Microsoft Copilot – A First Look at the Workday of the Future?
Microsoft 365 has around 350 million paying users worldwide . They use the Office Suite on a daily basis to create presentations, analyze data or collaborate in teams. Most of the daily work with the Office Suite is done manually, automation is not possible for all users due to lack of knowledge. But it is precisely this challenge that Microsoft wants to solve with the introduction of Microsoft 365 Copilot. What was only possible yesterday by trained specialists, should be possible today for every user in the company through the use of artificial intelligence (AI). But how does Copilot work and what are the possible applications? The following article explores these questions and presents fundamental implications of collaboration between humans and artificial intelligence in the workday of the future.
Artificial intelligence is the subject of controversial debate. It is therefore necessary to look at the topic from different perspectives such as strategy, organization, architecture, but also corporate culture. In a first step, the technology around the MS Copilot is analyzed from a technical perspective. The strategic, organizational, architectural and cultural implications for the Copilot and AI-supported everyday working life of the future follow in a second step.
The blog post “Data-centric AI” – A change in the AI mindset? from 2022 provides the ideal introduction to the topic of artificial intelligence and explains basic concepts of the subject area . In the paper, artificial intelligence is divided into the subsets of Machine Learning and Deep Learning. The MS Copilot can be classified in the Deep Learning area and uses so-called Large Language Models (LLM) , . These are AI models that are trained to understand and generate human language.
Figure 1: How Large Language Models work. 
LLMs use complex algorithms to process large amounts of text data and recognize patterns in speech. Based on these patterns, they can generate texts, answer questions, or even understand and respond to human speech (see Figure 1). Large amounts of data are needed to train an LLM .
The MS Copilot makes use of already trained language models  and allows the user to interact with the AI via text input. What distinguishes the service from existing approaches is the direct integration into other Microsoft services using the Microsoft Graph . The Graph links information from appointments, chats, and many other Microsoft 365 data points, enabling various use cases through the use of enterprise-ready AI .
Figure 2: Overview – Microsoft Graph 
The integration of Copilot in Office Apps enables new processes and workflows for the creation of corporate content. Presentations can be created based on existing files and documents, data analysis can be performed and visualized without complex queries. The manual creation of email templates and meeting minutes is a thing of the past thanks to the use of the new technology.
In addition to the use cases mentioned, the integration via Microsoft Graph introduces a new service category, the Business Chat. The chatbot combines all data points of the graph and can thus provide information about different areas of one’s work at any time (chats, e-mails, appointments, documents, contacts, etc.) , . The interaction between humans and machines thus reaches a new, previously unknown level and the everyday work life changes significantly. But what exactly will this change look like?
Implications for everyday work in the future
For this reason, the following section examines different areas of a company that change as a result of the use of AI-supported tools (strategy, organization, culture, architecture).
At the level of strategy (1), AI will enable companies to make faster and more accurate decisions by analyzing large amounts of data . Companies will be able to develop business strategies based on comprehensive data analysis and forecasting. AI will further enable companies to develop new products and services based on customer needs .
To make this possible, however, fundamental elements of the organization (2) must be adapted to the new technological possibilities. Processes and workflows change (e.g., through the creation of corporate content by the MS Copilot), and employees must acquire new skills and knowledge in dealing with the technology. The example of the Large Language Model presented earlier illustrates this issue. In order for the model to provide the desired answers and text output, communication and understanding between humans and machines must be aligned. This process is also referred to as prompt engineering. . In this process, the outputs of the AI are constantly compared with the manual inputs of the user. By adjusting the input commands, the AI can be guided to the desired result. An example illustrates the point: only after the human input has been adjusted several times does the AI generate an invoice in the correct format. Previously, this was generated for the wrong customer group because the human input did not allow any conclusions to be drawn about the customer group. The process of prompt engineering currently has to be performed by a human; a new organizational role is created, the prompt engineer. This person must know the limitations and weaknesses of the AI system in addition to the technical capabilities in order to generate the best possible output .
Figure 3: AI and the everyday working life of the future
From the perspective of corporate culture (3), strategic and organizational changes lead to further challenges. In addition to the intransparency of strategic decisions by AI, the topics of ethics and privacy determine the use of artificial intelligence in everyday work. Human values and norms must not influence the actions of AI to the systematic disadvantage of certain groups. Human Centered Design (HCD), in which humans and machines work together to develop the basis for a decision, is a solution to this problem.  or the concept of Trustworthy AI presented by the EU . . The latter characterizes trustworthy AI systems along three dimensions: lawful (1), ethical (2), and robust (3). Compliance with laws, ethics, and values, as well as technical security, make AI systems trustworthy. Only if these values are respected and accepted inside and outside the organization a fair and sustainable use of AI can be promoted in everyday work.
The fourth dimension architecture (4) deals with the topic of data and its security. The selection of data, its use by AI and security represent central tasks in the everyday work of the future . In particular, the topics of personal data processing and data integrity must be considered here. Microsoft uses various approaches in MS Copilot to ensure these challenges, such as two-factor authentication, compliance policies, and special application architectures . The integration of the Microsoft Graph provides new data sources for the use of AI. Only on the basis of a clearly defined and structured data landscape can services such as MS Copilot offer added value in everyday work in the future.
The Copilot for Microsoft 365 presented by Microsoft provides initial insights into the everyday working life of the future and shows various potentials in the use of AI. In addition to the efficient creation of corporate content, the personalized assistant Business Chat offers new possibilities in human-machine interaction. Nevertheless, different implications for companies can be identified when working with AI. In addition to strategic advantages, fundamental challenges arise in the further development of the company’s own employees as well as their trust in the technology. The everyday working life of the future will be accompanied by fundamental questions around privacy and transparency as well as technological issues. The Copilot offering nevertheless represents a milestone in the use of AI through the enterprise-ready AI approach.
The following introduction video shows the different areas of application of the Copilot: https://youtu.be/S7xTBa93TX8
References Daniel Ch, “Microsoft 365 Suite Revenue and Growth Statistics (2023),” SignHouse, 2023. https://www.usesignhouse.com/blog/microsoft-365-suite-stats (accessed Mar. 21, 2023).  L. Wrobel and T. Hackl, “”Data-centric AI” – A change in the AI mindset? | ccecosystems.news,” Apr. 29, 2022. https://ccecosystems.news/data-centric-ai-eine-veranderung-des-ki-mindsets/ (accessed Mar. 21, 2023).  J. Spataro, “Introducing Microsoft 365 Copilot – your copilot for work,” The Official Microsoft Blog, Mar. 16, 2023. https://blogs.microsoft.com/blog/2023/03/16/introducing-microsoft-365-copilot-your-copilot-for-work/ (accessed Mar. 21, 2023).  A. Lee, “What Are Large Language Models Used For and Why Are They Important?” NVIDIA Blog, Jan. 26, 2023. https://blogs.nvidia.com/blog/2023/01/26/what-are-large-language-models-used-for/ (accessed Mar. 21, 2023).  Co:here, “Introduction to Large Language Models,” Cohere AI, 2023. https://docs.cohere.ai/docs/introduction-to-large-language-models (accessed Mar. 21, 2023).  T. Wu, M. Terry, and C. J. Cai, “AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Prompts,” in CHI Conference on Human Factors in Computing Systems, New Orleans LA USA, Apr. 2022, pp. 1-22. doi: 10.1145/3491102.3517582.  OpenAI, “GPT-4,” 2023. https://openai.com/product/gpt-4 (accessed Mar. 21, 2023).  Microsoft, “Microsoft Graph overview – Microsoft Graph,” Mar. 16, 2023. https://learn.microsoft.com/en-us/graph/overview (accessed Mar. 21, 2023).  M. Crouse, “Microsoft adds Copilot AI productivity bot to 365 suite,” TechRepublic, Mar. 16, 2023. https://www.techrepublic.com/article/microsoft-copilot-ai-productivity-365-suite/ (accessed Mar. 21, 2023).  Deloitte, “Artificial Intelligence: New Dimension in Decision Making,” Deloitte Germany, 2023. https://www2.deloitte.com/de/de/pages/technology-media-and-telecommunications/articles/kuenstliche-intelligenz-ki.html (accessed Mar. 21, 2023).  S. Girard et al, “The art of customer-centric artificial intelligence.”  E. Eliaçık, “What Is AI Prompt Engineering: Examples, And More (2023) – Dataconomy,” Jan. 27, 2023. https://dataconomy.com/2023/01/what-is-ai-prompt-engineering-examples-how/ (accessed Mar. 21, 2023).  D. Holzer, “AI Creates New Jobs: What Does a Prompt Engineer Actually Do?,” BR24, Feb. 27, 2023. https://www.br.de/nachrichten/netzwelt/ki-schafft-neue-jobs-was-macht-eigentlich-ein-prompt-engineer,TX4P23Z (accessed Mar. 21, 2023).  European Commission, “Ethics guidelines for trustworthy AI | Shaping Europe’s digital future,” Apr. 08, 2019. https://digital-strategy.ec.europa.eu/en/library/ethics-guidelines-trustworthy-ai (accessed Mar. 21, 2023).  E. Van Gavrock, “Artificial intelligence design must prioritize data privacy,” World Economic Forum, Mar. 31, 2022. https://www.weforum.org/agenda/2022/03/designing-artificial-intelligence-for-privacy/ (accessed Mar. 21, 2023).
- What Is the Theory of Disruptive Innovation? - 27.09.2022
- Does Decentralization Guarantee Stability? - 03.01.2022
- Web3 and DAOs – the next hype? - 23.12.2021