Found Friday is back with a sharper focus: practical AI, Microsoft 365, SharePoint, governance, and the real workflows where business value — and business risk — now live.
The shift this week: Chatbots → Connected agents → Business workflows. The opportunity is not better chat. It is safer, clearer work across documents, approvals, systems, and decisions.
We still like SharePoint. In fact, SharePoint is more important than ever because it is where so much business content, collaboration, permissions, and process already live.
But the conversation is getting bigger. AI is starting to sit on top of those systems, connect into them, and act across them. This reboot is about separating what is useful from what is noisy, and helping business leaders decide where AI can create leverage, where it can create risk, and where they need a plan before they let the tools loose.
What I’m seeing this week
AI tools are becoming agents, not just chatbots.
Those agents want access to business systems, documents, email, calendars, workflows, and approvals.
Governance has to become practical, operational, and understandable before adoption gets away from the business.
This week’s stories all point in the same direction: AI is moving into the work layer of the business. That is where the value is. It is also where the risk lives.
Microsoft 365 + Agent Governance
Microsoft Agent 365, now generally available, expands capabilities and integrations
Source: Microsoft Security Blog
Summary: Microsoft announced that Agent 365 is generally available for commercial customers. It is positioned as a control plane for AI agents, giving organizations a way to observe, govern, and secure agents across Microsoft 365, endpoints, cloud platforms, and third-party ecosystems.
Why it matters
This is Microsoft saying the quiet part out loud: agents are going to spread faster than IT can manually track them. An AI agent is not just another app. It can access data, call tools, chain actions, and make changes.
What to do now: Start building an agent inventory. Ask what AI agents are already being used, what systems they touch, what identities or permissions they rely on, and who owns their behavior.
OpenAI + Deployment
OpenAI launches the OpenAI Deployment Company to help businesses build around intelligence
Source: OpenAI
Summary: OpenAI announced a new deployment effort intended to help organizations identify valuable use cases, redesign workflows, connect AI systems to data and tools, and move toward production deployment.
Why it matters
The bottleneck is no longer just model capability. It is implementation. Companies do not need another executive demo where AI summarizes a PDF. They need help connecting tools to real workflows, security requirements, messy data, adoption habits, and measurable outcomes.
What to do now: Stop asking only “Which model is best?” Start asking “Which workflow are we willing to redesign?”
Value rises when AI has a defined workflow, owner, and success measure.
Risk rises when AI gets broad access before anyone knows what it can touch.
Google Gemini + Proactive Assistants
The Gemini app becomes more agentic, delivering proactive, 24/7 help
Source: Google Blog
Summary: Google announced Gemini app updates pointing toward a more proactive assistant experience, including briefs, task support, file creation, desktop presence, and more cross-app help under user direction.
Why it matters
Google is pushing Gemini toward the “assistant that does things” category, not just the “chatbot that answers things” category. That can reduce friction, but it raises the same hard questions: What can it access? What can it change? What happens when it is wrong?
What to do now: Define practical use cases first. Keep human approval in the loop for anything that touches customers, money, legal obligations, or sensitive data.
Claude + Small Business
Anthropic wants small businesses to use Claude
Source: Axios
Summary: Axios reported that Anthropic is launching a Claude package aimed at small businesses, connecting Claude to tools including QuickBooks, PayPal, HubSpot, Canva, Docusign, Google Workspace, and Microsoft 365.
Why it matters
Small businesses do not need abstract AI strategy. They need help with invoices, follow-ups, proposals, hiring, marketing, customer service, and the work that falls between people. But they also have less room for expensive mistakes.
What to do now: Pick one workflow with a clear owner and low downside risk. Do not connect everything on day one.
Quick readiness pulse
Do we know which AI tools people are already using?
Do we know which systems those tools can reach?
Do we have an approval line before AI touches customers, money, or sensitive data?
Connectors + Microsoft 365 Data
Grok connectors arrive for web, iOS, and Android
Source: xAI
Summary: xAI announced Grok Connectors for web, iOS, and Android, with integrations into tools including SharePoint, Outlook, OneDrive, Google Workspace, Notion, GitHub, and Linear.
Why it matters
When non-Microsoft AI tools start connecting directly to SharePoint, Outlook, and OneDrive, the old assumption that “our data is safe because it is in Microsoft 365” gets weaker.
What to do now: Review third-party AI app consent, connector permissions, external app policies, and SharePoint access governance.
Found Friday signalConnectors turn permissions into strategy. The access model is now part of the AI strategy.
Three questions before you connect another AI tool
1. What can it see? Documents, mailboxes, calendars, SharePoint sites, customer records, or only a safe test space?
2. What can it do? Suggest, draft, summarize, send, approve, update records, or trigger workflows?
3. Who is accountable? Someone needs to own permissions, monitoring, exceptions, and the human approval line.
“The boundary is no longer the platform. The boundary is permissions, connectors, policies, monitoring, and user behavior.”
Closing Thought
The next phase of AI adoption is not about who has the cleverest chatbot.
It is about who can let AI safely interact with the real machinery of the business: documents, decisions, workflows, customers, permissions, approvals, and systems of record.
Before you add another AI tool, ask a boring question: “What exactly are we allowing this thing to touch?”
