Capital Wealth
Your Money · The Workplace File

Your Boss Wants the AI to Read Your Slack Messages. All of Them.

CEOs are ordering staff out of private DMs and into public channels — because the AI agents need to see everything to be useful. One CEO tracks what percentage of your messages are public. His CMO scores 98%. Welcome to the era of the transparent employee.

By Sean Anees Saifi · Capital Wealth · Published Tuesday, August 25, 2026 · Source: The Wall Street Journal, August 24 and August 25, 2026 editions
Key Points
98%
public-message score of Zapier’s CMO
~80%
the target public share for leadership roles
100M
emails and messages in the Spirit Airlines archive Google bid on
$10M
Google’s bid for that archive, delayed by a union
The modern workplace channel: executives increasingly want conversations public by default, where AI agents — and everyone else — can read them.
The modern workplace channel: executives increasingly want conversations public by default, where AI agents — and everyone else — can read them.
In one line: Every message you send at work is now potential training data. Write accordingly — and notice which companies get more valuable when conversations become data.

Somewhere in your company’s Slack right now, someone is typing a private message that begins “don’t share this, but…” — and somewhere above them, an executive is drafting a policy to make sure that message never happens again. Not to stop the gossip. To feed the machines.

The Journal reports that CEOs are pushing employees out of private DMs and into public channels because AI agents — the software co-workers now crawling company workspaces — are only as good as what they can read. “A big part of getting these agents to work really, really well is to make sure they have as much context as you can feed it,” says Wade Foster, CEO of Zapier. His instruction to staff: default to public. Zapier gets data from Slack on what percentage of each person’s messages are public. Foster recently weighed in at 79%. His chief marketing officer hit 98%. The leadership target is about 80%. Yes: your transparency now has a leaderboard.

The astonishing part, and the unsettling part

The astonishing part is what the machines do with the access. On a live customer call, Foster asked an AI agent to build a slide presentation from all Slack messages about that customer over two months. It finished during the call. At ElevenLabs — 800 employees across 50 countries, all internal communication on Slack — an agent digests the top 100 public channels into a podcast employees listen to on the commute. Nobody was “in the room” because the room is now searchable. Meetings are recorded, transcribed, and posted; many messages, the Journal notes, are now written primarily to be read by bots.

The unsettling part arrived in the same story. Google bid $10 million for the digital archives of the defunct Spirit Airlines — roughly 100 million emails, personnel records and Teams messages — to help train its AI models. The purchase was delayed only because the flight attendants’ union objected. Think about that pipeline: your Tuesday complaint about the expense-report system could outlive your company, get sold in a bankruptcy, and end up teaching a model how humans complain. A Berkeley professor made the human point plainly: private messages are where employees brainstorm freely. Take away the private space and you may get more context for the agents and fewer honest first drafts from the humans.

The two takeaways worth keeping

For you as an employee, the rule was always true and is now enforced by architecture: nothing you type at work is private. Not the DM, not the “deleted” draft, not the meeting you forgot was being transcribed. Write every work message as if it will be read by your CEO, a lawyer, and a language model — because the odds that all three eventually read it are no longer negligible.

For you as an investor, notice what quietly became valuable: the pipes and the archives. Salesforce (CRM), which owns Slack, told the Journal that channel usage is growing and called the accumulated record “the fuel that powers the innovation.” When a company’s internal chatter becomes training data, the company holding the chatter holds an asset that never shows up on the balance sheet. We flagged the same pattern when chips became loan collateral two weeks ago. The AI era keeps doing this — finding balance-sheet value in things nobody thought to price. Your DMs made the list. Congratulations, sort of.

What It Means For Your Portfolio

A workplace-conduct memo and a data-moat signal

No position change. The investing signal — communication archives becoming AI training assets — strengthens the data-platform theme we already hold through the technology sleeve.

Companies that own large proprietary conversation and workflow archives (Salesforce/Slack among them) gain a quiet asset as agent training becomes the norm. We’re watching rather than chasing — the theme is early and the privacy backlash (see the Spirit Airlines union objection) is a real regulatory tail risk.

Book a 15-Minute Review → Back to Edition No. 159 →