This was a week of vendors getting a little more honest, whether they meant to or not. One company shipped a tool that shows you your own AI habits back to you. Another quietly admitted a promised feature is not ready and pushed the date. And Microsoft put real money behind the uncomfortable stat everyone in this space has been avoiding: most enterprise AI pilots do not actually work.
1. Google Vids Lets You Star in Your Own AI Videos
What it is: On July 16, 2026, Google rolled out two updates to Google Vids: Gemini Omni and personal avatars. Omni generates and edits high-quality video clips from a plain-language prompt, plus optional reference images, and supports step-by-step chat edits (swap a background, fix lighting, add an effect) without starting over. Personal avatars let you upload a selfie and a short voice recording, then type what you want your avatar to say and have it deliver the message on camera, without ever picking up a camera. Every generated clip carries an invisible SynthID watermark. Avatars are tied to your own Google Account, restricted to your own likeness, limited to users 18 and older, and only available in certain regions so far. Both features are live for Google AI Pro and Ultra subscribers and Google Workspace business customers.
What this means for your workflow: If you have ever skipped recording a quick video update because you did not have time to get camera-ready, this closes that gap. A personal avatar is a reasonable stand-in for a fast team update or a personalized client message when scheduling a real recording is not worth it. The watermarking matters too: if you start using AI avatars for anything client-facing, know that the content is technically identifiable as AI-made, and set expectations with whoever is on the receiving end.
Source: Google – Create, edit and star in videos with two Google Vids updates (Jul 16, 2026)
2. Claude Reflect Shows You Your Own AI Habits
What it is: Anthropic introduced Reflect on July 9, 2026, a built-in Claude dashboard that visualizes how you actually use the tool: the topics you discuss, your overall usage patterns, and the kinds of tasks you turn to Claude for. It goes further than a simple usage log. Reflect occasionally surfaces questions like “What’s one thing you want to keep doing yourself, even if Claude could do it faster?” and offers tools to set quiet hours or schedule nudges to take a break. Anthropic says more sensitive conversations may appear in your insights, but only at a high level, and anything connected to a health integration is excluded entirely; none of the data is used for other purposes. Reflect is in beta for Free, Pro, and Max users who have memory turned on, with a future update planned to add a view of total time spent using Claude.
What this means for your workflow: Worth trying honestly, not just as a novelty. If Reflect shows you are repeatedly re-explaining the same project context, that is a real signal to start using Claude’s Projects feature instead. That said, go in with your eyes open: a dashboard that quantifies how much you rely on a tool is also, by design, a retention feature. It can be a genuinely useful audit of your own habits and a subtle nudge to use Claude more at the same time. Both things are true.
Source: Anthropic – Reflect with Claude (Jul 9, 2026) | TechCrunch – Anthropic’s new Claude feature is quietly selling you on AI
3. ChatGPT for PowerPoint Goes Generally Available
What it is: OpenAI made ChatGPT for PowerPoint generally available for ChatGPT Business workspaces on July 6, 2026 (it is also available globally in beta to Free, Go, Plus, Pro, Enterprise, Edu, Teachers, and K-12 users). It installs as a sidebar add-in directly inside Microsoft PowerPoint from the Microsoft Marketplace. You can turn notes, docs, spreadsheets, or an existing deck into new slides, ask ChatGPT what your presentation’s narrative is missing, tighten a deck for an executive audience, or add a chart slide from a table, all while the output stays fully editable in PowerPoint. Business usage is free through August 6, 2026; after that it draws from the workspace’s token-based credit pool the same way other ChatGPT Work agent runs do, a pricing model that took effect July 6.
What this means for your workflow: This is a direct competitor to Copilot’s deck-building features, running in the exact same app. If your company is on ChatGPT Business, this week through August 6 is a free window to actually test it against whatever you are already using in Copilot or Gemini, before usage-based pricing kicks in. OpenAI’s own beta disclaimer is worth repeating here too: review slide content, formatting, claims, and numbers before you share or rely on anything it produces.
Source: OpenAI – ChatGPT for PowerPoint (GA Jul 6, 2026)
4. Teams’ AI Kill Switch Just Got Pushed Back
What it is: Microsoft has been building an in-meeting toggle that lets licensed Teams organizers and presenters turn Meeting AI (Copilot, Facilitator, and recap) on or off in the middle of a live meeting, so a conversation can be captured when it is useful and shut off the moment it turns sensitive. It was originally scheduled to start rolling out in early July and reach general availability by mid-July. On July 9, 2026, Microsoft updated the official Message Center entry (MC1319216) and pushed the whole timeline back: targeted release now begins mid-August, and general availability is not expected until late August into early September. One technical wrinkle worth knowing regardless of the date: Meeting AI and transcription stay coupled, so turning on Meeting AI automatically turns on transcription, and starting transcription automatically turns on Meeting AI and recap. To keep a meeting free of both, you have to avoid transcription entirely.
What this means for your workflow: If you were counting on flipping AI notetaking off mid-meeting the moment a conversation gets sensitive, that control is not live yet and will not be for at least another month. Until it ships, the only real lever is your existing meeting options, set before the call starts, and simply not starting transcription for meetings you do not want captured.
Source: Microsoft 365 Message Center Archive – MC1319216, In-meeting toggle to turn Meeting AI on or off (updated Jul 9, 2026)
5. Microsoft Bets $2.5 Billion That Your AI Pilot Is Failing
What it is: On July 2, 2026, Microsoft announced Microsoft Frontier Company, a new operating business backed by a $2.5 billion investment and roughly 6,000 engineering and industry specialists whose job is to embed directly inside enterprise customers and build AI systems tied to measurable business outcomes, not just sell software and leave integration to someone else. The announcement leans directly on a well-known MIT Project NANDA finding: 95 percent of enterprise generative AI pilots deliver zero measurable impact on profit and loss. Microsoft’s Commercial Business CEO Judson Althoff framed it as going beyond the “forward-deployed engineer” model other vendors use, calling it the largest, most outcome-driven engineering organization in the industry. Early named customers include the London Stock Exchange Group, Unilever, Land O’Lakes, Novo Nordisk, and Accenture. Microsoft is not alone here: Amazon committed $1 billion to a similar internal deployment unit two days earlier, and OpenAI and Anthropic have both launched comparable joint ventures with outside capital in recent months.
What this means for your workflow: This is Microsoft publicly admitting what a lot of people already suspected: buying Copilot licenses and running a pilot does not automatically produce results, and the company that sold you the software now thinks the fix is putting its own engineers inside your building. If your organization’s AI rollout has stalled, this is not really a story about a new product you can use, it is a signal that you are not alone, and that the entire industry now treats “AI pilot that never becomes anything” as the default outcome to engineer around rather than the exception.
Source: Microsoft – Microsoft Frontier Company: AI engineering that amplifies and protects your intelligence (Jul 2, 2026) | TechCrunch – Microsoft launches its own AI deployment company
The Pattern This Week
Two of these stories are about AI vendors pushing further into tools you already have open all day: Google put a version of you inside Vids, and OpenAI put ChatGPT inside PowerPoint next to Copilot. Two more are about vendors being forced into a moment of honesty. Anthropic built a mirror that shows you exactly how dependent on Claude you have become, whether or not that was the point. Microsoft quietly admitted a meeting-privacy feature was not ready and moved the date instead of shipping something half-built.
The Frontier Company story ties all of it together. Every AI teammate, avatar, and in-app assistant launched this year assumes the underlying AI investment is already paying off. Microsoft’s own $2.5 billion says otherwise, and it says so about its own customers. If your company’s Copilot rollout has quietly stalled, the honest read is that you are the norm, not the exception, and the vendors themselves now know it.
If you want to stay on top of this without reading everything yourself, I cover these moves every week. The newsletter goes deeper than the roundup. Sign up below.