AI keeps changing. Who’s helping your team keep up?
If you’ve finally figured out which AI tool you like, only to open it and discover that the features you’ve been using have once again changed, I understand the frustration.
As someone who is at the forefront of this change, you are going to see many versions, iterations, and evolutions of these tools before society hits a norm. Keeping up and keeping your workflows functioning can feel like a full time job in addition to the actual work you need to do right now.
Two changes I highlight below bring that feeling into focus. OpenAI announced a transition from custom GPTs to plugins, which means organizations using those GPTs need to consider what happens to the tools they’ve built and the work those tools support. Meanwhile, Anthropic has begun bringing Claude chat and Cowork together (+ Claude Design!), changing how users move from asking a question to handing over a task.
These changes reach beyond where a button lives. A staff member who learned one way to prepare a report may open the same tool and find a different experience. A custom GPT built around your program information may need to move, be tested, and be explained to everyone who relies on it. The training you provided can still have value while the instructions that came with it need updating.
This is why I see AI training as an ongoing responsibility. A single workshop can give people a foundation, but they also need opportunities to ask questions as the tools evolve, practice with the work they actually do, and understand when a change affects their work.
That doesn’t mean scheduling a full training every time a vendor announces a feature. A short walkthrough, an updated how-to, or time in a staff meeting to demonstrate one relevant change can go a long way. Someone does, however, need to decide which changes matter to your organization and make sure that guidance reaches the people affected.
Your internal AI tools and processes need a named owner, with time and authority to maintain them. That person should know what the team uses, keep instructions and reference materials current, coordinate testing when something changes, and make sure there is a backup who can step in.
We can’t promise our teams that the tools will stay the same, but we can give them clear ownership, continuing support, and room to learn, so keeping up doesn’t become one more thing they’re expected to figure out alone.
With gratitude, love, and wishes for a sweet new year (Shana Tova!),
Amanda
NEWS THAT MATTERS, AND WHY
Image editing gets more practical
OpenAI has introduced ChatGPT Images 2.5, with improvements to targeted edits, consistency across revisions, and generation speed. New features include drawing a rough sketch to guide an image and placing comments directly on an image to explain what should change.
Why it matters:
For a small nonprofit marketing team, the useful test is whether you can revise a campaign image without having to rebuild everything around it. Try a specific change on a low-stakes asset, then check the wording, brand details, and anything the tool was supposed to preserve. Better editing still needs an attentive reviewer.
Read OpenAI’s announcement
When “recording” doesn’t look like recording
Apple has announced conversation-memory features for Apple Watch, with Live Rewind and Siri Recap expected in beta later this year. Live Rewind can display the previous 15 seconds of conversation as text, while Siri Recap creates a high-level summary.
Apple says these features do not create or store audio recordings and delete raw audio after processing. Live Rewind also provides an audible chime and visual indicators when activated.
Why it matters:
Your organization’s guidance may need to reach beyond meeting recordings and note-taking bots. Wearables introduce questions about notice, consent, saved summaries, and use during sensitive conversations, even when there is no audio file to replay.
Start by reviewing whether your policy covers transcription and summarization across devices.
Read Apple’s announcement
Documents, slides, and designs move into the conversation
Claude Docs and Claude Slides have launched in beta on paid plans, and Claude Design now works inside conversations. Enterprise administrators control when these capabilities are enabled.
Why it matters:
A program proposal and its board presentation could develop from the same conversation, reducing the need to explain the project again.
Read Anthropic’s announcement
Chat and Cowork are coming together
Anthropic began rolling out its combined Claude experience to Pro and Max users on September 16, with other plans following on different schedules. The change lets a conversation move from a question into more substantial work without the user choosing a separate mode first.
Why it matters:
Easier access to file-based work is a good reason to revisit staff guidance about connected information and when to review an action.
Read about the rollout
FREE RESOURCES
Prepare your custom GPTs for the move to plugins
The OpenAI notice supplied for this issue says custom GPTs will retire on December 11, 2026, with migration beginning September 22 as the option becomes available to creators. It also gives October 26 as the end date for creating new custom GPTs.
If your organization relies on custom GPTs, start by listing the ones people actually use, who maintains them, and which instructions, reference materials, and connections each depends on. Prioritize the workflows that support recurring work, then test their replacements with representative tasks.
Why it matters:
A successful migration means staff can still do the work they depend on. Keep a record of what each GPT is supposed to accomplish so you can evaluate its replacement against that purpose.
OpenAI’s custom GPT retirement and migration FAQ
OpenAI’s migration guide
Register for OpenAI’s migration webinar