The week's theme in one sentence: Whether you are watching UK government AI policy, choosing your next AI model, benchmarking yourself against competitors, or deciding which tools belong in your team workflow — this week moved all four dials simultaneously.
What happened in the last 48 hours
Four stories landed in the last 48 hours, each significant in isolation and more significant together:
- The UK government funded two new AI research labs at Oxford and UCL. £60 million over six years, backed by UKRI, with access to tens of millions of pounds of compute. The explicit goal: redesign the fundamental mathematics of AI to make it cheaper, more reliable, and more accessible to organisations that cannot currently afford it. This is the government betting that open-source, British-built AI can reduce the UK's dependence on a handful of US tech firms — and that smaller businesses will benefit from the cost reduction.
- Google's Gemini 3.5 Pro is inside its June 30 window — and it is still not out. The model was announced at Google I/O on 19 May with a June GA target. As of 24 June, it is available only to a small set of enterprise customers on Vertex AI preview. Prediction markets place the odds of a general release by June 30 at roughly 50 to 55 percent. UK operators choosing between AI platforms are now watching a six-day window.
- 54% of UK businesses are actively using AI — up from 35% in 2025 and 23% in 2023. British Chambers of Commerce data confirms that AI adoption among UK firms has more than doubled in two years. The majority use it for marketing and administration. The minority — 14% of micro businesses versus 36% of large firms — are being left behind not by cost but by a perceived lack of skills and identified use cases.
- ChatGPT Enterprise and EDU can now take actions inside Slack. OpenAI's June update adds Slack connector actions: join channels, create reminders, upload files, update Slack profiles, alongside the existing search. Small businesses running team operations through Slack have a new integration layer to evaluate — ChatGPT operating inside your existing communication tool, not as a separate tab.
What these four signals mean together
Individually, each story is interesting. Together, they describe a single underlying shift: the infrastructure, policy, adoption, and tooling of AI are all moving in the same direction at the same time, and they are moving faster in 2026 than in any previous year.
The £60m lab investment signals that the UK government has concluded AI is infrastructure — not a technology trend — and has committed to building sovereign AI capability. The Gemini 3.5 Pro delay signals that even at the frontier, the competition is intense enough that Google is taking extra weeks rather than risk releasing a model that underperforms its billing. The 54% adoption figure signals that most of your competitors are already using AI. And the ChatGPT Slack integration signals that AI tools are embedding themselves directly into the daily workflows of teams rather than sitting in a separate browser tab.
The three decisions that follow
The SOFAIR Lab at UCL and the BOLD Lab at Oxford are six-year commitments. Their explicit goal is to redesign AI architectures to make advanced tools cheaper and more accessible. For a UK service business, this is a planning signal: the AI tools you cannot currently afford to run at scale are likely to become substantially cheaper over a three-to-six year window. This does not mean wait — the productivity advantage of starting now is compounding — but it does mean your long-term AI platform choices should favour open-weight and open-standard compatibility where possible.
If you are due to make a platform decision involving Google AI tools in the next two weeks, the six-day Gemini 3.5 Pro window matters. Vertex AI enterprise preview customers report significantly improved reasoning, a 2-million-token context window, and Deep Think mode for complex multi-step tasks. If Google releases it to general availability by June 30, the April pricing announcement of £2.50 input / £15 output per million tokens makes it competitive with Claude Sonnet 4.6 for extended document and workflow tasks. If it misses the June window, reassess in July. Do not extend existing vendor contracts this week without noting this pending release.
The British Chambers of Commerce data is not a warning to start — it is a prompt to move from experimentation to consistent deployment. The gap between the 54% using AI and the firms extracting compounding value from it is large. Most businesses at 54% are using AI for one or two tasks, sporadically, without measurement. The businesses that will pull ahead in the next twelve months are those that commit one workflow, measure the result, and build from there. If you are already in the 54%, the question to answer this week is: which workflow is being done consistently with AI, and what is its measured impact?
The ChatGPT Slack integration: what to actually evaluate
The new Slack connector actions in ChatGPT Enterprise and EDU are notable because they shift ChatGPT from a tool you use in a separate context to a tool embedded in your team's live communication layer. For small businesses, the practical test is straightforward: pick one Slack channel and one task you currently do manually — creating a reminder at the end of a meeting thread, uploading a summary document, or updating your team status — and test whether ChatGPT can do it reliably from inside Slack.
The integration requires an admin to enable it under Apps in the ChatGPT dashboard. It is available on Enterprise and EDU plans, not on standard Business or Team plans as of June 24. If your business does not run Enterprise, this is not available yet — but the direction of travel is clear: AI actions are moving into communication tools, not requiring you to move into AI tools.
What to do this week
Four actions for UK operators this week
By today: Note the Gemini 3.5 Pro June 30 deadline. If you are evaluating Google AI tools, set a calendar alert for July 1 to check GA status before making platform commitments.
By Wednesday: If you are in the 54% using AI, name one workflow that will be consistently AI-assisted every week. If you are in the 46%, identify one task that meets this test: highly repetitive, structured inputs, takes 15+ minutes per occurrence. That is your starting point.
By Friday: If you use Slack on ChatGPT Enterprise or EDU, have your admin enable Slack connector actions and run one test. If not, note this integration for your next platform review.
Long-term: The £60m UK AI lab investment is a six-year signal. Factor open-weight and open-standard AI compatibility into your platform choices now, so you can benefit from the cost reductions when they arrive.
The businesses that will look back at June 2026 as a turning point are not the ones who read about these signals — they are the ones who made a decision because of them. Pick one. Act this week.