Why this matters
The data is already in Slack
The patterns the research hears about and sees at funds:
- A
#portfolioor#portfolio-updateschannel where forwarded founder emails, PDFs, and pasted update text land. - One channel per portfolio company (
#acme,#acme-updates), sometimes shared with the founder through Slack Connect. - Deal and partner channels where investors talk informally about how companies are doing: "talked to Sarah, they're closing the bridge next month."
- Monday partner meeting prep assembled by hand from those channels.
A Slack integration is less about adding a new channel of data and more about meeting funds where their data already is.
What if
Three modes: Slack as a source, PostMoney in the channel, and an agent that answers
Every scenario below comes from the research. Chips mark how it ranks the idea.
Inbound
Slack as a source
Watch the channels the fund chooses
A fund picks the channels PostMoney should watch. Any PDF, document, image, or long-form message in those channels becomes a source input to the existing update ingestion pipeline, the same way an inbound email does today. Short chatter is ignored by default.
Provenance is preserved: channel, message permalink, poster, timestamp. The biggest friction in getting updates into PostMoney goes away.
Map #acme to Acme once
The mapping is inferred from channel names against the portfolio, confirmed with a modal, and editable from the PostMoney settings page. From then on everything in #acme is company context.
The company page shows a "From Slack" timeline: the latest update, the latest partner comment, the latest founder message, each linking back to Slack. One click from the record to the conversation.
Partner chatter becomes cited memory
A founder is raising, a key hire left, a customer churned, a partner is worried. A lightweight extraction pass over mapped channels captures these as dated, cited observations attached to the company, with the Slack permalink as the citation.
This should be conservative. The value is a memory of what the partnership actually said, not an attempt to turn every message into a metric.
React to file it
React to any message with a PostMoney emoji to file it against a company. Use 🔥 for a highlight and ⚠️ for a concern.
Reactions are the lowest-friction input Slack has, and the reaction event carries everything needed to fetch the message and act on it. Zero forms, zero context switching.
Founders post in the shared channel
If a fund has Slack Connect channels with founders, a founder's update posted there is captured automatically. The founder is one message away from PostMoney without installing anything.
Removes the "please email us monthly" ask entirely for those companies, and pairs naturally with the founder update submissions work. Whose data it is, and whether the founder knows, still has to be decided.
Outbound
PostMoney in the channel
The update digest card
When an update is ingested and extraction finishes, a card lands in the fund's chosen channel: company and reporting period, three to five headline metrics with deltas from the prior period, flags for runway, revenue trend, and anything needing attention, plus buttons to open in PostMoney, ask a follow-up, or mark reviewed.
The single highest-frequency touchpoint. PostMoney is visible every day without anyone opening the app, and every other outbound feature builds on it.
Alerts the data model already knows
A dedicated alerts channel or thread for conditions that already exist in PostMoney: runway below a threshold, revenue down two consecutive periods, a company silent past its expected cadence, a task overdue or assigned to you, an update that failed extraction or needs human triage.
Nothing new to compute. Existing signals reach the partnership where it will actually see them.
Monday prep as a living canvas
Monday morning, PostMoney posts or updates a canvas: who reported, who did not, what changed, what needs discussion.
Because canvases are editable, partners annotate it in place and the meeting runs off the same document. The prep that used to be assembled by hand arrives assembled.
Run receipts and founder nudges
The batch updates run model already produces a lifecycle receipt; posting it to Slack closes the loop for whoever started the run. Cadence reminders to founders go to the shared Slack Connect channel instead of, or alongside, email.
A team record of what was processed, and reminders where founders respond faster than to email. Nice to have, follows demand.
Conversational
The PostMoney agent
Ask anything about the portfolio
"How is Acme's burn trending?" "Which companies haven't updated since June?" "Compare ARR growth across the fintech companies." "What did Acme say about hiring in their last two updates?"
Answers cite the update and reporting period and link back to the signal or data point in PostMoney. The AI layer, prompts, and search already exist; the agent is a new front door onto them.
Cross-source answers nobody else can give
With Real-Time Search, the agent combines PostMoney's structured data with what has been said in Slack: "Acme's July update reports twelve months of runway. In #acme last Thursday, a partner mentioned they are raising a bridge. Those two things may not line up."
Neither Slack AI nor PostMoney alone can produce that. Real-Time Search supplies the Slack half without ingesting the workspace. The differentiated piece, and the hardest for a generic Slack assistant to copy because it depends on our extraction pipeline.
Reply in the thread
Reply to a digest card with "why did gross margin drop?" and the agent answers in the thread with the relevant excerpt from the update.
The thread becomes the discussion record for that update.
Drafting and the App Home
"Write the LP letter paragraph for Acme." "Draft a reply to the founder asking about the hiring plan." "Summarize the last quarter for the fintech companies." The App Home tab becomes a personal portfolio view: the companies you cover, updates due this week, what changed since you last looked, open tasks.
Drafting leans on existing prompts and produces text the user can copy or send. Useful, but everything else can follow demand.
Automation and light workflow
Small actions that keep people in Slack
Workflow Builder steps
Publish steps such as "create company", "log a note against a company", "request an update from a founder". Fund admins wire them into their own no-code automations without touching PostMoney.
Slash commands
/postmoney acme for a snapshot card, /postmoney log to record a call note, /postmoney due for what is due this week.
Interactive triage
When an update arrives from a sender or channel PostMoney cannot map, a modal asks "which company is this update for?" Ingestion stays clean and nobody leaves Slack.
Task DMs
Tasks assigned in PostMoney arrive as a Slack DM with a button to complete them. Completing in Slack updates the task record.
Where PostMoney is different
Our edge is not summarization
Slack AI already does channel summaries, recaps, and search natively. If the PostMoney agent competes on generic summarization it loses. The edge is structured portfolio data, period-over-period deltas, extraction provenance, and citations. The agent should lean into numbers and sources, and use Slack context to enrich them rather than replace them.
Highest value first
The research's order
1
Update digest cards
Cheap, daily visibility, drives adoption of everything else.
2
Channel watching
Meets funds where their data already is and removes the biggest friction in getting updates into PostMoney.
3
The conversational agent
Especially with Real-Time Search for cross-source answers. The hard-to-copy piece.
4
Alerts and Monday prep
Natural extensions once the first two exist.
Everything else is nice to have and can follow demand.
Before getting excited
Things to know
| Topic | What the research says |
|---|---|
| Platform maturity | The agent surface and Real-Time Search API are generally available but new. Some capabilities are gated to paid Slack plans, and distribution through the Slack Marketplace requires app review. assistant_view shuts off in February 2027, so anything built now should use the agent experience and Agent Sessions API. |
| Permissions and trust | Reading channel messages broadly produces an alarming permission prompt. An explicit opt-in model, where the fund picks which channels PostMoney watches and the app only joins those, lands far better than "read everything." Funds will still want to understand what we store versus what we query live. |
| Slack Connect and founder data | Shared channels with founders raise a data question: whose data is it, and does the founder know PostMoney is listening. A deliberate product and trust decision, probably with visible disclosure in the channel, not something that falls out of the implementation. |
| Tenancy | Slack installs are per workspace; PostMoney accounts are per fund. The model should assume an account can have one Slack workspace and a workspace maps to exactly one account, with the OAuth tokens stored on the account. |
Under the hood
Platform facts, briefly
Terse. The full cross-integration picture lives on Foundations.
Surfaces
- Agent messaging experience:
features.agent_view, Agent Sessions API - Real-Time Search API and the Slack MCP server, permissions inherited from the user
- Events API:
message.channels,message.groups,app_mention,file_shared,reaction_added,member_joined_channel conversations.historyandfiles.info- Block Kit, modals, App Home, canvases, lists, slash commands, Workflow Builder steps
- Slack Connect shared channels
Limits
assistant_viewdeprecated, shuts off February 2027;assistant.*methods replaced by the Agent Sessions API- Some capabilities gated to paid Slack plans
- Marketplace distribution requires app review
Caveats
- Broad message reading produces an alarming permission prompt
- Founder-side disclosure needed in shared channels
- Our edge is not summarization
- One workspace maps to one PostMoney account
Existing systems it builds on: update ingestion and source inputs (email ingestion via Action Mailbox is the closest analog), the extraction pipeline and signals, cadence and expected update schedules, tasks and assignment, batch update runs and receipts, Ai::Client, PromptEngine, and pgvector search.