AI can draft due diligence responses from documents in minutes. This page explains how it works in ODD, what the security controls are, and how ChatGPT, Claude and MCP fit in.
AI is used in operational due diligence to read manager documents, draft questionnaire responses with sources cited, raise flags and detect changes since the last review. A person approves every answer. AlternativeSoft's AI uses a zero-retention external LLM (Anthropic's Claude by default), stores data only in AlternativeSoft on Microsoft Azure, and is protected by AES-256 encryption at rest, TLS 1.2+, MFA, role-based access and SOC 2. Its read-only MCP server also lets AI assistants query quantitative fund data.
The slowest part of ODD is finding answers in documents. AI reads the manager's offering documents, prior DDQs, regulatory filings and fact sheets, drafts a response to each question and cites the source. Where it finds nothing, it says so, rather than guessing. An analyst reviews and approves every answer before it is used.
AI also helps downstream: Raising flags where documents show a concern, and identifying what changed since the previous cycle so the annual review starts from the differences. The result is up to 90% less time on DDQ completion, with the analyst still deciding.
Due diligence data is confidential. These are the controls to ask any vendor about, and how AlternativeSoft answers.
| Question | AlternativeSoft | |||
|---|---|---|---|---|
| Does the AI provider keep my data? | No. The external LLM is zero-retention. | |||
| Which LLM is used? | Defaults to Anthropic's Claude; clients can lock the model or bring their own. | |||
| Where is data stored? | Only in AlternativeSoft, on Microsoft Azure. | |||
| How is data protected? | AES-256 at rest, TLS 1.2+ in transit, MFA and role-based access. | |||
| Is it audited? | SOC 2, with annual independent audits. | |||
| Can retention be configured? | Yes, configurable retention. |
Allocators ask whether they can simply use ChatGPT or Claude for fund analysis. General assistants are useful for drafting and summarising, but on their own they have no access to your reconciled fund data and no audit trail. AlternativeSoft's MCP server connects them to it: AI assistants such as Claude, ChatGPT, Microsoft 365 Copilot, GitHub Copilot and Codex can query quantitative fund data through a read-only connection.
MCP is read-only and covers quantitative data only; it does not cover DDQs or documents. It needs the REST API enabled and an MCP licence, arranged through support. See AlternativeSoft MCP and the AI security page.
Ask these before you approve any AI tool for due diligence data.
Confirm the AI provider does not retain or train on your data.
Confirm where data is stored and who can access it.
Confirm encryption at rest and in transit.
Confirm MFA and role-based access.
Ask for SOC 2 and independent audit evidence.
Confirm that a person approves AI output before use.
Bring a questionnaire and a set of documents. We will show the drafts, the sources cited and the security controls behind them.