Research Period: 8 June 2026 to 15 June 2026
1. ChatGPT just got a lot better at remembering your business
What changed
OpenAI has rolled out its biggest ChatGPT memory upgrade yet, built on a new architecture it calls Dreaming V3. For Plus and Pro subscribers, memory capacity has doubled and the quality of recall has jumped sharply. In OpenAI’s own testing, factual recall rose from 67.9% to 82.8%, how well it sticks to your stated preferences climbed from 55.3% to 71.3%, and accuracy over time improved from 52.2% to 75.1%. The upgrade is automatic, so there is nothing to install, and it is live globally including the UK, Europe, India, and Africa.
Why it matters
If you use ChatGPT regularly, you have almost certainly hit the wall where it forgets your writing style, your business context, or how you like answers laid out. This release targets exactly that. ChatGPT now updates its memory on its own, so you no longer need to say “remember this” every time. There is also a new editable summary view, which means you can see precisely what ChatGPT thinks it knows about you and correct anything that is wrong. For a busy owner who has explained the same context dozens of times, this is a meaningful reduction in repetition.
How to use it
- Open your ChatGPT memory settings and review the editable summary of what it knows about you.
- Refresh or add core business context now that capacity has doubled: your sector, tone, and audience.
- Correct any stale or inaccurate entries so future answers reflect your current situation.
- Test the improved recall by starting a fresh chat and checking it remembers your preferences.
- Decide what you do not want stored, and remove it whilst you have the summary open.
Source: OpenAI ChatGPT Release Notes
2. Gemini 3.5 Flash has landed in your Workspace tools
What changed
You may remember the Gemini 3.5 Flash announcement from Google I/O that we covered last month. It has now landed. Google has recently made Gemini 3.5 Flash the default model for all Gemini Enterprise users, so if you have that subscription you are already on a faster, more capable model with no action needed. Alongside this, Google launched its Managed Agents API in public preview, which lets developers run stateful AI agents inside secure Google-hosted environments.
Why it matters
For members on Google Workspace with a Gemini Enterprise subscription, this is a no-effort upgrade that quietly improves the model behind your everyday tools. The Managed Agents API is more forward-looking and aimed at developers, but it signals where Google’s agentic platform is heading. One important caveat for UK and EU members: Gemini Spark, the personal AI agent feature, remains blocked in the UK and EU pending an AI Act compliance review, with Q3 2026 the earliest likely date. The model upgrade is available now; the personal agent feature is not.
How to use it
- Open your Gemini settings to confirm the 3.5 Flash upgrade is active on your account.
- Continue using Gemini in Workspace as normal, since no migration is required.
- Note that Gemini Spark is not yet available in the UK or EU, and plan accordingly.
- Compare a few typical prompts against your previous results to gauge the improvement.
- Flag the Managed Agents API to any developer on your team exploring agent projects.
Source: Gemini API Changelog and Gemini Enterprise Release Notes
3. Your customers are asking AI about you before they call
What changed
B2B buyers are increasingly turning to AI assistants such as ChatGPT, Claude, and Perplexity for vendor recommendations long before they contact a sales team. A Harvard Business Review analysis lays out how this shifts the buying journey, and includes a pharmaceutical company case study where AI recommendations diverged noticeably from how the company expected to be positioned. The uncomfortable finding is that companies are losing visibility into this discovery process, because traditional marketing and SEO do not translate directly into how AI systems describe and recommend you.
Why it matters
Most guild members sell to other businesses. If your B2B customers are sizing up vendors through AI conversations first, and the AI describes you inaccurately or barely mentions you, you are effectively invisible for a growing slice of the buying journey. This is not a tool to install; it is a strategic shift to start auditing. The good news is that the first step is quick and free of jargon: see what the major AI tools actually say about you today.
How to use it
- Search “[your business type] in [your city/country]” in ChatGPT, Claude, and Perplexity.
- Record what each tool says, noting accuracy, prominence, and any obvious gaps.
- Compare those answers against how you describe yourself on your own website.
- Identify where AI gets your business wrong, and treat that as your starting list.
- Repeat the audit every few months to track whether your AI visibility is improving.
Source: Harvard Business Review
Additional Noteworthy
US government orders Anthropic to suspend Claude Fable 5 and Mythos 5 within 72 hours of launch: In a first, the US government ordered Anthropic to suspend Claude Fable 5 and Mythos 5 within 72 hours of launch, citing export controls. Anthropic has pushed back, and Claude Opus 4.8, the current flagship, is unaffected. It signals that AI availability is now subject to regulatory intervention. | Anthropic
Mollick: “I no longer steer; I commission” (Strategic Awareness): Wharton professor Ethan Mollick argues we have crossed a threshold where the better way to work with capable AI is to commission outcomes rather than direct every step. Write a clear brief, let the AI run, then judge the output, an approach that works today with Claude Opus 4.8 and ChatGPT. | One Useful Thing
UK AI investment surges at London Tech Week: Microsoft pledged £23B for UK AI, the NHS is rolling out Copilot to 505,000 staff, and the London Mayor launched a £12M SME AI support package. London-based members should look at the Mayor’s programme, which offers funded AI readiness assessments and expert mentoring. | BusinessCloud
Cursor Bugbot gets 3x faster: For members who use Cursor for development, Bugbot now reviews code in 90 seconds rather than five minutes, catches 10% more bugs, and costs 22% less per review. | Cursor Changelog
Survey: 64% of Americans fear AI job loss (Strategic Awareness): A large Anthropic survey of 52,000 respondents found job loss is the top AI fear at 64%, whilst 48% hope AI will help cure cancer. Useful context for members managing AI adoption with their teams: name the anxiety rather than ignore it. | Anthropic
Looking Ahead
The week’s governance news and the shift in how we work with AI point in the same direction. Whilst the most capable new models are temporarily unavailable, the commissioning mindset Mollick describes applies right now with the tools you already have. The opportunity this fortnight is less about chasing the newest release and more about getting better at briefing the AI you can use today.