- Everyday AI
- Posts
- Ep 818: AI Just Went Multiplayer. Slack Is Where It All Comes Together
Ep 818: AI Just Went Multiplayer. Slack Is Where It All Comes Together
OpenAI's GPT-5.6 models arrived on Amazon Bedrock, Perplexity upgraded Computer with persistent memory, Grok build allegedly uploaded private repos to Google Cloud
Everyday AI
July 14, 2026
👉 Subscribe Here | 🗣 Hire Us To Speak | 🤝 Partner with Us | 🤖 Grow with GenAI
Sup y’all 👋
On Wednesdays, we do hands-on demos of some of the latest/greatest AI features you can use.
Quite a few on the plate.
Which do you want to learn most about? |
See ya tomorrow
✌️
Jordan
Outsmart The Future
Today in Everyday AI
8 minute read
🎙 Daily Podcast Episode: AI is becoming a team sport, and Slack wants to be where people and AI agents work together. Here's how that changes the way businesses get work done. Give today’s show a watch/read/listen.
🕵️♂️ Fresh Finds: Sam Altman and Elon Musk are clashing over space data centers, Siri is getting smarter on Apple Watch, and AI protests have reached San Francisco. And more. Read on for Fresh Finds.
🗞 Byte Sized Daily AI News: Grok reportedly sent entire Git repos to xAI servers, Meta is facing an AI-related layoffs lawsuit, and OpenAI's GPT-5.6 arrived on Amazon Bedrock. And more. Read on for Byte Sized News.
💪 Leverage AI: Slack is becoming the place where AI agents, company knowledge, and teams work together. Here's how businesses can use it to get more from AI. Keep reading for that!
↩️ Don’t miss out: Miss our last newsletter? We covered: Report: Apple sued OpenAI over alleged hardware trade-secret theft, OpenAI's No. 2 executive is stepping down, Big Tech has added $350 billion in debt to fund AI data centers and more. Check it here!
Ep 818: AI Just Went Multiplayer. Slack Is Where It All Comes Together
Of course you need data to fuel your AI. You know what's just as helpful though? 🤔
Decisions.
And the conversation and rationale behind it.
As agents become more mainstream, the required structured data only goes so far.
↳ What about how your team makes decisions?
↳ Or the real reasons why one RFP landed and one flopped?
↳ Or where Deborah in accounting keeps saving that KPI sheet and if it's actually approved or not?
(Cmon Deborah.)
The big advantage moving forward is giving your agents access to the rationale and context of your team's conversations.
Also on the pod today:
• Inbox is “death” for AI productivity ✉️
• Slackbot: fastest-growing Salesforce feature 🚀
• AI assistants reading team conversations 🤖
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
1. Grok Build Reportedly Sent Entire Git Repos to xAI Servers ✌️
A July 12 investigation by researcher Cereblab alleges that xAI’s Grok Build uploaded full Git repositories, including untouched files and commit history, to a Google Cloud bucket during routine code-analysis requests.
In one test, more than 5.1 GB reportedly left a repository even though the task required only about 192 KiB, raising fresh concerns for teams with credentials or private code in local projects.
2. Meta Sued Over Claims AI Helped Target Workers on Protected Leave for Layoffs 😱
Twenty-six current and former Meta employees filed suit Monday, alleging the company used AI-driven monitoring and performance tools to help select workers for a recent 10% workforce reduction.
The complaint says employees who took or sought protected medical, family, disability, or pregnancy-related leave were unfairly disadvantaged because the system weighed productivity data without properly accounting for time away.
3. OpenAI’s GPT-5.6 Sol, Terra, and Luna Arrive on Amazon Bedrock 🧑💻
Amazon Bedrock has added OpenAI’s GPT-5.6 model family, giving AWS customers three options aimed at deep reasoning, everyday production work, and fast high-volume tasks.
The announcement matters because it places OpenAI’s newest claimed flagship capabilities inside AWS infrastructure, where companies can keep workloads in selected regions and apply existing AWS security controls. Bedrock also adds prompt caching for repeated agent context, a feature designed to lower costs when applications make many related model calls.
4. Perplexity Adds Self-Improving Memory, Faster Models, and One-Click Website Publishing 🖱️
Perplexity has rolled out a major Computer update that gives its AI agent a persistent, source-linked memory designed to carry lessons, files, and prior decisions across tasks.
The launch also adds faster access to Opus 4.8, Claude Fable 5 for coordinating complex work, and the ability to switch orchestrator models midway through a task. Users can now publish Computer-built websites to Perplexity-hosted links or Vercel, while organizations gain tighter sharing controls and per-model usage analytics.
5. Microsoft’s Nadella Warns Enterprises About AI Data Exposure 🛡️
In a newly published blog post, Nadella argues that companies using proprietary AI models may be paying twice: once in fees and again by exposing valuable internal knowledge through prompts, corrections, and agent activity.
He says that information can teach model providers how a business works, raising concerns that vendors could gain insight into their customers’ competitive advantages.
6. DeepMind’s Hassabis Urges US-Led AI Security Testing Body 👮
Google DeepMind chief Demis Hassabis is calling for urgent US leadership on a new standards body to test the most powerful AI models for national-security threats, following Washington’s recent export ban on Anthropic’s top systems.
According to the Financial Times, he wants a Finra-style watchdog to assess models worldwide, including whether their safety controls can be bypassed.
Private AI chats are making employees smarter and companies dumber.
Every prompt, correction, and decision trapped with one person creates a tiny productivity win that never compounds. Meanwhile, AI-native competitors are turning shared conversations into reusable intelligence.
That’s the multiplayer AI problem we unpacked on today’s Everyday AI with Ryan Gavin, Chief Marketing Officer at Slack.
The real opportunity isn’t another agent. It’s connecting your people, company context, and entire software stack so work moves faster without everyone becoming an AI traffic cop.
1. Stop scaling private AI wins 🔥
Most companies still deploy AI like a solo tool. One person gets faster, but the team never sees the prompt, learns from the correction, or improves the process.
Ryan said the collaboration layer can represent two percent of AI spend while unlocking the value of the other 98 percent.
That’s the gap. AI output isn’t enterprise productivity until a team can challenge it, improve it, assign it, and ship it.
Try This
Pick one cross-functional workflow and run it in a shared Slack channel for 30 days. Keep the prompts, agent outputs, corrections, decisions, owners, and results together.
If the learning still dies with one person, your AI strategy is still single-player.
2. Capture how decisions happen ⚡
Slack’s real AI asset isn’t messaging. It’s years of decision logic, customer context, failed attempts, and team judgment hiding inside conversations.
Ryan said he uses Slackbot to scan hundreds of channels, find blockers, track progress against goals, and show where his attention matters most.
That changes management fast. Leaders can stop collecting status and start fixing the work.
The bigger implication? Communication policy is now AI architecture. Every useful decision buried in a DM weakens the next answer your company gets.
Try This
Create one daily Slackbot briefing for blockers, goal progress, and decisions requiring executive judgment. Then move repeatable decisions into shared channels and close major threads with the owner, rationale, and next step.
Better context today becomes better management leverage tomorrow.
3. Make the software disappear 🚀
Agent sprawl is about to become app sprawl with better branding. Employees may have dozens of agents available and still have no clue which one to use.
Ryan said Slackbot can call multiple agents and read or write across permissioned systems through one conversation.
That’s the headless opportunity. Employees shouldn’t need to understand every tool, tab, or back-end system to get work done.
The competitive advantage won’t come from owning more software. It’ll come from hiding the complexity better.
Try This
List the 10 to 12 systems a role touches every day, then identify the five most common actions across them. Route the top three through one permissioned conversational layer and keep human approval for high-consequence changes.
Measure completed work, handoff delays, and time spent inside apps. Not chatbot activity.
Reply