Artificial intelligence is a hot topic that sparks debate. Organisations are starting to explore the possibilities of AI and consider how they can make the most of it in their own operations. However, AI experiments seem to fail easily as organizations rush to adoption without thinking about the goals. Technologies are introduced, people get bored, and stop using them before they even really get going. What could we do better to succeed with AI? This blog provides answers to this question and presents Gofore’s own experiences of piloting one AI instance, Microsoft Copilot.
In her blog, my colleague Elina Mattila calls artificial intelligence the next revolution of work. And it really is! AI itself is not new nor is it new to Gofore but from a knowledge worker’s perspective, it is. The latest manifestation of AI for knowledge workers is Microsoft Copilot. We at Gofore started our own pilot for Copilot a few months ago. It gave us the opportunity to have our own knowledge workers, consultants, testing Copilot and analyzing their experiences with our AI experts. We have been able to dive deep into the good practices of Copilot implementation and also identify pitfalls. This allows us to better serve our customers who want to start exploring the wonderful world of artificial intelligence and particularly Copilot .
In short: Microsoft Copilot is an AI-powered writing and productivity tool assistant. It can be used for e.g. files, emails, meetings and reports. It learns from the user’s style and preferences, and suggests content based on these. Contrary to what many technical experts have mistakenly thought about Copilot, it is not a search engine. That’s how it may look for sure but make no mistake about its appearance. It is a tool for learning.
Experiences of piloting
For this blog, I interviewed a few people from our own pilot project: Senior Consultant Harri Juntunen who is specialised in AI and change manangement Senior Consultant Päivi Pajunen. I also spoke with Amin Hassan from Gofore’s people operations development, who actively tested HR data and chatbots. I’m a knowledge worker myself and participated in the pilot as an end-user. I was happy to be part of the pilot since I’m also a long-term Microsoft enthusiast. My own use in Copilot focused especially on Teams and Word and summarizing different data sources.
Copilot is a change in human behavior
Both Päivi and Harri believe that the success of Copilot depends on people. “Using Copilot is about letting go of old habits, unlearning and learning new ways of working,” says Päivi. According to Harri, Copilot should not be considered as just one thing. “Copilot is not a system to be deployed, such as an ERP system. Copilot is more of a learning journey, a way to start working in a new way supported by artificial intelligence,” he describes.
Copilot is not just a technical solution, but a change in people’s behavior. It requires more than license sharing or technical training. The transition towards an AI-assisted way of working requires people-driven change management and support for continuous learning. Päivi describes that the core steps of change management guide the process, but perhaps in a slightly different way than usual.
“Whereas we usually first raise awareness and only then start supporting understanding, with Copilot the two must be done hand in hand from the beginning – awareness must be raised through concrete use cases. After this, we can reach the adoption phase, and in the end, the change must also be reinforced,” Päivi emphasizes. She says the most important thing in all these steps is to ensure two-way communication, the opportunity to participate and be heard, ongoing discussion.
Set goals to achieve benefits
Päivi emphasizes that with Copilot, the first question to focus on is why. It should be clear what the reasons and goals are for the implementation of Copilot. “Playing with copilot without focus leads nowhere. Utilizing artificial intelligence is a strategic decision that needs management support.” It is worth considering whether Copilot is specifically the right solution for existing needs. Copilot can help you find new options for doing things and opportunities that may have been overlooked in the past.
Expectation management is important
Amin talks about how people participating in the pilot need to understand what they’re getting into. It should be understood that Copilot is an evolving solution and still unfinished. He says that he also was expecting for the solution to be more complete after having used ChatGTP for some time already. “ChatGTP is ahead in many areas and I’ve liked it a lot, but the advantage of Copilots is working with files,” Amin describes.
Amin speaks of Copilot in the plural. This is because each application that can use this tool has its own version of Copilot. There is one Copilot in Teams, another in Word, and a third one in Excel. Each Copilot is developed individually, which is why they are at a very different stages of maturity. Each Copilot also learns from its user at a different pace. Users should be aware of this.
Prompting is a skill
We spoke with all three interviewees about prompting. Since Copilot is not a search engine, you do not type search terms, but you must give it prompts. Prompting enables learning. “Prompts need to be clear and detailed, a bit like precise task definitions. The better you prompt, the better the Copilot learns and the better it will be able to help. Making prompts is a skill!” Harri says.
A good example of a prompt is writing an article. At the prompt, you provide Copilot with the article title, keywords, outline, tone and style. Based on these, Copilot creates a draft that you can then edit, refine or continue. I confess that I used this method partly to write this blog as well. Another example is the creation of a new website. In this case, you would provide Copilot with information such as the layout, theme, functionalities and content of the website. Copilot generates code and HTML from these, which you can then customize and deploy.
Copilot is a personal assistant
While the examples above sound nifty, it’s good to be aware that Copilot isn’t a holistic solution that takes care of everything for you. It’s a tool you work with, it learns from user’s feedback and preferences: the more you use Copilot, the better it adapts to your style and needs. Copilot is a different user experience for each individual. “Copilot is a personal tool, so its benefits are also first personal before they become the benefit of the organization,” says Päivi.
Copilot is also not a flawless tool that always produces perfectly accurate and reliable results. By learning, it can improve over time and evolve through the corrections and suggestions you provide. At its best, Copilot is an effective support that can help the user in their work, but it is not an independent tool that can work without guidance. It requires user input, guidance and control. Copilot can help based on prompts, but it cannot replace the skills, knowledge and creativity of its user. Copilot is an assistant at work, not a substitute.
What should be considered with Copilot?
Copilot as a product is constantly evolving. Currently it works very well, for example, with Word and text editing as well as Teams, where it can transcribe meetings, summarize conversations, and suggest actions based on conversations. Something to consider with Teams is social etiquette. Just as we have gotten used to asking for consent to record Teams meetings, we need to start doing the same when launching Copilot. We also need to consider whether we want to discuss sensitive issues at all when Copilot is present at the meeting. We must remember to spend time on creating these types of practices when we start using artificial intelligence in our everyday work.
A lot of thought is also given to Copilot’s information security. It’s good to understand that Copilot is not a standalone application with separate access rights: it doesn’t have access to data that you don’t have access to. It’s an extension that uses your access to both your own data as well as your organization’s shared data. As such, Copilot does not add so-called security gaps. Copilot is just as secure as your organization’s data collection anyway.
Of course, the use of Copilot sets requirements for the data used. “If the organization’s data masses are fragmented, so are the answers provided by Copilot,” Harri describes. Amin adds that artificial intelligence thrives on different kinds of information than humans. “For example, in different intranets, a lot of effort is put into making the pages look nice, with text and pictures here and there. But Copilots don’t understand it and can’t act on that information.” For AI, data must be structured and clean.
The time is now
Many organizations are wondering when to jump on the AI train. Our recommendation is not to wait for the perfect time to get started – start learning now. The best way to get started with Copilot is to practice and learn. “Start small. There is also no reason to separately invent challenges to be solved with Copilot. The best way is to start by selecting existing problems or issues already identified, and let Copilot produce solutions to them. Choose 20 people who will start testing and plan the next steps from there,” Harri proposes.
Copilot is a powerful and innovative platform that can transform the way we work and learn. AI can be used to improve productivity, quality and creativity and achieve goals faster and easier. Copilot is a learning assistant that can help get the job done more smoothly.
If your organization wants to try Copilot and see how it can benefit your work, contact us and let’s start experimenting together!
How to start deploying AI in a people-driven way? Get inspired by our views on AI
AI is the next revolution of work
EU AI Act shapes AI and the ways we apply it
What should you know about AI right now
Four perspectives on utilizing AI successfully
Future of automation – making society not less but more human
