Your partners are pushing for a Claude rollout by Q3, while the associates have already gone rogue using Gemini on their personal accounts. Meanwhile, compliance is demanding guardrails, and the actual deal data that makes AI useful is still siloed across six different tools.
You’ve been handed the mandate every ops leader in private capital is hearing right now: make AI work for the firm. No new headcount or budget line, just a timeline set by stakeholders with competing priorities. Everyone agrees AI can help, but how do you build an infrastructure that connects it to your deal data, relationships, and pipeline without creating a security incident or another silo?
We’ve moved past worrying about AI capability. The real hurdle now is a fragmented data architecture that keeps your most valuable relationship insights siloed and unreachable.
Many private capital firms are trying to bolt AI tools onto disconnected systems. The research assistant can’t see your CRM. The investment committee dashboard pulls from a spreadsheet someone exported last Tuesday. LP reports get manually compiled from three sources every quarter. Without connected data, AI is a faster search bar. It’s useful, but not necessarily transformative.
The firms pulling ahead have solved this by treating their CRM as a data layer rather than a database. Instead of a system where people manually log activities and pull reports, they’ve built an architecture where deal and relationship data flows directly to AI tools like Claude, ChatGPT, Gemini, and Copilot through protocols like MCP (Model Context Protocol), and into warehouses like Snowflake for dashboards, LP reporting, and custom analytics.
That’s the shift: rebuilding the connective tissue between the tools you already have.
How leading firms are building AI-native infrastructure
One firm already built this. Bessemer Venture Partners connected a locally hosted instance of Claude to Affinity, Salesforce, and their other data sources through MCP, creating an internal system called “Brain.” This resulted in 70 different workflows feeding from the same data foundation, serving 20 partners simultaneously.
“We don’t want people just writing memos using Claude. We want to remove the parts of an analyst’s job that really don’t maximize value.” — Alec Robins, Bessemer Venture Partners
An investor asks “Who do we know at [Company]?” and gets an answer in 30 seconds by querying Affinity’s relationship data, replacing five to ten minutes of manual searching. Analysts reclaimed 234 hours per year because AI handles the rote research (scraping blogs, compiling company profiles) so they can focus on building relationships and doing diligence. The firm went from zero AI adoption to standard practice in 18 months, with full R.I.A. compliance maintained throughout.
The lesson from Bessemer is about architecture. Instead of buying 70 AI products, they built one data foundation and connected everything to it.
How does MCP connect a CRM to AI tools?
Every AI integration project starts with the same question: how does the model access our data? MCP is the open standard that answers it. When your CRM supports MCP, tools like Claude, ChatGPT, Gemini, and Copilot can read your deal pipeline, relationship history, and activity data without custom integrations or API engineering.
For ops teams, this means Affinity’s MCP Server turns the CRM into a live data source that any AI tool can access out of the box. There’s no middleware to configure, no CSV exports to schedule, and no six-month integration project to staff. An associate can ask Claude to prep a meeting briefing, and Claude pulls the relationship history, recent communications, and deal status from Affinity in real time. Setup is measured in hours, not quarters, and the connection respects your existing Affinity permissions structure, so compliance doesn’t need to open a new workstream.
This is different from the approach most CRM vendors take, where AI features are locked behind add-on pricing or limited to a single proprietary interface. With MCP, the AI layer is platform-agnostic. Your firm chooses which AI tools to use, and Affinity supplies the data.
What’s the best way to integrate CRM data with Snowflake?
While AI tools need real-time access, analytics teams need historical depth. MCP handles the first; Snowflake Data Shares handle the second.
Affinity’s Snowflake connector pushes deal and relationship data directly into your firm’s data warehouse, with no engineering resources required and no scheduled exports to manage. Once deal data lives in Snowflake alongside your other sources, your data team builds whatever they need. This could be IC dashboards with real-time pipeline status, LP reports that incorporate relationship metrics automatically, or custom analytics that combine deal flow with portfolio performance.
Affinity’s API suite, with dedicated endpoints for meeting transcripts, notes, and read access, rounds out the architecture. We connect MCP for AI access, Snowflake for analytics, APIs for everything custom.
What can firms build on this architecture?
The compound effect is what matters. Once deal and relationship data flows into AI tools and analytics platforms, each new use case becomes incremental rather than a standalone project.
Consider the investment committee meeting. Today, someone on the ops team spends Monday morning exporting pipeline data from the CRM into a spreadsheet, reformatting it for the IC deck, and emailing the package to partners by noon. When Affinity connects to Snowflake, that dashboard builds itself from live data. When it connects to Claude via MCP, a partner can ask “What’s changed in our Series B pipeline since last Tuesday?” and get a sourced answer in seconds. The ops team shifts from assembling the report to analyzing the data in it.
The same pattern applies across the firm. LP reports that automatically incorporate relationship metrics and deal activity cut quarterly prep from days to hours. Research agents that know your firm’s history surface smarter recommendations because they can see which companies you’ve already evaluated, who you spoke with, and what happened. (For a deeper look at specific MCP workflows, we’ve published a companion article to show what’s possible [link to MCP blog post].)
David Hefter, BlackRock’s AI Champion for Investments, puts the underlying principle simply: differentiation comes from connected data, not proprietary AI models. When AI tools have access to your firm’s relationship intelligence, every workflow they power gets smarter. BlackRock’s research team now covers five times more companies per day, and meeting briefings that took one to two hours are assembled in 15 minutes.
The architecture advantage
Adding more AI tools won’t move the needle if your data is still siloed. The firms that win this transition are the ones building a unified foundation where information flows between systems without manual entry or constant cleanup.
Affinity’s 2026 Predictions Report makes this case across the industry: as AI becomes strategic rather than tactical, the competitive gap widens between firms that have built connected infrastructure and those still bolting point solutions onto disconnected data.
And the behavioral data backs it up. In The Invisible Edge, an analysis of hundreds ofPE firms over 24 months, the top-quartile firms convert outreach into introductions at 17x the rate of the bottom quartile, with nearly identical team sizes. It usually isn’t a lack of effort holding these firms back. The real bottleneck is that their internal systems aren’t actually talking to each other, so all that manual work never scales.
Ops owns that architecture. The CRM is the foundation.
Get started with Affinity’s MCP Server →
Talk to our integrations team →
Frequently asked questions
How does AI improve deal sourcing in private equity?
AI improves deal sourcing in PE by automating the research, relationship mapping, and pipeline management that traditionally consume analyst time. When AI tools are connected to a CRM through protocols like MCP, they can query deal history, score relationship strength, and surface warm introduction paths in seconds. Firms using this architecture report reclaiming hundreds of hours per analyst per year and expanding research coverage by 5x.
How do leading PE firms use AI to maintain competitive advantage in deal flow?
Leading PE firms are building AI-native infrastructure that treats the CRM as a data layer rather than a static database. Bessemer Venture Partners, for example, connected Claude to Affinity and their other systems via MCP, creating 70 AI workflows on a single data foundation. The competitive advantage is the architecture that connects deal and relationship data to every tool the firm uses.
How to automate CRM pipeline management?
CRM pipeline management becomes automated when the CRM connects to AI tools and analytics platforms natively. Affinity’s MCP Server allows AI assistants to read and query pipeline data directly, while Snowflake Data Shares push live deal data into dashboards and reporting tools. The result is pipeline visibility that updates itself rather than depending on manual data entry or scheduled exports.
How to connect a CRM to AI tools like Claude or Gemini?
Affinity’s MCP Server connects the CRM to Claude, ChatGPT, Gemini, Copilot, and other AI tools through the Model Context Protocol standard. Once connected, AI tools can query your relationship data, deal pipeline, and activity history directly, with no custom integrations or middleware required. Setup is out of the box, and data access respects your existing Affinity permissions.
What AI workflows can you build with a CRM and MCP?
Affinity’s MCP Server supports both read and write operations, so teams can query and update their CRM without leaving their AI tool. On the read side, that includes pulling call transcripts from a recent meeting or asking “Who at our firm has the strongest relationship with [company]’s CEO?” On the write side, users can create notes attached to open opportunities, add industry tags to company records, move deals to new pipeline lists, update deal stages, or set follow-up reminders—all through natural language in Claude, ChatGPT, Gemini, or Copilot.
.png){kind=logo}
