OpenAI is introducing computer-use agents, improved clinical answers and new plugins for ChatGPT for Healthcare, describing capabilities that work across healthcare software. The executive significance is the proposed scope: assistance that extends beyond answering questions into performing work. [Source](https://openai.com/index/more-ways-to-move-care-forward-with-chatgpt-for-healthcare)
Action: Before authorizing a pilot, classify its permitted activities: read, recommend, draft or execute. Name the business owner and define which actions require human approval. Evaluate the announcement as a capability to validate—not evidence of compatibility with your applications or readiness for unattended operation.
AWS describes tenant-specific retrieval instructions for Amazon Bedrock Managed Knowledge Base, with metadata filters to separate tenant data, IAM tags to scope access and changes rolled out one tenant at a time. For a DSO evaluating a shared assistant, this offers an architecture worth testing for practice-level access boundaries. It is not a dental deployment announcement. [Source](https://aws.amazon.com/blogs/machine-learning/customize-answers-per-tenant-in-amazon-bedrock-managed-knowledge-base/)
Action: Start with an internal policy assistant rather than patient-facing or clinical execution. Map access by role and practice, distinguish enterprise policies from local procedures, and test whether users can retrieve information outside their authorization. Make those tests a release requirement.
OpenAI says it is working with NATO and allied cyber agencies to help trusted defenders find vulnerabilities, test fixes and protect essential systems. The relevant enterprise use case is bounded defensive work—not a general delegation of security authority. [Source](https://openai.com/index/strengthening-nato-allied-cyber-defences-ai)
Action: Have the CISO select one authorized vulnerability-validation workflow. Specify the systems in scope, permitted testing, evidence requirements and remediation owner. Judge the pilot on verified findings and validated fixes, with human authorization before production changes.
AWS describes integrating conversation storage, fact extraction and long-term memory through a single AgenticRetrieveStream call. Its example combines document permissions with private memory for each user. For executive oversight, treat persistent memory as a governed data asset rather than merely a convenience feature. [Source](https://aws.amazon.com/blogs/machine-learning/build-memory-aware-ai-agents-with-agentic-retrieve-on-amazon-bedrock-managed-knowledge-base/)
Action: Require a memory policy before enabling persistence. Define what may be retained, for how long, who may access it and how corrections or deletion will work. Test cross-user isolation and permission changes, including what happens when an employee changes roles or leaves.
Amazon Bedrock Managed Knowledge Base now supports tabular search across CSV, XLSX and XLS files, according to AWS. The described capability looks up values, calculates across relevant rows and combines results with related documents. For leadership, the evaluation should focus on whether an answer can be reproduced—not just whether it reads well. [Source](https://aws.amazon.com/blogs/machine-learning/get-answers-from-tables-with-tabular-search-in-amazon-bedrock-managed-knowledge-base/)
Action: Choose one management metric and establish its approved definition, source files, reporting period and reconciliation benchmark. Test calculations, conflicting documents and incomplete inputs. Require the assistant to expose its source basis and flag unresolved discrepancies before using its output in board reporting.
AWS reports that MoEngage built Merlin AI agents to draft, analyze and adjust campaigns through conversation. The implementation exposes existing APIs as MCP tools and maintains tenant isolation while accessing live platform data. The leadership implication is to evaluate tool permissions alongside the workflow itself. [Source](https://aws.amazon.com/blogs/machine-learning/how-moengage-built-agentic-customer-engagement-on-amazon-bedrock-agentcore/)
Action: Separate preparation from execution in the first automation pilot. Let the agent assemble a proposed change; require approval before publication or another external commitment. Review each tool’s privileges, record its calls and establish a recovery procedure. Measure completed, accepted work and exception-handling effort.
AWS’s manufacturing-focused guidance frames multi-agent coordination as a distributed-systems problem rather than a model problem. It offers design patterns, anti-patterns and a framework for deciding when multiple agents are warranted. The executive takeaway: require a reason for every additional agent. [Source](https://aws.amazon.com/blogs/machine-learning/taxonomy-of-multi-agent-design-patterns-anti-patterns-and-lessons-from-manufacturing/)
Action: Ask architecture teams to demonstrate why a simpler workflow or single agent is insufficient. Before approving added coordination, require explicit handoffs, timeout and retry rules, shared-state ownership and an end-to-end audit trail. Keep one business owner accountable for the complete process.
- [OpenAI: More ways to move care forward with ChatGPT for Healthcare](https://openai.com/index/more-ways-to-move-care-forward-with-chatgpt-for-healthcare) - [AWS Machine Learning: Customize answers per tenant in Amazon Bedrock Managed Knowledge Base](https://aws.amazon.com/blogs/machine-learning/customize-answers-per-tenant-in-amazon-bedrock-managed-knowledge-base/) - [OpenAI: Strengthening NATO and allied cyber defenses with AI](https://openai.com/index/strengthening-nato-allied-cyber-defences-ai) - [AWS Machine Learning: Build memory-aware AI agents with Agentic Retrieve on Amazon Bedrock Managed Knowledge Base](https://aws.amazon.com/blogs/machine-learning/build-memory-aware-ai-agents-with-agentic-retrieve-on-amazon-bedrock-managed-knowledge-base/) - [AWS Machine Learning: Get answers from tables with tabular search in Amazon Bedrock Managed Knowledge Base](https://aws.amazon.com/blogs/machine-learning/get-answers-from-tables-with-tabular-search-in-amazon-bedrock-managed-knowledge-base/) - [AWS Machine Learning: How MoEngage built agentic customer engagement on Amazon Bedrock AgentCore](https://aws.amazon.com/blogs/machine-learning/how-moengage-built-agentic-customer-engagement-on-amazon-bedrock-agentcore/) - [AWS Machine Learning: Taxonomy of multi-agent design patterns, anti-patterns, and lessons from manufacturing](https://aws.amazon.com/blogs/machine-learning/taxonomy-of-multi-agent-design-patterns-anti-patterns-and-lessons-from-manufacturing/)
Know an executive who should read it first? Forward this.
— BWP
Tell me what to research next.
Two questions: which topics matter most to you, and what challenges you're trying to resolve right now — including doctor or hygienist turnover. Your answers shape upcoming issues.
Take the surveyWas today's edition worth your five minutes? Your vote shapes what lands in your inbox next.
Know an executive who should read it first? Send it their way.
