To find out more about how Microsoft’s Copilot for M365 has been adopted by Norwegian businesses, we have interviewed several leading Norwegian businesses. The organizations were given early-access to Copilot, so they have used it for half a year.
Though there are many large language model offerings on the marked, Microsoft Copilot is the most sought after because of its interlinkage with Microsoft’s ecosystem. Microsoft 365 Copilot lets organizations enable LLM functionality directly within the Microsoft 365 suite of products, giving any applicable users access to a personalized agent that is grounded in proprietary data. This includes any data available in SharePoint and OneDrive, any conversational and meeting history in Teams, and your full mailbox history in Outlook
Why Microsoft Copilot for M365?
As noted by Microsoft in their 2023 Third Quarter Earnings report, there are now 382 million paid seats worldwide for the Office 365 commercial offerings. This emphasizes the potential reach of M365 Copilot as an easy turn-key solution for the general workforce when it becomes widely available in 2024. (Microsoft, 2023) Though Copilot is the one discussed in detail in this text, the principles are applicable to other LLMs too, as the challenges and opportunities
We had the pleasure to partake in BI’s first official completion of the new Executive programme Responsible AI Leadership, under guidance of associate professor Samson Yoseph Esayas and professor Christian Fieseler, along with esteemed guest lecturers. Throughout our examination of how Norwegian businesses have implemented Microsoft Copilot, these are the some of the most important features concerning a safe, responsible and effective adaptation that have been utilized:
These are 6 of the most important considerations prior to implementation
- Data access: Copilot for M365 can search through, process and generate an impressive amount of text in a matter of seconds. However, organizations must have proper data access is in place prior to adaptation. If classified data or business critical information is floating around in tenants accessed by Copilot without the proper safeguards in place, chances are this information will be sent to unauthorized persons thus compromising its confidentiality.
- Data lineage: Data access and data lineage are closely interlinked. While the initial term refers to the proper administration of access, lineage means that the data’s clearance level is ensured throughout its life cycle. Hence, confidential data cannot be used in texts or other media with a lower grade of restrictive use.
- Code of conduct: To ensure ethical and safe use, the Norwegian businesses with early-access to M365 Copilot have implemented a code of conduct giving its end users a framework to support their decisions. The code of conduct, must balance between a general and thus universal applicability whilst simultaneously being bespoke to address real-life scenarios and actual use cases. A too general code of conduct will lack the specific information necessary to aid end users. If the guidelines are too specific, they must be changed and updated frequently in accordance with new tools, options and features. This will make it harder to navigate and less predictable.
- Legal framework: Much like the principles applied to the code of conduct, most of the laws governing AI are general and often open for interpretation. Laws such as the General Data Protection Regulation has been criticized numerous times for being too general, making it challenging to navigate. Much of the same critique has been made of EU’s new Artificial Intelligence Act, which is set to be fully adopted within a couple of years. In addition, several other laws, both supernational and national, govern the use of AI. It is pivotal to understand these laws to ensure safe and responsible adaptation. The consequences can be dire, as the Artificial Intelligence Act can penalize severe misconduct with fines upwards of 6% of global turnover, or 30 million euros – whichever is highest.
- An incremental approach: Businesses that have had a successful implementation of Copilot, have applied it incrementally – one step at the time. Adopting Copilot in controlled environments have reduced the frequency of serious risks by notable margins. Giving selected end users the chance to test out generative AI in smaller segments before enrolling it to the entire organization has uncovered errors and challenges. This has enabled the organization to address potential hazards before they grew to bigger problems.
- Mapping out opportunities: Numerous reports state how important AI will be for business development and growth. An example, is NHO’s report stating that AI can create value of as much as 5600 billion NOK. This, however, is conditional. Norwegian businesses need to map out use cases and opportunities for utilization and adaptation of AI.
These are 3 of the most important considerations during and throughout implementation
- Continuous assessments: Technology change, and AI is no exception. In fact, the exponential growth of AI tools and software is unprecedented. In addition to technological development, the legal framework governing AI is constantly changing. Thus, continuously assessing data impact, risks, opportunities and judicial implications are fundamental to ensure sustainable safety and efficiency.
- End user training: Everyone has tried ChatGPT, and thus the misconception of AI being “easy to use” flourished. However, to efficiently use AI it is necessary with comprehensive end user training. Users must understand the code of conduct, learn to prompt, learn to categorize data, know how to look for and report mistakes and errors – and much more. Copilot and other generative AI also changes quickly, adding new features and possibilities. Thus, the end user training is continuous aiming to keep its organization up to date.
- AI adaptability: In 2023, the World Economic Forum reported that 23% of all the world's jobs will be affected by AI in the next 5 years. Furthermore, 83 million jobs will be discontinued and another 69 million will be created. These are transformative figures, ushering in an unprecedented need for both reskill and upskill. Businesses need to address how they will incorporate AI technology and the ramifications it will have on both current and future employees.