Coinbase CEO, Brian Armstrong: Some great insights on how they are using internally hosted AI Agents. "It’s connected to every Slack message, every Google Doc, and every Salesforce data confluence. Now, this is all linked up and the data is all aggregated, so you can ask these agents questions.... Every team is using it—legal, finance, everything. It’s like the "Oracle of Coinbase." I’ve started to ask it things that go beyond just simple prompting, like "Hey, can you write this kind of memo for me?" I’m asking these AI agents now, as CEO, "What should I be aware of in the company that I might not be aware of?" It will tell me, "Did you know that there’s actually disagreement on this team about the strategy?" I realized I didn't know that, but the AI does because it can read every Slack message and every Google Doc. Tobi, who is on my board, calls this "reverse prompting." Instead of telling the AI agent what you want to do, you ask it what you should be thinking more about." --- From The All-In Podcast YT channel (link in comment)show more

Imagine if employees knew one of their Slack messages were being read and summarized by an AI for the CEO? This could be the end of Slack.

We built this at my 40-person startup. Claude Code agents connected to Slack, Linear, GitHub, calendar, meeting transcripts. Every morning I get a briefing surfacing team misalignment, blocked work, initiative drift. Any founder can and should run this playbook today!

Surveillance Lurking Across Channels Konstantly

So the agent is spying on employees lol. These guys are so clueless man lol

Doesn’t sound very secure for a crypto platform

The part nobody talks about is the context window cost when you pipe in Slack plus Docs plus CRM simultaneously.

He talks about AI like a boomer

Thought police, who exhibited doubts about my decision making? Grab everything about him including personal, family, extended family etc...tech has a way of being used for other power control applications. Look out!

The ceo of any organization handling inter-team disagreements on strategy sounds like an incredibly inefficient use of his time

That level of integration is crucial for AI agents to be truly effective. The next step will be seeing how they handle data privacy and access controls to prevent sensitive information from being exposed.

What people aren't realizing with "reverse prompting" is you're only going to get information as good as the LLM was trained to give. And if a malicious LLM gets into your workflow... We're only at the beginning of the singularity.

Big brother

Built an all knowing Oracle of company knowledge and the first thing the CEO asked was what he was missing. Correct instinct.

crazy dystopian

My favorite "reverse prompting" I ask @meetgranola after a meeting, "What was the key truth that was left unsaid in this meeting." Shockingly useful insights.

Sounds familiar - every big company said the same thing about their data warehouse. The real question is whether employees will trust AI answers enough to act on them.

Did the ceo just replace himself before anyone else? Lmfao “Tell me what I don’t know about my company” “What’s the solution?” “Did you just take my job?”

This is why Jensen Huang highlighted the importance of Confidential Compute, so enterprises can apply AI agents at this scale/level of access. Ideally, this should be the best way to work with agents but then again, it still hosts lots of data privacy/security risks.

The ceo being able to just freely grab data from slack is wild. Not the flex he thinks.

So basically the company finally built the coworker who actually knows where everything lives. Pretty wild.

this is next level organizational transparency

In a traditional hierarchy, "disagreement on strategy" is often filtered out before it reaches the CEO. An agent that reads every Slack message identifies these patterns instantly, providing a ground-truth map of the organization's health.

Coinbase has a real edge here -- their agent is not a document retriever, it is already inside the workflow that moves money. every slack message indexed is also a decision that moved capital. that context is not available from a public model. the gap is not the AI. it is the data that only exists inside the company and the permissions to read it.

You should be aware they can do your job better than you :)

This is disgusting. Welcome to the age of distrust

This is the enterprise playbook in action. We hit the same inflection point at 15 agents — the tech works, but governance becomes the bottleneck. What we learned: schema-validated handoffs between agents caught 73% of errors before they cascaded. The companies winning here aren't just deploying agents, they're building eval suites with 200+ test cases per agent before prod. Coinbase's approach (Slack + Docs + Salesforce unified) is exactly right — context aggregation is the real moat, not the models themselves.

Reverse prompting concept is interesting but the privacy side of this is worth thinking about. Employees having candid conversations in Slack probably aren’t expecting the CEO to be able to ask an AI what they’re saying about company strategy.

This is the ultimate way to cut through the corporate BS. Most CEOs are kept in the dark by layers of middle management, but these agents give you the real ground truth. Huge win for transparency and speed.

does exactly that with a simple integration in a secure manner …

The reverse prompting angle is fascinating - asking "what should I be aware of" rather than "what should I do." But the real test is whether the AI surfaces disagreements that leadership actually wants to hear.

Giving agents access to every Slack message and GitHub commit is a massive context advantage. The question is precision — you can flood the agent with data or give it structured retrieval. What's the Coinbase approach: raw access or curated knowledge base?