deal data are not specified.

Not certain from the listing — While the agent interacts with external ecosystems (CRMs, email clients), it is unclear if it coordinates with other autonomous agents or operates within a multi-agent marketplace.