AI headlines for business this week (Aug 28, 2026)
McKinsey says 80% of workers feel more productive with AI, but only 37% see EBIT impact. Plus agent scaling, build-vs-buy shifts, and what to watch this week.
Another week, another pile of AI press releases. Here is the short list that actually matters if you run a business and you are trying to separate signal from spend.
Written for business members, not engineers. If one of these hits your industry, the deep dive below unpacks the McKinsey ROI gap in plain language.
1. McKinsey State of AI 2026: productivity up, EBIT flat
McKinsey published its annual State of AI survey on August 25 (QuantumBlack, 1,719 leaders across 97 countries, fielded May through June 2026). The headline that should worry your CFO: 80% of respondents say AI improved their individual productivity, but only 37% report any enterprise-wide EBIT impact from AI. That 37% figure is essentially unchanged from last year, despite another year of budget increases and wider agent deployments.
Why it matters for business: your team probably feels faster. Your P&L probably does not show it yet. That gap is not a reason to stop. It is a reason to stop treating AI like a software purchase and start treating it like a workflow redesign project with an owner and a metric.
2. Only 6% are "high performers" and they redesign workflows
McKinsey identifies about 6% of organizations as AI high performers (those attributing 5% or more of EBIT to AI). What separates them is not model choice. They are more than three times as likely to intend fundamental business transformation, they deploy across multiple functions, and senior leadership stays engaged. Bolt-on pilots do not compound. Redesigned workflows do.
Why it matters for business: if your AI strategy is "let each department experiment," you are optimizing for demos, not earnings.
3. Agentic coding is flipping build vs buy
Nearly 32% of McKinsey respondents said their organization decided not to buy one or more software products because agentic coding tools made an in-house build feasible. At the same time, 40% of companies with $1B+ revenue report scaling AI agents in at least one function, up from 27% last year.
Why it matters for business: mid-market operators face the same fork. Buy a vertical SaaS tool, or wire an agent to your existing stack? The answer depends on whether you have someone who can own the integration and the rollback plan. "We built it in a weekend" is not a strategy unless you can maintain it.
4. AI costs are starting to bite
About 20% of respondents said AI operating costs have constrained their AI use. Token bills, inference at scale, and surprise API overages are moving from engineering trivia to line items finance reviews. Agent FinOps is becoming a real discipline: track cost per workflow, not just cost per seat.
Why it matters for business: budget for variable AI spend the same way you budget for cloud. Flat-rate SaaS assumptions will break once agents run all day on live data.
5. Headcount expectations are rising faster than cuts
McKinsey reports 39% of respondents expect headcount declines from AI in the next year, versus 14% who say they already saw declines in the past year. Expectations doubled. Realized cuts did not. Plan for productivity and workflow change before you plan layoffs on a slide deck.
What I would do this week
- Pick one workflow with a measurable outcome ( cycle time, error rate, revenue per rep ).
- Ask whether you redesigned the process or just added a chat box.
- Read the deep dive on why individual gains are not hitting your bottom line yet.
Next post is the long read on the McKinsey ROI gap and what high performers do differently. If you want blog updates when they drop, join the notify list.
Matt Potter · Swift Media