|
All-Access · Members
All three operator guides live in your library, and every new one lands there the day it ships. Reply to this email with what you want the next Deep Dive or office hours session to cover and I'll prep it.
|
|
|
|
|
In AI this week
|
Slack just put AI agents in the group chat
The Thursday 3 took three weeks off while I was heads down in my companies. The news didn't, so today's three are the stories from those weeks that still matter, and the biggest one broke this morning. Slack launched Slack Code: tag a coding agent (Anthropic's Claude Code, Cognition's Devin, GitHub Copilot) into a conversation and it spins up a dedicated channel for the project, where your team watches the work happen live: code diffs, a planning doc, a working preview. Teammates comment and approve before anything ships, and the channel keeps an audit log and archives itself when the work is done. Slack says over 70% of these channels go from idea to merged pull request inside a single day. Live today on every plan, including free workspaces. (Slack)
What it means for you: If you have engineers, have them turn this on today. It costs nothing, and it inherits your existing Slack permissions, so IT has nothing new to configure or audit. If nobody on your team writes code, don't skip this one; study the shape instead. The agent works in the open, the team reviews as it happens, and nothing ships without a human approving. That's the template the rest of your company's agent work is going to follow, and you now have a free place to practice it.
|
|
|
xAI's new agents bring their own computer to work
On August 11, xAI launched Grok Bot, in beta: always-on agents that each get a persistent cloud computer (a browser, a terminal, files) and sign into your team's tools to carry a job all the way to done. xAI names the gap most AI leaves open: "There is a huge difference between 90% done and 100% done. Most AI gets you almost there. Grok Bot can finish the swing, because the work lands where a human would put it, in the actual tool." You can even put several Bots in one group chat, where they pass work between themselves and pull you in only for judgment calls. Individuals and teams get it now on the top Grok and Cursor plans; enterprise access is rolling out. (xAI)
|
Krista Letz
@kristaletz
|
𝕏 |
It’s been really fun leveraging @bot for GTM! Have spent a few weeks iterating on some of my workflows, so I wanted to share how I use Grok Bot at SpaceXAI. |
What it means for you: Put this next to the Slack story and the pattern is hard to miss: the agent is no longer a chat window, it's a coworker with hands. You don't need Grok specifically. What you need is the shape: one scoped job, access to the tool where the job happens, and a human on the judgment calls. It's the same shape as the email command center I run on my own inbox: the agent does the work, and nothing final happens without me. Pick the one job your team waits on most and sketch the agent version of it, under rules you write.
|
|
|
Cisco closed 145,000 support cases with zero humans involved
On yesterday's earnings call, Cisco CEO Chuck Robbins told Wall Street that 145,000 customer support cases were "resolved entirely by AI with zero human intervention" in fiscal 2026. The same call reported $9.3 billion in AI infrastructure orders for the year, about 4.5 times last year's total, and the company's highest revenue, operating margin, and earnings per employee in 30 years. That last metric is the one to sit with: the agents took on the support load, and the per-person numbers got better, not worse. (Cisco Q4 call transcript)
What it means for you: This is what done looks like when it's real: audited numbers a CEO signs in front of investors, not a pilot press release. And the quiet part is that none of it came from buying an agent. It came from redesigning support around one, run by people who know how to work with it. A study out Monday found the other side of that coin: two-thirds of companies report AI pushing costs up, not down, because they bolt it onto workflows built for humans (TTEC). The difference is mostly practice, and practice is exactly what this week's workflow installs.
|
|
|
|
This week's workflow
|
|
AI office hours: the 30-minute ritual that makes AI stick
|
|
|
WORKFLOW / THE +1
AI office hours: the 30-minute ritual that makes AI stick
> OUTCOME
I've paid for the slick AI training session. Everyone nodded, the glow lasted about 48 hours, and nothing changed by Monday. This ritual is what worked instead: a standing 30-minute session where one person brings a real task they're stuck on and your best AI person solves it live, in front of everyone, false starts included. The prompt that worked goes into a shared library. A few weeks in, the team's default question shifts from "how do we always do this?" to "could AI do this?", and new hires inherit the library on day one.
> SETUP
A recurring 30-minute slot, a respected peer to run it (not someone from IT; trust is the active ingredient), and a Slack channel where the working prompts live. Nothing to install.
> WHAT YOU SHOULD DO NEXT
Put the slot on next week's calendar and name your champion today. Then watch the first real task get solved live.
|
| |
↳ Full walkthrough _
|
|
|
Andrew
|
|
1:1 · OPERATOR TO OPERATOR
|
$1,000 · limited spots
|
Want to work through any of this directly?
I do a small number of Operator to Operator sessions each week. 60 minutes, screen-share, you bring the bottleneck and we leave with a concrete plan you can run that week. $1,000, pay then book.
|
|
|