Issue 011: Teams, Productivity and the AI Workplace

Most of the AI productivity gains in 2026 are still individual gains. Someone drafts faster, someone summarises faster, someone produces a cleaner first cut. Useful, often impressive, sometimes life-changing for the person holding the keyboard. Transformational at the team level, often not. Two of this week's stories make the same point from opposite ends. Cloudflare and General Motors are both reorganising work, not just speeding it up. The companies extracting real gains are doing the harder thing, changing how the team coordinates around AI, not just bolting AI onto unchanged rituals.

The pattern in unchanged rituals is easy to spot. The same meetings, the same review cycles, the same status updates, all running on AI-flavoured admin underneath. The pre-meeting briefing that used to be a 30-minute prep that nobody did is now a five-minute AI brief that everyone reads. Good. But the meeting itself is still scheduled for 45 minutes. The post-meeting summary used to be a 20-minute manual write-up that quietly got skipped. Now it is automatic. Good. But the meeting cadence has not dropped from weekly to fortnightly to take the saving out. The cross-team status update used to be a Slack thread that everyone skim-read. Now it is a structured weekly digest with owner-tagged actions. Good. But nobody owns whether the digest actually replaces a meeting.

First, a shared prompt library that is kept current, not a private stash on someone's desktop. Second, rituals like stand-ups, retros, and reviews redesigned so AI inputs arrive before the meeting, not as a novelty in the room. Third, named owners for the team's AI tooling, with enough authority to change the workflow when the tools allow it. These three signals are visible from the outside. If you cannot see them in your team, the AI gain is sitting on individuals, not on the team.

The strongest counter-argument is honest. AI without ownership manufactures noise. Bad summaries, vague action lists, and unowned automations make work feel busier, not lighter. That is exactly why redesigned rituals need named owners, review points, and a willingness to undo the redesign when it stops earning its keep. None of this is set-and-forget.

The recommendation is simple. Pick one ritual this week. Redesign it around an AI input that arrives before the meeting starts. Then take the saving out of the calendar, not just off the task list. If your stand-up still runs to 25 minutes after the AI pre-brief, the AI has not changed the work yet.

Granola is a meeting AI tool that takes notes alongside you while you type, then enhances those notes after the call with structured summaries, action items, decisions, and key quotes. The defining design choice is that Granola does not join meetings as a visible bot. It captures system audio directly from your laptop, which means no bot avatar in the call, no recording announcement, and no awkward moment when you explain why "Otter" is sitting in your one-to-one. For client-facing professionals, that single difference is why many are switching.

You join meetings as normal in Zoom, Google Meet, or Microsoft Teams. Granola transcribes in the background using Deepgram and AssemblyAI, then enhances your sparse notes using your choice of large language model on paid plans, including Anthropic's Claude (currently Sonnet 4.6 Thinking), Google's Gemini, and OpenAI's ChatGPT family. The output is genuinely good for stand-ups, client check-ins, recruiting calls, and weekly reviews. Granola is SOC 2 Type 2 certified and UK General Data Protection Regulation (GDPR) compliant. The Business plan also includes Model Context Protocol (MCP) support, so your meeting notes can be read directly inside Claude or ChatGPT without copy-paste.

The free tier caps meeting history aggressively, so you will outgrow it within a fortnight of real use. Battery hit on the laptop is real during long calls. And because there is no visible bot, you have to remember to start the recording yourself, which is an easy thing to forget after lunch.

Try Granola →

Some links are affiliate. Zymbos AI may earn a small commission at no extra cost to you.

The companies cutting hardest this week, Cloudflare and General Motors, are also the ones being most honest about why. Both are running the same playbook: keep revenue growing, change the role mix underneath, and let the calendar absorb the saving.

The next twelve months will separate organisations that adopt AI by adding it to existing work from organisations that adopt AI by redesigning the work itself. The first group will report modest productivity gains and rising headcount frustration. The second will report uncomfortable structural changes and material margin improvement.

Watch the FTSE 100. My prediction is that by Q2 2027, at least three constituents will publish an explicit AI-driven workforce restructuring plan, framed as a productivity gain rather than a cut. The signal to watch for is when "AI-first operating model" appears in an annual report. The question worth asking inside your own team this week: which meeting on next week's calendar would not survive an honest audit? If you redesign one ritual after reading this, hit reply and tell me which one. I read everything.