Intro paragraph – why demoing works
Helping people get the most out of their Copilot for M365 sub creates stickier customers – demonstrating the tools in a relevant way helps with this.
What is Copilot Analyst?
Microsoft 365 Copilot Analyst is an AI agent, released in March 2025 alongside the Researcher agent. It’s one of many agents published by Microsoft, and specialises in turning raw data into actionable insights. It thinks like a skilled data scientist, taking buckets of raw data and extracting key insights, spotting trends and changes, and A THIRD THING.
Copilot Analyst uses chain-of-thought reasoning to tackle the problem iteratively – it looks at a problem, taking multiple steps to produce a detailed, accurate, and well-reasoned reply. It tackles problems in a very human way, mirroring a human approach to an analytical challenge – if the agent gets stuck, or takes the wrong path to approaching an issue, it can un-stick itself without additional prompting.
Learn more about Chain of Thought Reasoning
Like data scientists, Copilot Analyst uses Python (the most popular programming language in 2025, according to Pluralsight) to tackle complex data queries. This allows the agent to work with different types of data: structured, unstructured, or semi-structured, in a variety of different formats (including Excel, CSV, JSON, XML, and PowerPoint files).
Some quote about data making good decisions
A clever person (1900)
Copilot Analyst uses these approaches – a scientific approach based on chain-of-thought reasoning, employing Python to tackle data queries – to turn data into actionable insights, empowering users to make better informed decisions.
So how do I demo it?
INTRO PARAGRAPH
Step 1: Find some relevant data
I prefer to demo using real-world data – taking somebody else’s dataset and being able to run a live demo with it proves a lack of smoke-and-mirrors. If users need to manually manipulate their datasets before analysing them with Copilot Analyst, this detracts from the benefits of using AI to analyse data.
1:1 Audience
Some ideas…
1:Many Audience
In a recent tour of the Baltics, I was tasked with demoing Copilot Studio to partners from Latvia, Estonia and Lithuania
Step 2: Identify relevant, relatable pain points and use-cases
Emotion, link to the audience – what challenges do they have
In a sales approach, you should start with this. When scripting demos, practicalities dictate that it comes later on…
Step 3: Build out your script
Demoing AI solutions introduces a different challenge compared to demonstrating other technology solutions. Praying to the demo gods is an important first step for any demo (“EXAMPLE THING THAT COULD GO WRONG 1” or “EXAMPLE NUMBER 2”), but when you introduce some artificial thinking into the mix, things become even less predictable.
When scripting your demo, think about the core features and capabilities that address the pain points and use-cases that you identified in step 2.
It’s a tough balancing act – over-scripting can leave you unprepared for when AI chooses to throw a curveball or gives you an unexpected response. RISK OF OVER PREPARING.
I like to go into demos with a single sheet of paper, covering the different features and capabilities that I want to show, reminding me of any key prompts, concepts, and ideas – I sit this on the desk in front of me, or on a separate screen for any remote demos, and use it to keep me on track and to avoid missing any key areas.
Key points for scripting your demo:
- Kick off early with a “DEFAULT INSIGHTS PROMPT” from your selected data file(s) – do this before giving too much of a description. The initial analysis can take a few minutes, which you can fill with showing the recipient through the dataset, where you sourced it from, or making smalltalk
- Rehearsing your demo is not an alternative to reading the responses from the Analyst agent whilst delivering your live demo – because Copilot Analyst is re-analysing the data every time you start a new session,
- For longer prompts, use the Windows 10/11 Clipboard History to have these lined up – typing anything on a live demo feels harder than climbing Everest (though I’ve only done one of those things…). Just make sure to clear any sensitive information out of the history first!
Step 4: Close out with a relatable human-decision scenario
Looping back to why a company or individual might want to use Copilot Analyst – turn data into actionable insights, empowering users to make better informed decisions – wrapping up a demo with a relatable, hypothetical scenario is a great way to highlight the potential impact of Copilot Analyst can have on your customer’s business.
You should link this back to the pain points and use-cases that you identified in step 2.
Leave a Reply