Staying ahead of the curve with Microsoft Copilot generative AI

Published on December 19, 2023 by Cristina Elizondo

Introduction

It is no secret that artificial intelligence (AI) has been transforming from a mere concept into an integral part of our daily lives, up to the point where it is present in almost every sector, e.g., healthcare, finance, manufacturing and transport. The 21^st century marked a significant shift in AI advances in areas such as machine learning, specifically deep learning, fuelled by the availability of large datasets and powerful computing. Thus, AI applications started to learn how to recognise patterns, process language and make informed decisions.

As AI continues to evolve, it has reached a point where it can replicate human intelligence through computational systems, and these systems have been innovating and creating new features, attempting to leave only the “creative” part of the job to the human. One such example is Microsoft’s new tool, Microsoft 365 Copilot. This leverages the power of machine learning by offering intelligent recommendations and insights based on a user’s work patterns and preferences (source: Walker, 2023)^[1].

This blog aims to define and analyse the different use cases of Microsoft 365 Copilot and how businesses can leverage it to increase efficiency. We also compare it with similar tools such as GitHub Copilot and discuss trends relating to the next productivity frontier in terms of generative AI tools and their main use cases.

The evolution of generative AI Financial Services

AI is known to have five main branches, as Figure 1 shows. Machine learning is expected to report the most growth, at a CAGR of 36.2%, by 2030 (source: Fortune Business Insights, 2023)^[2]. One subset of machine learning, combined with deep learning and natural language processing (NLP), is generative AI. This has seen an exponential increase in attention in recent years because of its ability to produce, with the help of algorithms and models, original content – from images, text and music to entire virtual environments (source: Nadeem A, 2023)^[3].

These generative AI financial service models learn from extensive amounts of data and generate new content that resembles the original data distribution (source: Great Learning, 2023)^[4]. There are five main models following this definition:

1. Variational autoencoders (VAEs)

• Main applications: image generation, anomaly detection and data compression

2. Generative adversarial networks (GANs)

• Text-to-image synthesis, creation of photorealistic images and video generation

3. Auto-regressive models

• Text generation, language modelling and music composition

4. Flow-based models

• Image generation, density estimation and anomaly detection

5. Transformer-based models

• Used for NLP tasks, completing text, answering questions, translating and summarising

What is Microsoft 365 Copilot?

In March 2023, Microsoft introduced what could be one of the greatest AI tools to facilitate daily, routine work. Microsoft 365 Copilot basically combines large language models (LLMs) with data in the Microsoft Graph and Microsoft 365 apps to become “the most powerful productivity tool on the planet”.

As Microsoft 365 Copilot is integrated into the well-known Microsoft apps, it would “unleash creativity, unlock productivity and uplevel skills”. It would do this by using NLP and machine learning (ML) algorithms to understand the context of the data and generate suggestions based on that context. This data includes emails, chats, documents and meetings, enabling the tool to learn your daily activities and tasks, make informed recommendations to ease the tedious aspects of your daily work and leave just the “creative” aspects to you (source: Finnegan, 2023)^[5]. Figure 2 shows its use cases:

Advantages and drawbacks in business

The most evident advantage of this tool is increased productivity and efficiency, together with enhanced creativity and accessibility across roles within a team. Copilot also eliminates the learning curve associated with Office applications. The main advantage is that a user needs only to ask Office to perform a task; they do not necessarily need to know how to perform the task themselves. Thus, it creates space for new freedoms within certain workflows (source: Omnidocs, 2023)^[7].

However, these “freedoms” have limitations and must be handled carefully, as they could result in issues relating to compliance, transparency and privacy (source: Credo AI, 2023)^[8], such as copyright infringement and claims, violation of regulatory and organisational jurisdiction and Copilot’s inability to deal with highly sensitive data or data subject to strict regulation. All these consequences are compounded because Copilot is still not available for universal use, and Microsoft has not yet announced a date for a possible official launch, increasing uncertainty among potential clients.

Similar AI tools

The speed at which gen AI financial services technology is developing is impressive. ChatGPT was released in November 2022; in March 2023, OpenAI released a new LLM called GPT-4 and Microsoft announced the release of Microsoft 365 Copilot. Just three months later, GitHub Copilot released its program (source: McKinsey, Chui et al., 2023)[9].

Although all these tools come from the same type of technology, their infrastructure differs as does the way in which they increase business efficiency. GitHub Copilot, for example, is geared for use by developers and writes code based on natural language prompts (source: Nagel, 2023)^[10]. It increases business efficiency basically by programming faster. It provides suggestions, by either starting to write the code you want to use or by writing a natural language comment describing what you want the code to achieve. It can, thus, write complete functions and smaller programs on its own to ease and expedite the process by working faster and with fewer errors (source: 1902 Software, 2023)^[12].

While GitHub Copilot and Microsoft 365 Copilot share some architectural similarities at the back end, Microsoft 365 Copilot enables businesses to improve efficiency by almost completely eliminating routine work by being able to answer questions, provide recommendations and even generate content such as emails, social media posts or complete decks (source: Walker, 2023)^[1].

Potential trends

Generative AI has seen an exponential increase in reach over the past decade, but its power and capabilities relating to the different sectors are still at an early stage. The latest McKinsey report on “The economic potential of generative AI” predicts that this type of AI would transform roles and boost performance across functions such as sales and marketing, customer operations and software development. It could also add value to the global economy given its impact on productivity.

Four key insights from this report indicate what the future holds for generative AI in terms of economic impact:

• Generative AI is expected to handle 75% of customer operations, marketing and sales, software engineering and research and development

• It is expected to change the anatomy of work by automating most daily activities, freeing up time to augment other capabilities

• It is estimated to add USD2.6-4.4tn annually across 63 use cases, increasing the impact of all AI by 15-40%

• Such technical automation will likely accelerate the pace of workforce transformation

Source: McKinsey, Chui et al., 2023^[9]

The era of generative AI is just commencing, and businesses would need to adapt to these continued, significant changes, especially in determining new skills and capabilities and restructuring the core business process, including resilience training. Microsoft 365 Copilot would eliminate the need for routine work, helping staff enhance their creative skills and have more time to explore other areas, affording a better work-life balance. It must be noted, however, that this and most generative AI tools come with certain drawbacks, especially in terms of informational risks and violating jurisdiction agreements. This new productivity tool does not have a clear launch date, but once available, would undoubtedly “unleash creativity, unlock productivity and uplevel skills”.

References

• [1] Tony Walker (March 2023). Microsoft 365 Copilot vs ChatGPT: A Comparative Analysis. AI wars has just begun. Retrieved from Microsoft 365 Copilot vs ChatGpt: A Comparative Analysis. AI wars has just begun (linkedin.com)

• [2] Fortune Business Insights (May 2023). Machine Learning Market: Market Research Report. Retrieved from Machine Learning Market Size, Share, Growth | Trends [2030] (fortunebusinessinsights.com)

• [3] Nadeem, A (June 2023). The Evolution of Generative AI and Generative Machine learning. Retrieved from The Evolution of Generative AI and Generative Machine learning (linkedin.com)

• [4] Great Learning (June 2023). Generative AI Models. Retrieved from Generative AI Models (mygreatlearning.com)

• [5] Matthew Finnegan (June 2023). M365 Copilot, Microsoft’s generative AI tool, explained. Retrieved from M365 Copilot, Microsoft’s generative AI tool, explained | Computerworld

• [6] Colette Stallbaumer (March 2023). Introducing Microsoft 365 Copilot – A whole new way to work. Retrieved from Introducing Microsoft 365 Copilot | Microsoft 365 Blog

• [7] Omnidocs (April 2023). Microsoft 365 Copilot: The good, the bad, and the ugly. Retrieved from Microsoft 365 Copilot: The good, the bad, and the ugly (linkedin.com)

• [8] Credo AI (July 2023). Microsoft 365 Copilot. Retrieved from Microsoft 365 Copilot – AI Vendor Risk Profile – Credo AI

• [9] Michael Chui, Eric Hazan, Roger Roberts, Alex Singla, Kate Smaje, Alex Sukharevsy, Lareina Yee and Rodney Zemmel (June 2023). McKinsey Digital. The economic potential of generative AI: The next productivity frontier. Retrieved from Economic potential of generative AI | McKinsey

• [10] Becky Nagel (April 2023). AI & IT: What's Up with Microsoft Copilot? A Q&A with Brien Posey. Redmond Magazine. Retrieved from AI & IT: What's Up with Microsoft Copilot? A Q&A with Brien Posey – Redmondmag.com

• [11] 1902 Software (June 2023). How GitHub Copilot is Changing the Game for Software Development. Retrieved from How GitHub Copilot is Changing the Game for Software Development (1902software.com)

---

---