24–31 AUGUST 2026 · MARKETING · SALES · CUSTOMER SERVICE
Everyone saved an hour. Finance is still looking for it.
Plus: agents that need permission to fail, campaigns that all sound equally excellent, and the hidden kill switch in your AI stack.
Hello, fellow laundry inspectors.
This week’s AI developments contained an uncomfortable message for marketing, sales and customer-service leaders: better AI does not automatically produce better operations.
Productivity gains need somewhere valuable to go. Agents need permission to stop. Creative teams need rubrics that reward more than polished sameness. Multi-model products need actual supplier resilience. And credible-looking research still needs credible researchers.
Let’s empty the pockets.
Productivity rose. The P&L stayed put.
McKinsey’s latest State of AI survey contains two numbers that belong next to each other.
That financial-impact figure was essentially unchanged from last year. Meanwhile, only around 6% qualified as AI high performers—also essentially unchanged.
AI adoption is growing. Enterprise-wide scaling reportedly rose from 38% to 44%. At large organisations, the share scaling AI agents increased from 27% to 40%.
The tools are spreading. The value is not spreading at the same speed.
The missing conversion
Imagine an account executive uses AI to prepare for a meeting in 15 minutes instead of 45. Thirty minutes have been saved. Where did they go?
- Prepare for more opportunities?
- Improve the meeting or move an opportunity forward?
- Correct the AI’s research?
- Attend another internal meeting?
- Simply finish work earlier?
All are possible. Only some create measurable organisational value.
AI makes campaign variants faster. Did that create more experiments—or 40 versions nobody had time to review?
AI writes notes and follow-ups. Did it improve CRM completeness and next steps—or add another summary nobody reads?
AI drafts responses. Did it reduce repeat contact—or move difficult work into a less visible queue?
Time saved is inventory. It becomes value only when the operating system decides what happens to it.
The workflow redesign gap
McKinsey found that organisations reporting stronger results were much more likely to redesign workflows rather than place AI inside the existing process. The survey cannot prove causality, but the distinction is useful.
Take the existing task, ask AI to do part of it, then keep every meeting, review and approval around it.
If this part now takes five minutes, what should the entire process become?
Do not stop at “How many hours did we save?” Ask: What changed because those hours were saved?
Source: McKinsey’s State of AI 2026 ↗
THE REST OF THIS WASH
Five more stories are waiting in the drum.
Continue with the agent that needed permission to fail, the campaign rubric that missed originality, the hidden kill switch in your model stack, and two more commercially useful oddities.
Subscribe free. One thoughtful briefing each week. No service pitch.