The most expensive thing in your business never shows up on the balance sheet. It is the potential of your people. It drains away every time they spend an hour on work a machine could do. This week's AI news was busy. New models landed. Agents kept spreading. The rules shifted. But all of it comes back to one question. Are your best people free to do their best work, or buried in busywork? Here is the week, in plain language, with our take on each story.
The model race did not slow down this week. OpenAI's new GPT-5.6 family is live, but only about twenty vetted partners can use it so far. Google's Gemini 3.5 Pro is still stuck in a limited enterprise preview with no firm launch date. Anthropic's Fable 5 is now back on sale worldwide, after the US lifted its export controls at the end of June. The headline is not the raw power. It is access. Who can use these models, and at what price, is what really shapes your options. For most businesses, the newest model is not the thing that matters. What matters is whether the work around it has been redesigned to use it well.
Our take: A better model is not a plan. Value comes from redesigning the work around it, not from owning the newest tool.
Read the original at LLM-Stats
AI agents are the story of the year, and the gap between hype and results is widening. Almost every executive says their company rolled out agents in the past year. Yet only about a third have one running in real production. Gartner now warns that more than four in ten agent projects could be scrapped by 2027, on rising costs and unclear returns. Where agents do work, the median payback lands around five months. The businesses getting value are the ones who fixed the work first, then handed a clean, well-designed task to the machine.
Our take: Agents pay back when they own a redesigned task, not when they are bolted onto a broken one.
Read the original at Accelirate
The money is moving from building models to helping people use them. Microsoft is putting 2.5 billion dollars into a new team of about six thousand people who will sit inside customer businesses and help AI actually stick. Anthropic added tighter admin controls for its enterprise users, with richer analytics, model-level permissions, and spend alerts. Even the companies that build the tools now admit the hard part is not the technology. It is changing how people work.
Our take: If the makers of the tools say adoption is the hard part, that is where your effort should go too.
Read the original at Build Fast with AI
The jobs picture got sharper this week. PwC's 2026 Global AI Jobs Barometer found that AI is rewarding human skills like judgement, creativity, and leadership. Roles where AI takes over the routine work are growing faster and paying more. At the same time, Bloomberg reports that tech and finance are shedding around twenty-eight thousand jobs a month. Some routine work is going away, and the value of human judgement is going up. The lesson is not fear. It is focus.
Our take: The safest role is the one built around human judgement. AI should clear the routine so your people can get there.
Read the original at PwC
On 29 June the Council of the EU signed off a package that simplifies the AI Act and pushes back its toughest deadlines. The rules for high-risk systems now move to late 2027. Smaller and mid-sized firms get lighter paperwork, softer fines, and access to testing sandboxes. More time can feel like permission to wait. It is not. It is a window. The businesses that use it to redesign their work will be ready when the rules do bite.
Our take: More time is not a reason to stall. Use the window to redesign the work now, while the pressure is off.
Read the original at Latham & Watkins
New models. Real agents. Big budgets. Shifting rules. None of it matters if your best people still spend their days on work a machine could do. That is the potential draining away, quietly, every week. The good news is that it is fixable. We embed a Fractional Chief AI Officer in your team. An AI leader who finds the wasted hours, redesigns the work, and frees your people to do their best work.
.svg)
.png)
.svg){kind=small}
.png)