IThe question that used to take ten minutes now takes thirty seconds
A lookup that used to cost five or ten minutes of screen-hopping — who else was in that seed round with us? — now takes about thirty seconds: someone types the question into Slack and the answer comes back in the channel. That's the entire promise of the CRM Slack integration we build for VC funds, and it isn't a mockup. We call it the Affinity Chat Bot — a plain-English layer over Affinity that lives inside Slack — and there are multiple versions of it deployed across client funds today, each tuned to that fund's own setup.
The time saving isn't a number I invented to sell you something. It's Affinity's own. In their case study, Stephen Lantz, a Principal at Bain Capital, describes the same collapse: a post-meeting task that used to sprawl across multiple screens and apps, now handled in a single natural-language request. Same pain, same fix. We just put it where the fund already talks all day, so the question and the answer land in the same window.
Picture the alternative, because it's what most funds still do: a partner throws out a question in a meeting, and someone opens a fourth browser tab to go dig for an answer that's technically already in the CRM — just nowhere near where the question got asked.
IIWhy funds are still digging through the CRM by hand
Here's the uncomfortable part. People dig because the CRM is rarely trustworthy enough to just glance at.
Original 2026 survey research on venture firms puts CRM field completion at roughly 30–40% — most records are half-filled at best. So even when the data exists, nobody's sure it's current, which means they go verify it manually anyway. These aren't cheap systems to ignore, either: the same research finds a mid-size firm spends $60,000–$80,000 a year across three or four record-keeping tools, and a slim majority (54%) run Affinity as their system of record. The data has been paid for. It just can't be reached from where the questions get asked.
Meanwhile Affinity's own research finds 76% of investors pull from four or more data sources to research a single deal. Four tabs, minimum, per question. Then there's the tax of the tools themselves. Slack's own workplace research found that 68% of workers lose at least 30 minutes a day just toggling between apps, and more than half say the switching actively makes it harder to do the real work. Stack that on top of what one operator study calls "the stupid work" — an estimated 30–40% of investment-team hours lost to logging, chasing, and reformatting instead of writing the actual memo — and you can see why "just ask the CRM" is the wrong instruction to give a human. You're asking them to be the query engine.
IIIWhat we actually built: a plain-English layer over Affinity, living in Slack
The idea is simple: stop making people be the query engine. Let them ask in words.
Under the hood, the bot wraps the Affinity API and surfaces it two ways — inside Slack, built on Slack's Block Kit so answers come back as clean, formatted messages instead of a wall of raw data, and as a Custom GPT for the people who live in ChatGPT instead. It reads across the things a fund actually asks about: companies, deals, co-investors, notes, and network relationships. And it doesn't only read — it can write structured updates back, so a note or a stage change can start as a sentence in Slack.
The part that makes it useful rather than a novelty is the tuning, and it's worth being specific about what that means. Every fund's Affinity looks different. One fund's early pipeline stage is called "Tracking"; the next fund calls the same thing "Watchlist." "Committed" means a signed term sheet at one shop and a wired check at another. Custom fields, list names, the shorthand a team types without thinking — all of it is built up over years, and none of it matches a generic template. So we fit the bot to each client's actual schema: when someone asks "what's in diligence?" or "who's committed this quarter?", it maps those words to that fund's real stages and fields rather than guessing. Skip that step and a bot will still answer confidently — it'll just be wrong, which is worse than no answer at all.
It's the same instinct behind a Slack bot that searches your fund's LinkedIn network — meet people inside the tool they already have open, and answer in their words. That's not a small edge: Slack now connects to more than 2,600 apps, and teams that lean into it report a 47% productivity bump over email-and-meetings workflows.
One practical gate worth naming once: this runs on the Affinity API, so a fund needs to be on Affinity's Scale, Advanced, or Enterprise plan for it to work at all. That's the whole of the fine print.
IVThe loop in practice: the three questions funds actually ask
Three patterns show up in almost every deployment. The exchanges below are composite examples — drawn from the shapes people type across live deployments, anonymized and never any single client's data — but the questions, and the structure of the answers that come back, are exactly what the bot handles day to day.
aQuery 1 — "Who else has co-invested with us in climate in the last 18 months?"
This is the classic four-tab question, and it's one line in Slack. The bot reads across the fund's deals, notes, and relationships, then posts back a tidy list: the co-investors you've shared rounds with in that sector, roughly when, and which of your partners owns each relationship. What used to be a manual crawl through past deals becomes a message you can send from your phone between meetings.
bQuery 2 — checking deal stage and momentum without opening Affinity
"Where are we with [company], and when did it last move?" The answer comes straight back in the channel — current stage, the owner, and the date of the last logged activity — without anyone opening the CRM interface. This is where the per-fund tuning earns its keep: "stage" only means anything if the bot knows your stages. Answered in-channel, it deletes the exact tab-switching cost quantified earlier — the question and the answer now happen in the same window where the deal is already being discussed.
cQuery 3 — writing a note back into the CRM from Slack
Reading is only half of it. After a call, someone can type "log a note on [company]: founder wants to reconnect in Q1, still pre-revenue," and the bot writes a structured update back to Affinity. Because everything runs through the Affinity API, it inherits Affinity's permission model — every action is bounded by the permissions of the person asking, and the write side can be scoped down or switched off entirely for a fund that wants the bot to answer questions and nothing more.
VSetting up the Slack side: what it actually takes
The Slack side is lighter than people expect. Most of the real work is the tuning, not the plumbing. Here's the honest breakdown:
The last row is where the value is. A generic bolt-on that doesn't know your stages will confidently give wrong answers; the mapping step is what makes the answers trustworthy enough to act on without double-checking them first.
VIWhy this pattern is suddenly everywhere
If this feels like it's showing up everywhere at once, that's because it is. The plumbing underneath — the Model Context Protocol, or MCP — was only released by Anthropic in November 2024. By mid-2026, more than 10,000 MCP servers were running in production. The pattern went from research curiosity to table stakes in about eighteen months.
The big platforms piled in. Salesforce rebuilt its own Slackbot into a full AI agent in January 2026, with internal users reporting up to 90 minutes saved a day. And 80% of private equity firms are now exploring or piloting agentic AI, per Grant Thornton's 2026 survey.
Affinity itself validated the whole idea: on May 26, 2026 it launched a hosted MCP server that lets Claude, ChatGPT, Gemini, or Copilot read and write live CRM data in plain English — reframing the CRM, in their words, "from a system of record into a system of action." Big funds are already building bespoke versions: Bessemer connected a local Claude instance to Affinity and Salesforce and stood up an internal system running 70 workflows for 20 partners off one data foundation.
I'm not going to re-litigate our Slack bot versus Affinity's native server here — I wrote a separate piece on wiring Affinity's own MCP server into Claude and Cursor if you want that comparison. The point for this article is narrower: the market caught up to the plain-English-CRM idea, which means the question is no longer whether your fund should have one, but where it should live. My answer is: where your team already is — which, for a CRM Slack integration, means the channel where the deal is already being discussed.
VIIThe honest verdict: what this doesn't fix
Here's the part the vendors skip. A question-and-answer layer only surfaces what's already in the CRM. It does not fix the 30–40% field-completion problem — if anything, it makes the gaps more visible, because now someone's asking questions the empty fields can't answer.
So the honest split is this. If your Affinity is reasonably well-kept and your team asks it a lot of questions, a Slack layer is one of the highest-leverage things you can add — it turns an underused system into something people actually reach for. If your CRM is mostly empty, don't build this yet. A plain-English interface over blank fields just hands you fast, confident nonsense. Fix the hygiene first — and if you're still deciding which CRM is even worth committing to, that's a conversation I've had before in Affinity vs. Attio: the VC CRM that actually sticks.
The tool is worth building for funds with data worth querying. For everyone else, it's a very good reason to finally get the data in order.
VIIIFAQ
Does it work from ChatGPT as well as Slack? Yes. The same layer surfaces as a Custom GPT for people who live in ChatGPT, alongside the Slack version. Most funds default to Slack because that's where the conversation already happens, but the two run off the same connection to Affinity.
Can it write back to Affinity, or only answer questions? Both. It reads across companies, deals, co-investors, notes, and network, and it can write structured updates — a note, a stage change — back into Affinity. The write side is optional and can be scoped down or turned off per fund, so you can start read-only and add writes later.
What if our fund isn't on Affinity's Scale, Advanced, or Enterprise plan? Then it can't run yet. The bot works through the Affinity API, and API access is gated to those plans. That's an Affinity requirement, not ours.
How long does it take to tune it to our fund's own stages and fields? The Slack and Affinity plumbing is usually a same-day job. The tuning — mapping the bot to your actual deal stages, custom fields, and naming so the answers are trustworthy — is the part that matters, and it typically lands in the first week, with a partner checking the results.
— Michael Rouveure
