Audition AI

Primary category: enterprise-ai-assistant. Also ai-governance-platform, dlp.

One-liner — A compliance-first enterprise AI platform purpose-built for hedge funds and regulated financial firms: it deploys secure AI assistants, agents, and automations inside the customer’s own Microsoft Azure tenant (or on-prem), with identity-bound access, DLP, and immutable audit trails baked in.

Categories — enterprise-ai-assistant, ai-governance-platform, dlp

What it does

Audition AI lets a regulated fund use frontier LLMs against its internal documents and workflows without sending data to public AI services, while producing the audit and compliance evidence (SEC, Reg S-P) examiners expect. It targets real fund workflows — analyst research assistants, DDQ/RFP drafting, compliance policy Q&A, investor-letter and reporting automation, meeting/email summarization, due-diligence extraction.

The product is a build-and-deploy AI platform, not a single chatbot. Components (vendor marketing):

• Local Agents — agentic workloads on Windows, Linux, cloud, or on-prem.
• Sidekick — personal desktop agents with browser, shell, and file access.
• Agentic Swarms — coordinated parallel agent teams for high-volume work.
• Enterprise Intelligence Engine — “sub-second” search across SharePoint, Egnyte, email, Teams, CRM, databases, and local files, with citations back to source documents.
• Persona Marketplace — pre-built managed skills/assistants.
• Integration Forge — MCP servers + arbitrary APIs and connectors (S3, Azure, Office 365, Slack, SQLite, Azure Managed SQL).

The differentiating idea is a secure enterprise AI workspace where governance is part of the platform, vertically tuned for capital-markets compliance — rather than a horizontal assistant you bolt a guardrail onto.

Where it sits in the stack

Primary home is enterprise-ai-assistant (the ux layer) — it is the sanctioned, in-tenant assistant/agent platform a fund’s people actually use, an alternative to the horizontal players (openai-chatgpt-enterprise, anthropic-claude-enterprise, microsoft-365-copilot, glean). It carries two strong secondary feet: ai-governance-platform via its GRC console (policy enforcement, violation tracking, compliance dashboards) and dlp via built-in PII/SSN/PCI pattern matching plus behavioral “Generative Rules.” Its permission-aware retrieval also overlaps entitlement-aware-rag, and its org-wide chat visibility / violation alerts edge toward comms-surveillance.

Risk-control profile:

• Sensitive-data — strong. In-tenant deployment, permission-aware retrieval (“if a user doesn’t have access to a document, the AI doesn’t either”), and DLP keep fund data inside the customer’s cloud. This is the product’s center of gravity.
• Egress — partial. Keeping inference inside the customer’s Azure tenant and brokering model access (instead of users pasting into public ChatGPT), plus DLP and audit on outputs, constrains where data flows; but there is no outbound network-egress firewall or shadow-AI CASB.
• Untrusted-input — weak/partial. “Generative Rules” provide some prompt-content policy enforcement, but no published prompt-injection defense for the Sidekick agents that have shell + browser + file access — an open risk the vendor’s own blog gestures at.

It lives in the yellow zone: a trusted UX/agent surface through which untrusted content and privileged tools meet.

Deployment & architecture

Three deployment models: fully-hosted SaaS (Pro tiers, from ~$200/mo single-user); deployed inside the customer’s Azure Resource Group / tenant (“your compute, storage, AI resources” — the Business tier); or on-prem. Listed on the Azure Marketplace (publisher: Saberin Data Platform Inc).

• Identity/access: Azure Entra ID (microsoft-entra), described as zero-trust with admin-granular control of M365 permissions and permission-aware responses.
• Models: multi-model via Azure AI Foundry — Claude, GPT, Gemini, Llama, Mistral; claims 11,000+ models and new frontier models “within hours of release” (marketing).
• Integrations: MCP + APIs (“Integration Forge”); SharePoint, Egnyte, Office 365, Teams, Slack, CRM, S3, Azure SQL/SQLite; policy enforcement extends to AI IDEs (Cursor, Copilot).
• Governance plumbing: immutable audit trails, real-time email violation alerts, org-wide chat visibility, AI-generated compliance reports/dashboards. No published SIEM, DSPM, or non-Entra IdP integrations — a gap (absence of evidence).

Positioning & differentiators

The core differentiator is vertical focus on regulated finance — built by a team with ~17+ years serving capital-markets compliance, with a “runs in your cloud / your tenant” data-residency pitch and a “compliance-first,” SEC / Reg S-P framing. Secondary differentiators: dual-layer guardrails (Generative Rules + traditional DLP) and citation-grounded answers.

Against nearest neighbors:

• vs the horizontal assistants (openai-chatgpt-enterprise, anthropic-claude-enterprise, microsoft-365-copilot, glean) — those are broad, deep, well-resourced general-purpose platforms; Audition AI trades breadth/scale for a finance-specific, in-tenant, governance-baked-in package and white-glove fit.
• vs ai-governance-platform vendors (credo-ai, holistic-ai, modelop) — those are model-risk / attestation platforms (SR 11-7); Audition AI’s GRC is usage governance over its own AI workspace, not a model-risk register.
• vs witnessai / harmonic-security (CASB-for-AI) — those govern employees reaching outward to third-party AI; Audition AI is the sanctioned inward platform itself.

A notable credibility signal: Eldon Sprickerhoff (founder of eSentire, MDR pioneer) is listed as Cybersecurity Advisor — independently corroborated.

Ownership, funding & M&A

Independent and private. Audition AI is a product of Saberin Data Platform, Inc., a Saberin Group company. Saberin Group was founded 2007 as a data-first automation/reporting firm for capital markets; Audition AI itself appears to be a recent (2025-era) product offshoot. HQ: Hauppauge, NY (Long Island).

Leadership: Sharon Saberin (Founder/CEO), Benjamin Saberin (Founder / Developer-Architect), Eldon Sprickerhoff (Cybersecurity Advisor, eSentire founder), and Michael “Mags” Maguire (Director of Customer Success).

No external/VC funding was found in any search (no seed/Series rounds, no TechCrunch/Crunchbase funding entries; Crunchbase org pages returned 403). The structure strongly implies it is bootstrapped/self-funded under the privately held Saberin Group — an operating software/services company, not a VC-backed startup. No funding figure is recorded here. No M&A. Ownership confidence: high (clear private parent); funding: “none found,” confidence medium (absence of evidence).

CTO / hedge-fund lens

For the wiki’s core reader — a hedge-fund / asset-manager CTO — this is one of the few vendors purpose-built for exactly that buyer, which makes it Day-1-relevant as a candidate for the sanctioned enterprise assistant slot. Its strengths line up with fund requirements: in-tenant data residency, permission-aware retrieval, DLP, immutable audit, and explicit SEC / Reg S-P framing — the things that otherwise make a fund nervous about ChatGPT/Copilot. It can also double as the agent platform (research, DDQ/RFP, reporting automations) and a usage-governance console.

The caveats are vendor maturity and verification: it is a small, bootstrapped shop, its “30+ hedge fund customers” claim is unverified, and key architecture details (runtime isolation, prompt-injection defenses for tool-using agents) are not publicly documented — the latter a real concern given Sidekick agents wield shell, browser, and file access. No formal model-risk attestation (SR 11-7) capability — its GRC is usage governance, not a model-risk register. Fit: high as a finance-vertical assistant/agent option to evaluate, weighed against the scale and roadmap of the horizontal incumbents and standard vendor-risk diligence on a small provider.

Competitors / alternatives

openai-chatgpt-enterprise, anthropic-claude-enterprise, microsoft-365-copilot, glean, credo-ai, holistic-ai, witnessai, harmonic-security

Open questions / to verify

• Exact Audition AI product launch date and whether it’s a standalone entity vs. a Saberin Data Platform product line (appears 2025-era).
• The “30+ hedge fund customers” claim — no independent corroboration.
• Whether there is any outside funding (none found; presumed bootstrapped).
• Real architecture details: runtime isolation and prompt-injection defenses for the tool-using (shell/browser/file) Sidekick agents.
• Pricing above the published $200/mo Pro tier (Business/Enterprise are custom).

Sources

• Audition AI homepage + GRC/Agents/Leadership/Team pages — fetched 2026-06-30 — supports: in-tenant Azure deployment, product components, dual-layer DLP + Generative Rules, Entra zero-trust, permission-aware retrieval, leadership, Hauppauge NY HQ, Saberin parent, hedge-fund target, “30+ funds” claim (vendor/marketing); confidence: high (that they claim it) / med (accuracy)
• Saberin Group / Saberin Data Platform — fetched 2026-06-30 — supports: parent operating company founded 2007, private/bootstrapped structure, capital-markets data heritage; confidence: high

History

• [2026-06-30] Created and researched from vendor request. Established primary fit as a finance-vertical, in-tenant secure enterprise-ai-assistant + agent platform, with ai-governance-platform (usage GRC) and dlp as secondary tags. Ownership: private/bootstrapped product of Saberin Data Platform Inc (Saberin Group, founded 2007), Hauppauge NY; no external funding found; no M&A. Risk coverage: strong sensitive-data control, partial egress control, weak untrusted-input screening (flagged Sidekick shell/browser/file agents as an open risk). Cached 1 source file.