Too Much to Do, No Budget to Hire—Can Excel Copilot Help?

____________

If you are based in London and are interested in this topic please consider signing up to the meetup event on 18th February at the Microsoft Reactor in Paddington (5.30-8.30 pm): Click here to register

____________

The ability and necessity to wear multiple hats has often been portrayed as a self-proclaimed badge of honour for startup founders. Whilst it’s an incredible skill to have (It also helps explain why entrepreneurship is so attractive for neurodivergent folks), the downsides are rarely explored.

Some areas of business are easy enough to teach yourself as you go along. Many aspects of sales and marketing are so unique to individual businesses and industries anyway that going through this self-learning journey quickly is almost a pre-requisite to getting any business off the ground.

Finance and operations come along shortly afterwards, and are generally more standardised, but also more rigid in terms of how they need to be implemented. Getting your accounts filed correctly is not something you can afford to do wrong and, for this reason, compliance is often outsourced.

But, outside of returns, the financial data of a business is its lifeblood and should be used to drive data-driven decisions on an almost daily basis in order to thrive. And this is where the lack of financial/data skills becomes a challenge.

Education clearly plays a role and it’s why many startup founders choose to invest tens of thousands of dollars in an MBA to gain this generic foundational knowledge. ?Whilst I’m an MBA holder myself, I went down a less expensive route – in terms of both time and money – through Quantic, and I’m not sure paying 10-20x the cost (or more including living costs/the opportunity cost of lost earnings for a full time in-person program!) would have reaped more benefits, especially compared to just investing that additional money and time in actually running a business. ?It’s clear to me that a fundamental disruption in the education system across the board (including professional education) is required to deliver what people need to learn when they need it, rather than being forced to go through fixed syllabuses and trying to remember the content months or years down the line when they finally have an opportunity to apply it.

Technology, and Generative AI in particular, can help with this. As someone who runs my communities as essentially ed-tech businesses, I’m particularly fascinated with how AI can evolve individual learning journeys to be more relevant, engaging and timely.

But the promise of Generative AI goes well beyond this. The (albeit future) promise of Generative AI is to be the staff you can’t yet afford to hire. Like humans, it still needs a broad direction and an idea of what the end goal is, but it potentially already has the financial and data skills required to do the actual analysis without you having to learn it all yourself. You need to know just enough to check that what is being returned is sensible. But this relies to a large extent on how well you engineer your prompts and understand how it is using that (based on the engines/models that lie behind Gen AI) to provide you with a response.

AI is transforming the way businesses analyse data, making advanced insights accessible to everyone, even non-technical users.

Generative AI tools, like Microsoft Copilot in Excel, help users perform complex tasks like filtering data, calculating key metrics, and even running Python-based analytics without deep coding expertise.

Let’s bring this to a realistic scenario. Imagine a startup founder, Lisa, who runs a growing e-commerce business. She has hundreds of invoices in Excel and wants to:

• Identify late-paying customers.
• Calculate monthly revenue trends.
• Find opportunities to improve cash flow.

Traditionally, this would require manual filtering, complex Excel formulas, or even writing scripts in Python. ?With Excel Copilot, Lisa can achieve this effortlessly using natural language prompts.

?How Excel Copilot Can Help: Automating Data Analysis with AI

1.????? Filtering & Summarizing Invoice Data Using Copilot

Instead of manually filtering overdue invoices, Lisa can simply type: “Show me all invoices that are overdue by more than 30 days.”

• Copilot instantly applies the correct FILTER function in Excel.

She can also ask: “Summarize the total amount outstanding by customer.”

• Copilot generates a SUMIFS formula or suggests creating a PivotTable.

?2.????? Generating Financial Metrics Automatically

Lisa wants to calculate key business KPIs such as Total Revenue per Month, Average Invoice Value and percentage of Overdue Invoices.

Instead of manually writing formulas, she types: "Calculate total revenue for each month and show it as a table."

Copilot suggests using SUMIF or PivotTables and even formats the table for easy analysis.

?

3.????? Advanced Analysis Using Python in Excel

For deeper insights, Lisa can leverage Python in Excel, powered by Copilot. For example, to forecast future revenue trends, Lisa could ask: "Use Python to predict next quarter’s revenue based on past 12 months of invoices."

To detect anomalies. She could ask "Highlight any unusually large invoices compared to past trends."

To visualise cash-flow she could ask "Plot a line chart of monthly revenue and mark overdue payments."

In all these scenarios, Copilot writes and executes Python code behind the scenes using libraries like Pandas and Matplotlib.

?

How Copilot Works: The AI Principles at Play

Lisa doesn’t need to know all the intricate details of how Large Language Models, neural networks and probabilistic Python libraries work in order to take advantage of Generative AI. Much in the same way as she doesn’t need to know how a Carburettor, Continuously Variable Transmissions?or Pistons work in order to drive a car. But to drive a car she needs to know how the core controls – the breaks, accelerators and mirrors for example – all work in order to determine how she needs to input into the process of driving.

In the same way, it is very useful to understand the concepts on which Generative AI models are based, so that you can ensure your inputs (prompts) are optimised to deliver the best results. A prompt is different from an instruction – an ideal prompt should contain context (think of things like persona, examples, audience and intended outcomes!).

All of these things are taken into account when the prompt is converted into tokens and the knowledge base (whether that is internal or external) is referenced, Some of the key AI concepts at play when using Copilot:

• Tokenization & Embeddings: Converts Lisa’s query into a format AI understands.
• Prompt Engineering: Copilot fine-tunes responses based on best practices in Excel/Python.
• Retrieval-Augmented Generation (RAG): Copilot fetches relevant Excel functions from Microsoft’s knowledge base.
• Reinforcement Learning (RLHF): Over time, Copilot improves based on user feedback.

?

The Bigger Picture: Why Generative AI is a Game-Changer for Business Owners

With tools like Copilot in Excel, small business owners no longer need deep technical skills to extract insights.

• AI enables: Faster decision-making by reducing manual work. Improved accuracy in financial analysis. Scalability, as businesses can analyze large datasets without hiring data analysts.

The rise of AI-powered business intelligence will reshape how companies operate, making data-driven decision-making the norm. Generative AI is no longer just for tech experts—it’s now embedded in everyday tools like Excel.

Startup founders and business professionals can:

? Use Copilot to automate Excel formulas.

? Leverage Python for deeper analysis.

? Gain AI-powered insights for smarter decision-making.

What do you think? Have you used AI-powered Excel features? Share your thoughts in the comments! ??

If you are based in London and are interested in this topic please consider signing up to the meetup event on 18th February at the Microsoft Reactor in Paddington (5.30-8.30 pm): Click here to register

_________________