Data and AI Training

Home | Prices | Contact Us | Courses: Power BI - Excel - Python - SQL - Generative AI - Visualising Data - Analysing Data

How to get the best from Microsoft 365 Copilot and other AI tools

This is a talk for Midlands Insights Placement Scheme regional development day on 11th June 2025.

Generative AI tools like ChatGPT and Microsoft’s Copilot can boost productivity. This session will demo various AI tools, focusing on Copilot in Teams, Word, PowerPoint, and Excel. Since finance teams use Excel, we’ll explore how generative AI can assist with Excel tasks, such as data reconciliation and chart creation.

The key to maximizing these tools is effective communication with AI, known as prompt engineering. The session will offer tips on how to do this well.

Warning

AI can make mistakes. Always check important results. Follow your trust policy on AI. And never put anything confidential into an AI tool unless your trust policy allows it.

What are (generative) AI tools?

We can have a conversation with a Generative AI tool and it will (hopefully) offer good advice and information. The core of a generative AI tool is a large language model (LLM).

Most of the major software companies offer an AI tool. Examples of such tools are ChatGPT from OpenAI, Gemini from Google, Claude from Anthropic and Copilot from Microsoft. These are all available through a browser and have a free version.

For more information:

What is M365?

Microsoft 365 (M365) is a suite of office applications provided by Microsoft. These are ubiquitous especially in organisations. The most well known applications are Word, Excel, PowerPoint, Outlook and Teams.

What is Copilot?

Copilot is an AI assistant that is integrated into these M365 applications to help make users more productive.

It use natural language to help create, summarise, analyse, and automate tasks
It is built on LLMs from Microsoft and OpenAI.
It is context-aware: it can read your emails, calendar, files, and chats (within your organization)

For example, Copilot can:

in Teams: provide a transcript and a summary of a meeting
in Outlook: review a draft email and improve the clarity and tone of the message before it is sent
in Word: draft a full report based on a few bullet points of essential information
in Excel: summarise data in pivot tables, draw charts, solve Excel error messages

Demos

For these demos I have prepared some sample prompts. We may look at some of these.

Copilot in Word
Copilot in Excel
(possibly and briefly) Copilot in PowerPoint
ChatGPT

Should I use Copilot or other AI tools?

M365 Copilot is embedded in the M365 applications. This makes it very convenient to use. It also has “read” the the contents of the open document (Word, Excel spreadsheet) so it know the context of any prompts that you may provide.

On the other hand, Copilot may not use the most powerful LLMs. For the best answer to a tough questions, we may want to use an external LLM e.g. a ChatGPT reasoning model, even though it is less convenient to do so.

Follow your trust policy on AI.

How to have a better conversation with an AI tool

Communicating with the AI tool is the crucial skill, (This is known as prompt engineering.) We can help AI tools help us with more relevant, accurate responses if we provide more tailored prompts. Here are a few pointers to obtain better responses from AI tools.

Make the prompt more specific.
Tell the AI tool what role you want it to play.
Talk, don’t type
Provide context and relevant background.
Create a new chat and ask a new question on a different topic.
Once you have received a response, you may want to cross-examine the AI tool: for example, “provide examples that support …”
You may want to use the same prompt with different AI tools.
Don’t underestimate the AI tool.

For more information:

Tips to get better responses from AI tools

Skills we need to thrive in the age of AI

These are not necessarily technical or computer skills.

Communication
Critical thinking
Creativity

Final (longer term) thoughts

So far, we’ve thought of an AI tool to answer an immediate question or task. AI can do better than this.

Project (subject) level AI expert

We can provide an AI with a set of documents that cover a topic and ask it to base its answers on the information in those documents only, an not on its general training or by searching the web. (This is a popular use-case with organisations, know as retrieval augmented generation or RAG.)

For example, Google’s NotebookLM is designed exactly for these purposes.

Customised AI

We can ask the AI to act in a particular role (or use an AI the some-one else has customised). In education, the use of LLMs as tutors is very exciting.

For example Khanmigo from the Khan Academy is a great tutor for school kids.

Open Source AI

So far, the AI tools mentioned have been a proprietary (and paid for) remote service. It is also possible to run an open-source AI tool locally. This has some advantages: much more specialisation; no data leaves the organisation; no subscription costs paid to a third-party.

For more information:

Agentic AI

This is the manufacturer-led buzz at the moment. AI will be able to do tasks on our behalf rather than simply advising us on a course of action. For example, as well as planning our holiday, it will book travel and accommodation.

Excel and Copilot and Python form a wonderful three-way partnership

Python is a very popular computer language and the most used language for data analysis and generative AI. It is now available in Excel.

Excel’s Copilot now has an advanced analysis mode where it uses Python to do tasks such as analysing data and building charts. This works really well.

Suggestion: If you are building a significant financial report in Excel, it may be worth considering learning a bit about Python and using Copilot’s advanced analysis mode.