AI Leadership Weekly · Issue #89 · Tuesday 23 June 2026 · 08:00 BST
Good morning.
Anthropic added interactive artifacts to Claude Code, while Apptio previewed a more conversational way to examine technology spend. Microsoft’s strategy article puts model choice and enterprise control alongside those product developments. For a leadership team, the useful opportunity is to make the result easier to inspect: not just another answer, but something a colleague can question, understand and act on.
IN 60 SECONDS
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.
Apptio previews conversational spend analysis. Apptio announced Conversational Insights and other AI capabilities for technology-spend analysis, with the conversational experience initially in preview.
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.
CEO / COO / CXO CHECKLIST
CEO: Define a completed, acceptable unit of work before reporting productivity.
COO: Include human review and exception handling in the operating view.
CXO: Check who can see the data behind a shared interactive report.
TOP STORIES
1. Claude Code adds interactive artifacts
Anthropic · 18 June 2026
What happened. Anthropic introduced artifacts in Claude Code on 18 June in beta for Team and Enterprise. The feature supports interactive work products derived from a session and connected information, with authenticated organisation sharing. It should not be confused with publishing a report openly on the public web.
Why it matters. Our take: An interactive report can help a colleague inspect the evidence rather than accept a static conclusion. It also creates maintenance questions: which data is current, who owns the logic and what happens when a connected source changes? Treat the artifact as a small product with an audience and lifecycle. A polished interface should not disguise an unmaintained calculation or make restricted information easier to share.
What to do. Choose one repeated status report. Prototype a view that lets the reader inspect the underlying evidence. Label its refresh point, source owner and audience, and test permissions before inviting a wider group.
2. Apptio previews conversational spend analysis
IBM / Apptio · 16 June 2026
What happened. Apptio announced Conversational Insights on 16 June as a preview, alongside cost-management and forecasting updates. The interface is intended to help users ask questions across technology-spend information while inheriting existing permissions. General availability of Conversational Insights was expected later in 2026, not established by the preview announcement.
Why it matters. Our take: A conversational answer is only as useful as the cost definitions beneath it. Ask whether a reported AI cost includes integration, support, review and failed attempts, and whether shared costs have been allocated consistently. For decisions about expansion, connect spend to a meaningful unit such as an accepted case or resolved request. Otherwise a lower model bill can conceal a higher total operating cost.
What to do. Pick one service and list its costs, including estimates where precise figures are unavailable. Make the uncertainty visible. Then compare the cost per acceptable outcome with the current method using the same boundaries.
3. Microsoft sets out its AI approach
Microsoft · 16 June 2026
What happened. In a 16 June strategy article, Microsoft argued for retaining organisational knowledge, using diverse models and connecting governance with cost management. This was a supplier’s position on enterprise AI, not a new product launch or independent evidence that a particular platform guarantees value or avoids lock-in.
Why it matters. Our take: The useful test is whether you can change a component without losing the service. Instructions, source connections, permissions and evaluation often matter as much as the model endpoint. Portability should therefore be demonstrated on a small task. Keep the cost of moving visible, while also recognising that deliberately chosen integration may save operating effort today.
What to do. Select one bounded workflow and identify what would need to move to another provider. Separate reusable business assets from supplier-specific configuration. Estimate the work before describing the service as portable.
SIGNALS FROM THE LAST MONTH
3 June · Gemma 4 gains a 12B model. Google introduced Gemma 4 12B, adding another option for organisations considering local or self-managed inference. Source
2 June · Microsoft connects its agent stack. Microsoft outlined an integrated enterprise agent platform, emphasising the systems needed to run AI alongside existing business operations. Source
28 May · Endava describes Codex adoption. An OpenAI case study described how Endava was using Codex and reusable expertise in its delivery work; outcomes were customer-reported. Source
28 May · Mistral launches Search Toolkit. Mistral introduced components for building search and retrieval, giving developers more control over how agents find relevant information. Source
IN BRIEF
More dated updates from the preceding 30 days.
12 June · GPT-5.2 retires from ChatGPT. OpenAI announced retirement of GPT-5.2 models from ChatGPT. A consumer-product change should not be read as an identical API withdrawal. Source
11 June · IBM and ServiceNow expand collaboration. The companies announced work connecting enterprise data and workflows, with initial new offerings planned for the second half of 2026. Source
9 June · Gemini expands live translation. Google introduced Gemini Live 3.5 Translate, broadening speech-translation capabilities while retaining product-specific availability conditions. Source
8 June · Claude connects with Apple’s framework. Anthropic announced Claude support for Apple’s Foundation Models framework, providing a route beyond the framework’s on-device models. Source
THE 15-MINUTE PLAYBOOK
Minutes 0–4: Define one unit of completed work and who accepts it. Include a rule for work that is reopened or corrected later.
Minutes 4–8: Select four measures: accepted volume, elapsed time, review or rework, and total cost. Add a specific safeguard where the consequences require it.
Minutes 8–12: Mark each measure as observed, estimated or currently missing. Agree consistent boundaries so improvements cannot be created by moving work outside the report.
Minutes 12–15: Name the service owner and the decision the scorecard will support. Start with a simple table and a short explanation of exceptions before investing in a more elaborate interface.
DATA WAVE MOMENT
The useful output of an AI initiative is an operating improvement the business can see and explain. Build the evidence into the service from the start: what arrived, what happened, what was accepted and what it cost. That gives leaders a practical basis for improving, expanding or stopping the work.
QUESTION FOR READERS
Which number would tell us that our AI service is creating more work than it removes?
Brought to you by Data Wave — your AI & Data Team as a Subscription.
