Hello AI Enthusiast,
This week's stories all point in the same direction: building AI turned out to be the easy part, getting people to use it is the real challenge. Sam Altman admitted adoption will take longer than expected, Anthropic launched free AI courses with Claude Academy, and OpenAI shipped a plugin to help admins manage AI at work. Let's dive in.
The Big Picture 🔊
Sam Altman Admits AI Adoption Will Take Longer Than Expected
Speaking on David Senra's podcast, Sam Altman admitted the industry has been too ambitious on AI adoption timelines. People stick to familiar tools, and changing behavior is harder than Silicon Valley expected. He compared it to people still renting from Blockbuster in Netflix's early days. The admission comes amid public backlash against data centers and fresh cybersecurity concerns, with OpenAI slowing development of its Astra model to strengthen safety testing.
|
||
|
There are two ways to read this. Either Altman genuinely never considered that people need time to adopt new tools, or the narrative has to change now that OpenAI needs to justify its numbers. And when he says behavior change is harder than "the tech nerds realize," we have to point out: that's him. What we can confirm from our own work is the pattern he's describing. Companies treat AI as a technology purchase, then discover nobody is using it. Adoption is a change process that takes months, and it starts with people understanding what these tools are actually for. |
Anthropic Launches Claude Academy to Teach the World AI
Anthropic shared its approach to AI education through Claude Academy, a free learning platform at academy.claude.com. The curriculum mirrors how Anthropic trains its own employees: organized around real problems rather than product features, built on durable mindsets like "verify in proportion to the stakes," and full of hands-on practice. It covers Claude but also model-agnostic AI skills, with plans for more personalized learning as models improve.
|
||
|
Credit where it's due, this is genuinely well designed. It's problem-based, it works on mindsets instead of feature lists, and OpenAI offers something very similar in its own Academy. But we know the numbers on self-serve video courses: completion rates sit around 3%, and that's among people who chose to enroll. Now picture the colleague who never wanted AI in the first place. For Anthropic, training is customer success, not a business, so it will never get the resources that real behavior change requires. That gap is exactly where structured, live training comes in. We'll go deeper on this in an upcoming blog post which you can look out for here. |
Free courses are everywhere. Behavior change isn't.
Even Sam Altman now admits that adoption is the hard part, and no video library fixes that. Our Corporate AI Training is built around your company's tools, roles, and real work, with live sessions people actually complete. If AI adoption is on your September agenda, this is the moment to plan it.
OpenAI Gives Admins Their Own AI Plugin
OpenAI released an Admin plugin for ChatGPT Work and Codex that lets workspace admins check usage, manage members and permissions, and adjust spending limits from a single conversation. It can also automate recurring workflows, like routing approval requests to Slack or Teams, while respecting existing roles and permissions. OpenAI's own IT team reports that AI workflows now resolve about 45% of its support tickets.
|
||
|
This fits the theme of the week. Rolling out AI at scale is not just giving everyone a chat box, there's real administrative work behind making these tools something a business can trust. This plugin is a quiet admission of that, and we prefer it to pretending the problem doesn't exist. From our client work the pattern is clear: setting these tools up is quick, setting them up well is not. Missing connectors, access limits, spend questions, the issues show up after launch. |
Bits and Bobs 🗞️
DeepSeek released Harness, a developer preview built around the idea that everything is a plugin.
Claude's memory now works everywhere, across both Claude chat and Cowork, and users decide what goes in it.
OpenAI is cutting API pricing for GPT-5.6 Sol by over 20% for the next three months.
OpenAI now offers Zero Data Retention for its frontier models.
Mistral announced a partnership with Saudi AI company HUMAIN.
Google is giving publishers a new way to fight AI-driven traffic losses.
Google launched new AI study tools for students across Search and Gemini.
ChatGPT can now send texts for you through a new Apple Messages plugin.
Apple announced changes for apps in the European Union.
Anthropic is funding research into better evaluations of AI's impact on wellbeing.
Anthropic is bringing Claude Mythos 5's cybersecurity capabilities to more defenders through Claude Security.
From Our Founder’s Channels 🤳
In his latest LinkedIn post, Gianluca Mauro shares his take on Claude Academy: why it's great, why it's destined to fail, and what it actually takes to boost AI adoption. Read it here.
That's a wrap on our newsletter!
If you're looking to build AI skills in your team, check out our Corporate Training. Every program is tailored to your team and built around hands-on practice.
Catch you next week! 👋
