Your AI roadmap starts here
AI Pathfinder
layer 1 layer 2 layer 3 layer 4 layer 5 abstract shapes

AI That Works On Demand: Episode 3

Episode 3

18 Jun 2026

8 min read

Knowing which tools to use and how to write a good prompt is a solid start. But the real question is: what does this look like in practice, on a normal working day? 

In this third session, Fraser Dear put Copilot to work live across four applications your team already uses. No slides, no theory. Just real prompts, real outputs, and a clear picture of where the time savings are. 

Watch the recording and you’ll learn: 

  • How to get on top of long email threads, pull out actions, and draft replies without reading every message.
  • How to turn rough notes into a polished document in the time it takes to write a prompt.
  • How to clean up a messy spreadsheet, build charts, and answer data questions without touching a formula.
  • How to catch up on meetings you missed and know exactly what’s been assigned to you.  

Episode 3 – Small AI Wins: Everyday Hints and Tips Hidden in Your Work Tools

Key Takeaways: AI That Works Episode 3 

Episode 3 focused entirely on practice. Fraser ran four live demos, each using real prompts in real M365 applications, to show the kind of small, repeatable changes that add up to significant time savings across a working week. 

Think Small to Win Big 

The session opened with a straightforward bit of maths. 

Save two minutes on a task, five times a day, and you’ve saved nearly 40 hours over a year. That’s not a transformation project. It’s just doing the same familiar tasks a bit faster, a bit more reliably, and with less effort. 

All four demos used M365 Copilot, running entirely within your Microsoft tenant. Your data doesn’t leave your organisation. That’s one of the core reasons BCN recommends it as the right starting point. 

1. Outlook: Stop Drowning in Email Threads 

Long, complex email chains are one of the most common drains on a working day. Reading back through them properly, especially after time away, takes longer than anyone has time for. 

Fraser showed how Copilot can cut through this instantly. Using the CRAFT prompting framework covered in  

Episode Two, a single structured prompt pulled the following from a 30-message email chain: 

  1. A three-sentence summary of everything discussed 
  2. Every key decision made 
  3. All outstanding actions with named owners 
  4. Anything still waiting on a response 

From there, Fraser showed how to draft a reply directly, asking Copilot to produce two versions side by side: one warm and friendly, one short and professional. The whole demo took four minutes. 

A question from the session:

Is the experience the same in Outlook Desktop as in the web version? 

Yes. The functionality is identical. The sidebar looks slightly different, but everything works the same way, as long as you have an  M365 Copilot licence. 

Expert View

If I can save 2 minutes a task, five times a day, that’s almost 40 hours a year. That’s a huge saving.

Fraser Dear, Head of AI and Data Innovation

2. Word: From Scattered Notes to Finished Document

Creating a polished document from rough notes is one of those tasks that always takes longer than expected. You know what you want to say. Getting it structured and written is the slow part. 

Fraser started with five files in OneDrive: keynote notes, session notes, a developer platform write-up, a sources list, and a genuine stream-of-consciousness brain dump. He attached them all inside Copilot in Word and used one structured prompt, specifying: 

  1. Context: Just returned from Microsoft Build 2026 with files of rough notes. 
  2. Role: Act as a senior technology analyst writing crisp executive briefings. 
  3. Action: Write a first draft covering the big picture, key announcements, availability, and what this means for our Copilot practice. 
  4. Format: Title, sections, tables, and headers. 
  5. Tone: Confident, concise, and commercially focused.

The document came back formatted, structured, with tables and headers, before he’d finished reading the prompt aloud. 

He then showed how to go further: asking Copilot to rewrite the draft in your own voice. You can describe your style in the prompt directly, or attach a document you’ve already written as a reference. Copilot works through the text in real time, adjusting language while keeping the substance intact. 

3. Excel: Clean the Data, Build the Charts, Answer the Questions

The Excel demo started with a deliberately messy CSV file. Inconsistent region names, mixed-case entries, dates in multiple formats, unit prices with random currency symbols, and totals that simply didn’t add up. The kind of file that lands in your inbox and takes an afternoon to sort. 

Step one: make it a table. 

Before using Copilot in Excel, convert your data to a proper Excel table. It’s a two-click change that doesn’t alter the data at all. It simply tells Copilot where your data lives. Without it, Copilot can’t work effectively. 

From there, one prompt produced: 

  1. Cleaned and standardised region and status columns 
  2. Corrected price formatting throughout 
  3. A calculated column showing expected totals and flagging discrepancies 
  4. Duplicate rows identified and removed 
  5. Colour-coded highlights on rows with issues 
  6. A plain-English action list explaining every problem and how to fix it 

Fraser then showed chart creation from scratch, asking Copilot to select the right chart type, build out sales by region and a monthly trend view, and reformat the axis labels when they came back incorrectly. No formulas. No chart wizard. Just a prompt. 

Finally, he asked three analyst-level questions directly over the data: 

  1. Which sales rep generated the most revenue, and how much? 
  2. What percentage of orders were cancelled? 
  3. If we focus next quarter on our single best-selling product by value, which should it be? 

Copilot answered all three instantly, pulling directly from the table data. 

4. Teams: Never Miss What Matters in a Meeting

The Teams demo tackled a scenario most people deal with regularly: missing a meeting and needing to get up to speed fast, without watching back the whole recording. 

Fraser opened a meeting recap he hadn’t attended and asked Copilot for a quick, friendly catch-up: what the meeting was about, what was decided, and what he’d missed. Copilot returned a three-sentence overview and a full action table with owners and timings. 

He then went further, asking Copilot to surface every moment he’d been mentioned by name, what was said, and exactly what was now expected of him. The response came back with all of it, including a light-touch note that he’d been assigned rather a lot of actions in his absence. 

The final demo showed how to use that catch-up proactively. Fraser asked Copilot to summarise ideas raised in the meeting and add one new suggestion of its own. He then drafted a reply referencing that idea, so his colleagues could see he’d read back, engaged with what was discussed, and was contributing something new, even without having been in the room. 

Three Things to Take Away

1. Pick one task this week.

Not a project. Not a workflow review. One thing you do regularly that takes longer than it should. Try it with Copilot. That’s how confidence builds.

2. Use CRAFT in the early days.

Context, Role, Action, Format, Tone. If you need a reminder, go back to Episode Two. Once it becomes habit, you won’t need the structure. You’ll just write better prompts. 

3. Share what works.

If a prompt saves you time, share it with a colleague. The fastest way to build Copilot confidence across a team is for people to show each other what’s possible. 

Next Steps 

Episode Four is on Thursday 13 August at midday. It will cover the difference between Power Automate, declarative workflows, process automation, and agentic workflows. Fraser will also build a live agent inside M365 Copilot during the session, so you can see how quick and straightforward it is to get started. 

Discover what’s possible with BCN’s AI services.

AI and Automation down down down