AI Leadership Weekly · Issue #77 · Tuesday 31 March 2026 · 08:00 BST
Good morning.
Google expanded Search Live, Anthropic introduced an auto-mode preview for Claude Code, and OpenAI updated product discovery in ChatGPT. These changes make interaction smoother in different ways. For leaders, the interesting question is where less friction helps—and where a deliberate pause still has value. Easier conversation, easier comparison and easier execution should not be treated as the same kind of change.
IN 60 SECONDS
Search Live expands internationally. Google expanded Search Live, allowing more users to ask questions through voice and camera input while searching.
Claude Code previews auto mode. Anthropic introduced a research preview that automates some permission decisions in Claude Code while retaining checks for higher-risk actions.
ChatGPT improves product comparison. OpenAI announced shopping updates in ChatGPT, including more visual product discovery and side-by-side comparison.
CEO / COO / CXO CHECKLIST
CEO: List the actions your assistant can take, not just the systems it can read.
COO: Identify which decisions require explicit approval and which have bounded standing permission.
CXO: Test an ambiguous request before celebrating the straightforward demonstration.
TOP STORIES
1. Search Live expands internationally
Google · 26 March 2026
What happened. Google announced Search Live’s expansion on 26 March to more than 200 countries and territories where AI Mode is available. The mobile experience supports conversational voice and camera input, powered by Gemini 3.1 Flash Live. It is a search experience, not a promise that a business’s own service processes have been integrated.
Why it matters. Our take: A customer can show an unfamiliar object or describe a problem without knowing the correct term. That is useful for discovery and support. The next question is whether your product information is good enough to support the conversation: clear model identifiers, usable instructions, current specifications and understandable exception routes. An assistant that recognises a problem still needs an authoritative source for the answer.
What to do. Ask a colleague unfamiliar with a product to explain a problem using only a photo and everyday language. Check whether your published material helps them reach the right answer without insider terminology.
2. Claude Code previews auto mode
Anthropic · 24 March 2026
What happened. Anthropic introduced auto mode for Claude Code on 24 March as a research preview. It uses an additional check to assess proposed tool actions instead of asking the user about every action. The original announcement warns that the system can still allow risky actions or block benign ones and recommends isolated environments.
Why it matters. Our take: The management choice is not simply “human approval” versus “autonomous”. You can define low-risk actions that proceed automatically, constrain where they run and reserve attention for meaningful exceptions. The difficult part is proving that those categories behave as expected. Repeatedly clicking approve is not thoughtful oversight, but an automated judgement is not infallible either.
What to do. Review one approval queue. Remove needless repetition only where the action has clear limits and a recoverable outcome. Keep a sample of decisions for review, including cases the system blocked and cases it allowed.
3. ChatGPT improves product comparison
OpenAI · 24 March 2026
What happened. OpenAI’s 24 March release notes describe richer shopping results, side-by-side comparisons and improved product-data coverage and freshness, rolling out during that week. These are discovery improvements, not a guarantee that every price, specification or stock position is correct.
Why it matters. Our take: For a retailer or manufacturer, product data becomes part of the sales conversation before a customer reaches your site. Treat differences between nearly identical variants as important, not cosmetic. For an internal buyer, a comparison is still evidence to check, rather than permission to order. The purchasing process needs a reliable transition from suggestion to the exact item, supplier and conditions that will be approved.
What to do. Choose five products that are easy to confuse. Check whether their identifying features, restrictions and current descriptions are unambiguous in the source material. Record the owner of any correction.
SIGNALS FROM THE LAST MONTH
16 March · NVIDIA introduces NemoClaw. NVIDIA announced NemoClaw, adding runtime and security tooling around OpenClaw-based personal agents rather than replacing business-level controls. Source
16 March · Mistral releases Small 4. Mistral introduced Small 4 under Apache 2.0, combining capabilities within an open model that organisations could evaluate for their own deployments. Source
10 March · Gemini expands across Workspace. Google announced new AI-assisted creation and analysis features in Docs, Sheets, Slides and Drive, with availability varying by product and plan. Source
9 March · Microsoft previews Copilot Cowork. Microsoft announced Copilot Cowork for longer, multi-step work, initially in limited research preview with broader Frontier access planned. Source
IN BRIEF
More dated updates from the preceding 30 days.
23 March · Mistral expands its voice offer. Mistral announced a Voxtral update, extending its speech-focused model offering beyond text-only interactions. Source
17 March · OpenAI adds GPT-5.4 mini and nano. OpenAI launched smaller models for lower-cost workloads. Access routes and supported features differed between mini, nano, ChatGPT and Codex. Source
17 March · Mistral offers enterprise model training. Mistral introduced Forge, an enterprise offering for building models around organisational data rather than only customising prompts. Source
17 March · Personal Intelligence reaches more users. Google expanded Personal Intelligence for eligible US personal accounts. The announcement did not grant equivalent access to Workspace business accounts. Source
THE 15-MINUTE PLAYBOOK
Minutes 0–4: Choose a service request. Write four stages: identify, explain, propose and act. Put each current system permission beside the relevant stage.
Minutes 4–8: Decide which actions are reversible and which create an external commitment. Define what the assistant may do without another conversation.
Minutes 8–12: Write two ambiguous requests and one deliberately conflicting request. Ask a process owner what the correct next step should be in each case.
Minutes 12–15: Select the point where approval must be recorded, who can give it and how a colleague can see exactly what was authorised. Test this before widening access.
DATA WAVE MOMENT
A better interaction should remove unnecessary effort without removing necessary judgement. The practical work is to join a clear customer conversation to a clear operational process. That means good source information, limited permissions and an obvious handover when the system reaches the edge of its authority.
QUESTION FOR READERS
Which apparently simple request could cause the most damage if an assistant interpreted it too broadly?
Brought to you by Data Wave — your AI & Data Team as a Subscription.
