Copilot、Agent Skills 与表格：TLDR AI 本周综述

2 Share Hello and welcome to TL;DR AI . Each week, I sort through the noise to bring you a curated digest of important AI developments and practical advice for business leaders. In this edition:

Copilot gets more “admin friendly,” so rolling it out company wide gets less scary

Gemini for Workspace quietly turns Slides and Sheets into your analyst intern

Anthropic’s Agent Skills make Claude feel less like a chatbot and more like a teammate

A simple AI playbook to triage your inbox without losing your mind

Microsoft Copilot “readiness” packages aim to de‑risk rollouts for IT and leaders Summary: Microsoft is introducing Copilot Readiness Packages in Microsoft 365, giving admins guided assessments and secure deployment presets for rolling Copilot out across Outlook, Teams, and the broader suite. The update also includes better troubleshooting in the Teams admin center, with automated issue detection and Copilot-driven suggestions for meeting and call problems, which should reduce support tickets for hybrid work setups.

Why it matters: This makes it easier to move from Copilot pilots to organization-wide deployment without creating chaos for IT or end users.

Source: Reddit summary of January Microsoft 365 changes (aggregating official docs)

Summary: A new guide on Gemini for Google Workspace highlights how Gemini now sits directly in Slides and Sheets to generate new slide decks, create and clean up images, draft agendas, and build editable charts that stay linked to your spreadsheet data. In Slides, Gemini can generate slide content and visuals from prompts like “annual planning proposal,” while in Sheets it can analyze data and help automate repetitive text and categorization work.

Why it matters: If your team lives in Slides and Sheets, this effectively adds a junior analyst who can set up decks and first pass analyses so humans can focus on decisions.

Anthropic’s “Agent Skills” turn Claude into a reusable playbook for your org Summary: Anthropic has rolled out enterprise “Agent Skills” for Claude, reusable instructions that encode workflows like brand voice, email formats, Jira task creation, or support flows, and made the standard open so skills can be portable across tools. These skills integrate with common workplace apps such as Jira, Confluence, Figma, Canva, Stripe, Notion, and Zapier, with no extra fee beyond normal Claude plans, and early users report faster outcomes on finance, data, and operations tasks.

Why it matters: This lets you codify tribal knowledge once and reuse it across teams, so the AI behaves like a trained employee instead of a clever stranger.

Source: Axios – Anthropic aims to tame workplace AI

Summary: OpenAI’s GPT‑5.2, released for professional and enterprise use, focuses on long context, structured outputs, and multi-step projects like creating spreadsheets, developing presentations, and coordinating complex workflows. The model comes in three variants, tuned for quick responses, deep reasoning, or long form work, and is optimized for tool integration so companies can build more reliable automations for coding, analysis, and document heavy processes.

Why it matters: This shifts AI from “ask a question, get an answer” toward “describe the outcome and let the model orchestrate the work around your existing tools.”

Source: VentureBeat – GPT‑5.2 is here: what enterprises need to know

The AI stack is going vertical, and that changes your tooling map Summary: A recent analysis of the AI landscape argues that OpenAI, Anthropic, Google, and xAI are shifting from “just models” to vertically integrated stacks spanning infrastructure, models, agents, and distribution, with a new Agentic AI Foundation trying to prevent fragmentation. For executives, this means vendor choices are no longer just about model accuracy, but about which ecosystem of agents, connectors, and partner apps you want your operations to live in.

Why it matters: Your AI strategy is increasingly a platform strategy, so consolidation around a few ecosystems will affect everything from procurement to talent and integration choices.

Summary: A late‑December review of 2025 notes that AI systems shifted from static content generators to agents that can use tools, act on your calendar, and coordinate across software to complete tasks without constant prompts. The piece highlights growing expectations that in 2026, agents will increasingly handle logistics like scheduling, travel rebooking, and routine approvals in the background.

Why it matters: As agents get permission to “touch” real systems, leaders need clear policies on what agents can change, how to audit them, and how to keep humans in the loop for high-impact decisions.

Source: The Conversation – AI agents arrived in 2025

Consumer chatbots predict “proactive operators,” not just assistants Summary: A year‑end experiment asking ChatGPT, Gemini, and Claude to predict 2026 found strong consensus that everyday AI will shift from waiting for prompts to proactively taking action, especially in personal logistics and scheduling. Gemini in particular framed 2026 as the year people experience AI as something that quietly updates calendars, reschedules appointments, and coordinates changes with minimal user input.

Why it matters: As employees get used to this behavior at home, they will expect workplace systems to be just as proactive, raising the bar for internal tools and workflows.

Source: TechRadar – What does 2026 have in store for AI?

Summary: A recent outlook on Microsoft 365 trends argues that by 2026 Copilot will shift from being an add-on feature to the default interface for how people use Word, Excel, PowerPoint, Outlook, and Teams. Instead of hunting for features, users will increasingly start with natural language instructions, letting Copilot orchestrate multi-app workflows like drafting proposals, summarizing meetings, or preparing reports.

Why it matters: If Copilot becomes the front door to Microsoft 365, training, change management, and governance move from “nice to have” to core parts of your operating model.

This week: Turn your inbox into a triage line, not a guilt museum

If your inbox feels like a to-do list that other people write for you, this is the week to deputize an AI triage nurse. You can use Microsoft Copilot in Outlook or Gemini in Gmail to sort, summarize, and prioritize instead of reading every email like it is a cliffhanger.

Here is a simple workflow you can pilot with a small group:

Create three labels or categories: “Action today,” “Action this week,” and “FYI only.”

In Outlook with Copilot, select a batch of emails from the last 24 hours and ask: “Summarize these emails, group them by topic, and tell me which require a response today versus this week versus just reading.” In Gmail with Gemini, use a similar prompt on a conversation view or search result set.

For each group, ask the AI to draft one line of next steps and suggested replies, then quickly accept, edit, or delete.

Block a 25-minute “triage block” each morning where you only process AI summaries and suggested replies, not the raw inbox.

Measure success by two metrics at the end of the week: number of unanswered critical emails, and minutes spent in email compared to your usual estimate.

Why it matters: This turns AI from a novelty into a daily operating habit, and gives you a repeatable way to claw back time without missing important messages.

Pro Tip: Give your AI a job title, not just a task

Instead of “rewrite this email,” try “You are my sales operations manager, rewrite this email to a key customer to be concise, clear on next steps, and aligned to our Q1 renewal goals.” One extra line of context often produces outputs that feel closer to how your team would actually communicate.

Why it matters: Framing the role and objective nudges the model toward more on-brand, business-savvy responses, which reduces editing time and makes adoption easier for busy teams.

Thanks for reading today, have a great week!