Most AI advice about fixing this assumes you have Claude or ChatGPT, but most companies actually gave you Copilot. Good news: Copilot's automation layer, Power Automate, can build custom workflows with no separate AI subscription.

It takes about 10 minutes to set up once.

Here's how to automate meeting notes being transcribed to a Teams message and a running OneNote.

Why? Meeting recordings expire after 60 days.

If the only copy of what got decided lives in a Stream recap, it's gone the day that recording ages out.

Step 1: Get the transcript out of Teams before it disappears

If you owned the meeting invite and it was recorded, open the Meeting Event and click "View Recap."

Download the transcript as a .vtt file and save it to a dedicated OneDrive folder. Call it "Transcripts" so the next step can find it.

This only works for meetings you organized; if someone else owned the invite, you won't have Recap access to it this way.

Step 2: Build the trigger in Power Automate

Go to make.powerautomate.com — this is the same automation suite from the companion no-code guide, just pointed at a more specific job.

Click "+Create" → "Skip" → Blank page

Then set the trigger to OneDrive: When a file is created, linked to your Transcripts folder.

Add the following in this order:

OneDrive: Get file content then

Compose (with Inputs set to "Body") so the workflow has the transcript text in hand.

Step 3: Have AI summarize it in your format

Add the trigger Run a prompt, choose Claude or ChatGPT, and give it a fixed instruction so every summary comes back:

"You are an executive assistant creating professional meeting notes. Analyze this meeting transcript: {{transcript}}.

Return plain text only.

Do not use markdown, tables, or numbered lists.

Keep the entire response under 10 bullet points where possible.

Be concise and remove repetitive information. If an owner or due date is not explicitly stated, use 'TBD.'"

That last line stops the summary from inventing an owner or deadline the meeting never actually settled.

Step 4: Send the summary to two places at once

Add OneNote: Get sections in notebook then

OneNote: Update page content, filling in the notebook key, section, and page ID, with the summary appended to the body.

Then add Teams: Post a message to yourself with the same summary as the text body.

Now, a OneNote page becomes your running project log and a Teams message that puts the summary somewhere you'll actually see it that day.

Caution

The summary still needs an edit before it's trusted.

AI-generated notes are a first draft of the record. Read it before you forward it, alwys

It only works for meetings you own.

No ownership, no Recap access, no transcript file to trigger on.

Transcripts disappear on their own schedule, not yours.

Build the OneDrive copy step before the next meeting you'd regret losing, not after you've already lost one.

Most advice about AI is written about Claude. But most companies actually licensed Copilot or Gemini.

So the advice you saw this week is probably not applicable in your 9-5, and you're not getting any better at AI.

Here's the version that is applicable: your job is a stack of admin and busy work, and you can build a real automation for a piece of it with whichever tool your company already gave you.

Today, let's talk about Copilot's Power Automate platform.

What is Copilot's Power Automate?

Instead of manually searching for connectors, you simply type what you want (e.g., "Send an email when a new item is added to SharePoint") and Copilot generates the flow structure for you.

You can create your own flows or use the owns that already exist within the Power Automate platform.

Here's an introduction to Power Automate, using the example of automating getting a Teams notification anytime you are tagged in an Outlook email.

Caution: I'm a blunt person, and I will be honest when I say this: Power Automate is not as easy as it's marketed to be. It will take hours of trial and error. This guide will save you the hours I spent figuring it out.

Identify a repeatable, administrative-heavy task

Choose a task you keep redoing by hand, i.e. the notes you copy from one place to another.

In my case, it'd be checking Outlook religiously to make sure a client hasn't @'d you.

AI workflows only pay off if it replaces something genuinely repetitive. Don't add complexity to your day for the sake of "automation."

What this looks like in a real workday:

• Checking your Outlook when you live in Teams messenger

• Sending meeting recap notes in one internal chat, then copying it into a running notes document

• A status change in a tracker that three other people need to know about

Caution: the biggest waste of a no-code build is automating something that only happens twice a quarter. Save the effort for admin you're doing weekly or more.

Example workflow: Automate getting a Teams notification when you are tagged in an email.

Describe the workflow in plain language

Description-based builders let you describe the outcome in a sentence instead of wiring triggers and actions by hand.

An example is: "When I get an email from this address, send me a message in Teams" is a complete instruction.

You're not required to know what a trigger or a connector technically is.

Descriptions you could ask Copilot to build for your job:

• "When a new email arrives from this client, post it to my project channel in Teams"

• "When someone fills out this intake form, notify the owner and add a row to the tracker"

• "When a ticket's status changes to blocked, message the assigned lead"

Caution: write the instruction as one plain sentence before you open the tool at all. If you can't say it in a sentence, it isn't ready to automate yet.

Example prompt: "When I am tagged in an email, notify me in a Teams chat to myself."

Let the platform suggest the steps

Once you've described it, the platform proposes the workflow in the chat.

This is the part that used to require real technical setup, and mostly doesn't anymore.

Caution: first drafts are often broader or narrower than you meant. This is where you will have to tweak triggers or outputs in the workflow.

Example prompt, What Copilot suggests: "You can automate this by creating a flow with these steps:

1. Use the “When a new email arrives (V3)” trigger from Outlook, and set it to filter for emails where you are tagged (mentioned).

2. Add a condition to check if you are mentioned in the email.

3. If true, use the Teams connector to send a message to yourself in a 1:1 chat.

Would you like help setting up this flow step-by-step?"

Build and Test

Copilot isn't at the point to where it can physically piece the workflow for you together.

You still have to follow the Copilot suggestions to complete the flow setup. Ask Copilot questions on fine-turning the parameters (i.e. what will the text say in my message) and conditions.

Always, always test the workflow before shipping it to a broader Teams channel or email.

While AI speeds up the process, if you're in a disjointed system like Copilot and can't create loops that correct itself, always exercise human judgement.

If you are new to AI, do not waste weeks trying to compare all the tools. This is what each AI tool is actually for in most 9-5 jobs, where it falls short, and the first task worth trying today.

Stick to one platform, ideally the one your IT department has already approved.

If you have more flexibility at work, then pick one platform to master based on its capabilities. After that step, then you can choose the platform for each specific task.

Claude: The sharp writer and designer

The best for nuanced thinking, this is the colleague you hand a messy draft to and get back something clear and well thought through.

Claude is best-in-class for writing materials that sound like a person wrote them.

Claude is also best for UX, brand identity design and coding websites, thanks to its connector to Figma, where you can turn production code into editable designs.

If you're more technically-inclined, you can create custom workflows, dashboards or web apps with Claude Code.

What this looks like in a real workday:

• Paste a 40-page document and ask: "What are the three things in here that could pose as risks later in the project?"

• Hand it your rambling bullet notes and get a status update an executive will actually read (without sounding like "It's not X, it's Y."

• Give it 5 customer complaint emails and ask what pattern connects them

• Upload last quarter's report and say "write this quarter's in the same structure"

• Creating a website mockup for a new landing page

• Ingesting local files on your computer into a dashboard via Claude Code

Where it's weak: no image generation. Fewer integrations into other apps than ChatGPT.

First task to try: paste your longest unread work document and ask for a summary with 3 callouts of risks.

ChatGPT: The Office Swiss Army Knife

The coworker who can do a bit of everything: text, images, voice conversations, web research, quick charts.

While it's not the deepest at any single thing, it has the widest toolbox — especially helpful for image generation. ChatGPT also has a popular coding assistant, Codex.

What this looks like in a real workday:

• Make a quick image for a slide or a mockup of a living room renovation

• Talk to it on your commute: rehearse a salary conversation or a difficult feedback session out loud

• "Research our top 3 competitors' pricing pages and put the differences in a table"

• Drop in a CSV export and ask for a chart of monthly trends

Where it's weak: writing outputs sound like AI and have more shallow strategic thinking. Quality varies more between tasks.

First task to try: use voice mode to prep for your next meeting

Gemini: The one who lives in Google Suite

Gemini lives inside the Google ecosystem: Gmail, Docs, Sheets, and Drive.

Its superpower is that it can already see your work files without you uploading anything.

What this looks like in a real workday:

• In Gmail: "read my emails and tell me what follow-ups I owe to whom"

• In Docs: highlight a section and ask it to tighten the wording — without leaving the doc into another chat

• In Sheets: "which rows shows a decline in [insert metric]?"

• In Drive: "find the deck where we discussed the Q3 pricing change"

• In Meet: automatic meeting notes with action items, saved to Drive

Where it's weak: outside Google's apps, it's a middle-of-the-pack chatbot. The value is the integration.

First task to try: open your longest, most vague email thread and ask Gemini to summarize it and draft the reply.

Copilot: The one that works with Microsoft

Same idea as Gemini, Microsoft edition: built into Outlook, Teams, Word, Excel, and PowerPoint.

This is the tool your IT department is most likely to have approved, which, at a big company, decides the question for you.

What this looks like in a real workday:

• In Outlook: "draft a reply agreeing to the timeline but flagging the budget risk"

• In Teams: join a meeting late and ask "what did I miss?" — it recaps the conversation live

• In Excel: "add a column showing % change and highlight anything below target"

• In PowerPoint: "turn this Word doc into a 10-slide deck" — then fix the design, but the skeleton's done

• In Word: first drafts of proposals pulled from your existing files

Where it's weak: same as Gemini, the standalone chat is unremarkable. And results in Excel/PowerPoint often need cleanup. It is my preferred platform to create LinkedIn infographics.

First task to try: ask Copilot in Teams to recap a meeting you skipped.