From Diligence Calls to Investment Memos: How Fellow and Claude Streamline PE Deal Work
•
10
MIN READ
AI Summary by Fellow
Every PE analyst knows the tradeoff. You sit down for a management call with a target company, you have a list of questions you need to get through, you need to steer the conversation, and you need to capture what was said. You can do two of those three well. Not all three.
As an analyst at a large SEC-registered private credit fund put it on a recent call:
"The way most diligence calls work is you have a list of questions you're trying to get through or steer a conversation a certain way, and referencing your list while taking notes diverts your attention."
That divided attention has a cost, and the cost scales. Each analyst on a typical PE team reviews 100 to 150 deals per year, each one requiring management calls, customer references, expert network sessions, and counterparty discussions. Multiply that across a deal team, and the firm is running thousands of hours of conversations where the person asking the questions is also the person writing them down.
The result is predictable. Notes get thinner as calls get longer. Critical details, the offhand comment from a CFO about working capital assumptions, the customer reference who contradicted management's churn numbers, get missed or forgotten. By the time the investment memo is drafted, the raw material is already degraded.
If you're evaluating ways to close that gap, see how a secure AI meeting assistant captures diligence calls so your analysts can focus on the conversation, not the transcript.
What an AI investment memo actually is
An AI-assisted investment memo is a first-draft investment committee memo generated from the structured transcripts, summaries, and notes of a deal's diligence calls, fed into a large language model that synthesizes across the full body of conversations.
The AI does not replace the analyst. It compresses the gap between raw diligence and a working draft, so the analyst spends time sharpening analysis and judgment rather than reconstructing what was said across 40 hours of calls. The workflow typically looks like this:
A secure AI meeting note taker like Fellow captures every diligence call (management meetings, customer references, expert network calls, site visits, CEO follow-ups) with high-quality transcription and structured summaries.
Those summaries follow a custom template that maps to the firm's diligence framework, so the same categories (management quality, financial KPIs, competitive positioning, red flags) appear consistently across every call.
The meeting data connects to an LLM through a native integration (such as Fellow's Claude Connector). The analyst prompts the model to synthesize across all calls for a given deal, and the model returns a structured first-draft memo with quotes, themes, and contradictions surfaced.
The analyst reviews, revises, and sharpens the draft. Human judgment stays in the loop; the grunt work of synthesis does not.
Why generic AI note takers fail in PE
Most AI note takers are built for general business meetings. That assumption breaks in private equity for four specific reasons.
Consent and recording posture: PE diligence involves external counterparties, management teams under NDA, expert network calls with strict recording restrictions, and advisors who cannot be recorded at all. One PE firm described unauthenticated sharing links as a "hygiene factor dealbreaker," and ruled out multiple otherwise-capable tools on that basis alone.
Recordkeeping implications: The SEC's Rule 204-2 requires SEC-registered investment advisers to preserve books and records relating to their advisory activities for at least five years. Dodd-Frank extended that rule to hedge funds and private equity firms. If an AI transcript includes investment recommendations, portfolio discussions, or action items, it may be treated the same as an email or written memo under the rule. A note taker without admin-enforced retention, deletion, and access controls forces the firm to retrofit compliance around a tool that was not designed for it.
MNPI and information barriers: Deal conversations routinely contain material non-public information. The SEC's 2026 examination priorities emphasize that recordkeeping obligations remain comprehensive regardless of the channel, and firms must be able to produce complete records quickly across every platform employees use for business purposes. Tools that allow unauthenticated external sharing, lack role-based access controls, or cannot block recording on specific counterparty domains fail the CCO review.
Integration architecture: A summary locked inside the note taker is useful to the person who attended the call. It is not useful as an input to an investment memo, because there is no path for an LLM to read across 40 hours of diligence conversations in a secure, governed way.
These four gaps are why PE buyers often arrive at Fellow after evaluating and disqualifying other tools. If these gaps sound familiar, a purpose-built alternative solves them at the platform level rather than through workarounds.
How Fellow captures the diligence call
Fellow is a secure AI meeting notetaker built for regulated industries such as finance. It captures diligence calls across Microsoft Teams, Zoom, Google Meet, and in-person meetings, with the capture method chosen to fit the conversation.
Botless recording for external-facing calls: Fellow's botless recording runs through the desktop app and captures audio locally in the background, without sending a visible bot into the call. This matters for diligence conversations where a bot in the participant list would spook management teams, advisors, or counterparties. Internal participants are always notified, external guests see no visible indicator, and the audio stays inside the firm's governance framework.
Bot-based recording for internal meetings: When explicit in-meeting disclosure is preferred (IC meetings, internal partner debates, board prep), Fellow supports bot-based recording with the same security posture.
Zoom Native Capture with bot fallback: For compliance-heavy orgs that prefer Zoom's built-in recording consent mechanism, Fellow supports Zoom Native Capture.
Opt-in, not auto-record: Fellow supports manual recording triggers as a workspace-level policy, so analysts control when a call is recorded rather than having the tool default to capturing everything.
Domain and attendee-based recording blocks: Admins can configure Fellow to automatically prevent recording when specific domains (counsel, regulators, counterparties) or individuals are on the call, taking the compliance burden off the user.
The capture method is the first layer. What makes the data useful for investment memo drafting is what happens after the call ends.
Summaries built for deal work
Every diligence call in Fellow produces a summary. What makes those summaries usable as investment memo inputs, rather than generic meeting recaps, is two things: custom templates and structured output.
Custom templates aligned to your diligence framework: Fellow supports custom output templates with no cap on how many a workspace can run. A PE firm can create templates for management Q&A capture, customer reference synthesis, financial KPI extraction, red flag surfacing, and competitive positioning notes. Every call tagged to a deal runs through the relevant template, producing consistent categories across the entire diligence process. One Fellow AE described a large customer with more than 300 custom templates in production, which gives a sense of the ceiling.
Multi-LLM reconciliation for accuracy: Fellow's summary generation uses a multi-model reconciliation process designed to surface what was actually decided, not just everything that was said. This matters for diligence calls where a founder changes their answer mid-call, or where a customer walks back a critical comment. Generic tools capture both statements as action items. Fellow's summaries are built to reflect the conclusion, not the noise.
Summaries delivered within minutes: Post-call, a structured summary is available within roughly two minutes, so analysts can act on the conversation while context is fresh.
The output is not a transcript dump. It is structured meeting data organized by deal, call type, and custom template, which is exactly the shape an LLM needs to draft a memo.
From call to Claude: the native MCP integration
Fellow has a native Claude connector built on the Model Context Protocol (MCP). Analysts can connect Fellow to Claude once, then prompt Claude to read across all Fellow meetings relevant to a specific deal, company, or portfolio theme.
The workflow in practice:
The deal team tags every diligence call in Fellow to a specific deal (for example, "Project Orion").
Over the course of diligence, Fellow captures 40 hours of conversation across management calls, customer references, expert network sessions, and CEO follow-ups. Each call runs through the firm's custom diligence templates.
When it's time to draft the IC memo, the analyst opens Claude, connects Fellow as a data source, and prompts: "Draft an investment committee memo for Project Orion based on all diligence calls. Include sections on management quality, revenue model, competitive positioning, customer feedback, and red flags. Quote directly where possible."
Claude reads across the full body of Fellow meeting data and produces a first-draft memo with quotes, themes, and surfaced contradictions.
The analyst reviews, sharpens, adds analysis Claude can't infer from the calls alone (valuation work, deal structure, return modeling), and produces the final memo.
The discoverability of this pattern is a story in itself. One technology lead at a financial services firm in a recent Fellow evaluation described how they found Fellow specifically because they were searching for products that connect to Claude. Anthropic has been investing in this segment (Anthropic launched Claude for Financial Services in mid-2025 with partnerships including Databricks, FactSet, Morningstar, PitchBook, S&P Global, and Snowflake), and PE firms with Claude Enterprise deployments are actively shopping for the integrations that extend Claude's usefulness into their deal workflow. Fellow's MCP connector is one of those integrations.
If your firm runs Claude and needs to close the gap between meeting data and memo drafting, this is the native path. Book a call with our team to see the connector in action.
What this looks like in practice: customer reference synthesis
The full IC memo is the end state. A more concrete, provable use case along the same workflow is customer reference synthesis.
Every PE diligence process includes 10 to 30 customer reference calls. Analysts currently write them up by hand, one by one, and the synthesis (what customers agree on, where they diverge, which quotes support which thesis) happens in the analyst's head or in a scratch doc.
With Fellow and Claude:
Every reference call is captured with a standard reference-call template: customer context, product usage, what's working, what's broken, renewal likelihood, competitive displacement.
After all reference calls are complete, the analyst prompts Claude: "Across all customer reference calls for Project Orion, produce a synthesis organized by theme. For each theme, note the number of customers who raised it, direct quotes, and any contradictions with management's framing."
Claude returns a structured synthesis with verbatim pull-quotes organized by theme, flagging contradictions between customer sentiment and management's diligence answers.
The synthesis that used to take a senior analyst a full day of work arrives as a first draft in minutes. The analyst still owns the interpretation, the framing, and the judgment. The mechanical work of reading across 30 transcripts and organizing themes is gone.
Compliance without compromise
AI-assisted memo drafting is only viable if the compliance layer is solid. The finance-specific controls that make Fellow deployable at SEC-registered PE firms:
Zero-day retention for recordings and transcripts
Fellow supports workspace-wide, admin-enforced policies (such as zero-day retention) to delete audio and transcripts after processing, while preserving AI-generated summaries, action items, decisions, and key takeaways. Many firms treat the summaries as analyst notes rather than communicated records, which aligns with how an emerging consensus views internal-only AI-generated content as not constituting a "communication" record under Rule 204-2 unless it is externally shared. Retention for recordings and transcripts can be configured independently.
Admin-enforced recording restrictions
Domain-based and attendee-based blocks prevent Fellow from joining meetings with counsel, regulators, or sensitive counterparties, enforced at the policy level regardless of individual user behavior.
No unauthenticated sharing links
Admins can remove "anyone with link" sharing from the workspace entirely. This is a dealbreaker for most PE firms and is one of the specific controls that disqualifies many competing tools.
SOC 2 Type II, GDPR, HIPAA
Standard enterprise compliance posture. Fellow does not train AI models on customer data, with contractual guarantees from all AI sub-processors.
Global Relay integration
For firms with communications archiving obligations, Fellow's Global Relay integration supports compliant archiving of notes and transcripts.
Super Admin API for SEC exams
Full transcripts, summaries, and metadata can be exported in JSON for audit purposes, with C-suite-level access controls.
Explicit consent capture
Pre-meeting disclosure emails, audible consent trails embedded in recordings, and a participant consent workflow that produces an auditable consent log.
The broader regulatory direction matters here. The SEC's 2026 examination priorities emphasize recordkeeping comprehensiveness across all channels, with a focus on whether firms can produce complete records quickly regardless of the platform employees use.
The off-channel communications enforcement wave of 2022 to 2024 produced billions in fines, and the same scrutiny is extending to AI-generated records. Adopting an AI notetaker without admin-enforced governance creates the same category of exposure. Adopting one with proper controls is how firms stay ahead of the regulator.
Beyond Claude: why the MCP architecture matters
The Fellow MCP Server exposes meeting data through the Model Context Protocol, an open standard. Any MCP-compatible AI tool can read from Fellow in the same way Claude does. For firms that have standardized on ChatGPT Enterprise, or that run a multi-model stack, the same memo-drafting workflow works with a different model at the other end.
The practical implication: adopting Fellow today does not lock the firm into a single AI vendor.
Frequently asked questions
What is the best AI meeting assistant for private equity firms?
The best AI meeting assistant for private equity firms combines botless recording, admin-enforced zero-day retention, no unauthenticated sharing links, SOC 2 Type II certification, and a native LLM integration for synthesis across diligence calls. Fellow is built specifically for this use case, with custom output templates for deal-specific workflows and a native Claude connector for first-draft investment memos. General-purpose AI notetakers typically fail the CCO review on recording governance or sharing controls.
How do PE deal teams use AI to draft investment memos?
PE deal teams use AI to draft investment memos by capturing all diligence calls in a structured AI meeting assistant, then feeding the resulting transcripts and summaries into a large language model through a secure integration. The LLM synthesizes across the full body of conversations to produce a first-draft memo with quotes, themes, and contradictions. Human analysts review and sharpen the draft, adding analysis the model cannot infer (valuation, deal structure, return modeling). Fellow's native Claude connector is the most common integration for this workflow.
Is an AI meeting assistant compliant with SEC Rule 204-2?
An AI meeting assistant can be compliant with SEC Rule 204-2, but only if it supports admin-enforced retention policies, access controls, and governed sharing. The rule requires SEC-registered investment advisers to preserve books and records of advisory activity for at least five years in an accessible format. AI-generated transcripts and summaries that relate to investment recommendations, advice, or securities transactions may fall under the rule when shared externally. Fellow supports configurable retention (including zero-day deletion of recordings and transcripts while preserving summaries), admin-enforced governance, Super Admin API exports for audits, and Global Relay archiving integration.
Can I use AI notetakers for external diligence calls without a visible bot?
Yes, botless recording captures meeting audio through the desktop app without sending a visible bot into the call, so external participants see no change in the meeting environment. Fellow's botless recording is governed by the same admin-enforced retention, access, and sharing policies as bot-based recording, and internal participants are always notified that recording is active. This matters for diligence calls with management teams, advisors, or counterparties where a visible bot would alter the conversation.
Does Fellow work with Claude?
Yes, Fellow has a native Claude integration built on the Model Context Protocol. Once connected, Claude can read across all Fellow meetings to draft investment memos, synthesize customer reference calls, surface decisions across management diligence, or answer cross-meeting queries like "what did the CFO say about working capital?" The integration is available out of the box and is the most common path for PE firms running Claude Enterprise who want to connect their meeting intelligence to their AI workflow.
What about firms using ChatGPT or other LLMs instead of Claude?
Fellow's MCP Server is model-agnostic. Any MCP-compatible AI tool can read from Fellow in the same structured way Claude does. Firms that standardize on a different model (or run a multi-model stack) get the same memo-drafting workflow. The meeting data, custom templates, and governance controls carry over regardless of which LLM the firm adopts.
How does Fellow handle MNPI on deal calls?
Fellow handles MNPI through admin-enforced recording restrictions (domain and attendee-based blocks), no unauthenticated sharing links, role-based access controls, and configurable retention policies. Admins can prevent Fellow from joining meetings with specific counterparties or domains entirely. Deal meetings are private by default, visible only to invited attendees. For firms with stricter posture, zero-day deletion of recordings and transcripts removes the raw data entirely while preserving AI-generated summaries as analyst notes.
Your diligence calls already contain the memo
The material for the IC memo is already in your diligence calls. Every management Q&A, every customer reference, every expert network session is a layer of the argument your analysts are eventually going to write up. The only reason that process takes a week of grinding through notes is that the raw material is trapped in formats that do not connect to each other.
Fellow changes that. Diligence calls become structured, searchable, compliant meeting data. A native Claude connector turns that data into a first-draft memo. Your analysts spend time on judgment, not reconstruction.
Every diligence process without this workflow is context your team can't search, commitments no one can reference, and synthesis that happens by brute force. That's the status quo that Fellow was built to replace.
Book a call with our team for a walkthrough of the Claude integration and Fellow's finance-specific compliance controls.
Record, transcribe and summarize every meeting with the only AI meeting assistant built with privacy and security in mind.
