AI Leadership Weekly · Issue #92 · Tuesday 14 July 2026 · 08:00 BST
Good morning.
OpenAI released GPT-5.6, Google made AlphaEvolve available on Cloud, and three additional FireSat satellites extended a real-world sensing programme. It is a varied week, with a useful reminder that AI’s business value need not arrive as a conversation. Optimisation, detection and assistance solve different problems. This edition looks beyond model announcements to the work each development could make more effective.
IN 60 SECONDS
OpenAI launches GPT-5.6. OpenAI released GPT-5.6 in multiple variants, giving customers different capability and cost options across its product and developer routes.
AlphaEvolve becomes available on Cloud. Google announced general availability of AlphaEvolve on Google Cloud, bringing its algorithm-optimisation approach to customer workloads.
FireSat adds three satellites. Google reported the launch of three additional FireSat satellites, extending work on AI-assisted detection of wildfires.
CEO / COO / CXO CHECKLIST
CEO: Look for an expensive decision or calculation, not only a writing task.
COO: Compare the result against a completed job with a known outcome.
CXO: Check access conditions and operating costs before starting the trial.
TOP STORIES
1. OpenAI launches GPT-5.6
OpenAI · 9 July 2026
What happened. OpenAI launched GPT-5.6 on 9 July across ChatGPT, Codex and its API, with access varying by plan. The family includes Sol, its flagship; Terra, a balanced tier; and Luna, a more cost-focused option. OpenAI also introduced an “ultra” setting that coordinates multiple agents for demanding work.
Why it matters. Treat the extra choice as a purchasing question, not an instruction to upgrade everything. A complex investigation and a routine classification task do not necessarily justify the same model or computing effort.
The decision should include the cost of failed attempts and human corrections. A cheaper answer is only cheaper if it is useful.
What to do. Select a routine task, a difficult task and an exception case. Compare your existing setup with the new options. Keep whichever produces the most dependable value for each category.
2. AlphaEvolve becomes available on Cloud
Google · 9 July 2026
What happened. Google made AlphaEvolve generally available to Google Cloud customers on 9 July. Give it a baseline algorithm and an objective, and it searches for improved versions, returning readable code. Google names BASF, JetBrains and Kinaxis among early users.
Why it matters. This is a different proposition from writing another document. My takeaway is to look for value in how a business schedules, routes, allocates or calculates—not just how it explains its decisions afterwards.
It is not an automatic operations director. The opportunity depends on expressing the problem, constraints and success measures clearly enough to test candidate solutions.
What to do. Ask your operations or analytics lead for one recurring optimisation problem. Establish the current baseline and practical constraints before evaluating whether this kind of tool fits.
3. FireSat adds three satellites
Google · 7 July 2026
What happened. Three FireSat satellites launched on 7 July, expanding the Earth Fire Alliance-led programme supported by Google Research and Muon Space. Google says the earlier pilot detected small, low-intensity fires missed by existing satellites. The new launch extends the network; it does not establish complete global coverage.
Why it matters. This is a reminder to consider AI that improves what you can observe, not only what you can generate. Earlier, more useful information could matter to organisations managing physical assets and disruption.
The business value would depend on coverage, reliability, access and an actual response—not the alert alone.
What to do. Identify one decision your organisation makes with information that arrives too late. Ask what earlier warning would change, and who would be able to act on it.
SIGNALS FROM THE LAST MONTH
23 June · Mistral releases OCR 4. Mistral introduced OCR 4, adding richer structural information and paragraph-level locations to extracted document content. Source
23 June · Azure adds an Observability Agent. Microsoft announced general availability of Azure’s Observability Agent, applying agents to the investigation and operation of cloud services. Source
18 June · Claude Code adds interactive artifacts. Anthropic introduced artifacts in Claude Code in beta, allowing users to create and share interactive outputs from their development work. Source
16 June · Microsoft sets out its AI approach. Microsoft published its perspective on model diversity, enterprise control and implementation; it was a strategy article, not a new model release. Source
IN BRIEF
More dated updates from the preceding 30 days.
6 July · ChatGPT comes to PowerPoint. OpenAI made ChatGPT for PowerPoint generally available, offering in-application assistance with editable presentations, subject to account limits and workspace settings. Source
2 July · Microsoft funds hands-on implementation. Microsoft announced a $2.5 billion initiative involving 6,000 industry and engineering experts to work alongside customers on AI implementation. Source
30 June · Sonnet 5 arrives. Anthropic launched Sonnet 5 with introductory API pricing of $2 per million input tokens and $10 per million output tokens. Source
24 June · Gemini 3.5 Flash adds computer use. Google announced computer-use capabilities for Gemini 3.5 Flash, extending how the model can interact with software interfaces. Source
THE 15-MINUTE PLAYBOOK
Do not start by asking everyone to brainstorm AI use cases. Start with one department’s working day.
Minutes 0–4: Name a repeated decision: how to allocate work, which items to prioritise, how to schedule capacity, or when to investigate an exception.
Minutes 4–8: Write down the inputs, constraints and consequences. Identify what is genuinely measurable and what still needs judgement. Avoid reducing a multi-dimensional decision to one convenient number.
Minutes 8–12: Ask why the current decision is difficult. Is information late? Are there too many combinations? Are people repeatedly applying rules? Or is the disagreement about objectives?
Minutes 12–15: Choose the smallest useful next step. That might be a data sample, a baseline calculation or a discussion with the people making the decision. It does not have to be a purchase.
A tool can help solve a defined problem. It cannot choose your business priorities for you.
DATA WAVE MOMENT
When assessing AI opportunities, separate three questions: can we do the work with less effort, can we make the decision better, and can we offer something we could not sensibly offer before?
Each question needs a different business case. Keeping them separate makes it easier to spot the worthwhile opportunities without pretending every experiment is a transformation.
QUESTION FOR READERS
Which decision gets made so often that even a small improvement would be worth pursuing?
Brought to you by Data Wave — your AI & Data Team as a Subscription.
