There is a lot of talk and experimentation about (generative) AI, but the number of truly practical applications lags far behind the attention AI is getting in the world. In the near future, I will blog regularly about practical applications of AI that you can emulate.
That's why it was so valuable to hear Wolfgang Schöffel's experiences with Copilot. Wolfgang is Corporate & Internal Communications Manager Germany & Austria at Konica Minolta.
Copilot: Microsofts AI-assistent
In an online session by Advatera, the European community of professionals in communication, marketing and digital transformation - Wolfgang discussed the pros and cons of Copilot. He shared his personal experiences and explained how he integrates this tool into his daily workflow.
Copilot is an AI-powered conversational interface developed by Microsoft. It acts as a kind of AI buddy that helps you with various tasks, such as information search and retrieval, text generation, image creation, code writing, and integration with Microsoft 365 Apps. You can use Copilot on the Copilot Web site, in Windows 11, via Bing Search and in an Edge Web browser.
AI is here to stay
Wolfgang has been using Copilot as a test user since January, so about five months now. He is also trying out AI tools from 2023 for content generation. Right now, generative AI is at the height of the hype. Wolfgang wonders if genAI will ever diminish. In his view, developments around AI are as essential and permanent as the Internet and the World Wide Web.
Wolfgang explains how for years his division had English-language press releases from Japanese headquarters translated by a certified translator. Then the images were attached. This whole process took about two weeks. That is now completely different. The English-language press release is now translated with DeepL Translate and Deepl Write. Sometimes Wolfgang's team uses DALL-E to generate images. That now takes only two hours, instead of two weeks.
Shielded data is essential
Copilot, on the other hand, is not as helpful in this press release process. But Wolfgang points out that Konica Minolta is very strict in its information policy. Information must not be processed outside the secure organizational network. Although Copilot uses ChatGPT from OpenAI, Microsoft promises that the LLM does not have access to the data, nor are the questions and answers used for training purposes. That's why Konica Minolta chose Copilot.
Wolfgang explains that they use Copilot Lab and not the integrated Copilot in Office, Bing or Edge. With an organization account, all data is shielded. This does not happen when you log in as an individual (private) user.
AI as a lightning-fast assistant
After this introduction, Wolfgang shows us how Copilot works for him. First step is to open a website. For Wolfgang, the “killer app” is that he can ask Copilot to summarize the website in five “bullets. Something similar Wolfgang does with storyboards. He does this regularly and for him it is a godsend.
The next example is Excel. Wolfgang shows an Excel worksheet. He asks the generic question, “Give me some data insights.” Copilot sees a clear sales trend and visualizes it in a graph. Wolfgang asks increasingly specific questions that Copilot quickly answers. And this is still the test version of Copilot, Wolfgang assures us.
AI as a presenter
Copilot can also create PowerPoint presentations. Konica Minolta has a ppt template. Wolfgang gives a link to a text and asks Copilot to make a presentation in 10 slides with notes in the organization template. Copilot immediately goes to work and in between keeps us updated on what the AI tool is doing.
The first result is already fine. There is a summary, there are notes for the presenter. Now Wolfgang himself can work to make it even better. The big advantage for him is that PowerPoint can now use the corporate template.
AI as communicator
Wolfgang now opens Outlook. He types a line with the gist of the text. Then he tells Copilot to create a longer version of the mail. Again, this is done at lightning speed. Wolfgang only has to make a few minor adjustments to make it more personal.
The second Outlook example is about replying to an e-mail. The colleague sent an e-mail about the application of AI. Wolfgang gives a short reply and Copilot makes a longer version of it. Copilot also saw that he has a meeting with that particular colleague and automatically adds a reminder in the text.
AI as author
Now Wolfgang has to write a five-page article on AI. Wolfgang gives a “prompt” explaining what is intended. Right after that, Copilot starts creating an article in Word.
Wolfgang notices a big difference from the “regular” ChatGPT. Copilot uses information from Konica Minolta because the tool is inside its secure network. The information is therefore much better focused on the context of the organiSation.
AI as a collaborator
For Wolfgang, Copilot's integration with MS Teams is the biggest win. Wolfgang provides a search for files posted in the past 30 days with AI as the subject. In this way, Copilot takes knowledge management within the organization to the next level.
Wolfgang asks a similar query where he mentions a colleague to be included in the search. The question now covers all conversations, planning and everything recorded in MS Teams. “What are the main points and what is still open?”
This is ideal for Wolfgang to prepare for a meeting. This way Wolfgang is more certain that he hasn't forgotten anything. He regularly gets compliments from his manager that the meetings are always so well prepared.
Next question, “Which of my and my colleague's documents are currently in preparation?” Again, Copilot knows the answer. Next step for Wolfgang is Copilot preparing an invitation in MS Teams Chat for colleagues to the next meeting with the agenda items and other relevant information.
Copilot then creates a summary of everything discussed in the meeting. This saves Wolfgang an enormous amount of time. And the minutes are better than before, too.
Conclusion: is Copilot really useful?
Wolfgang concludes his demonstration with his conclusions. Copilot in Excel is not recommended yet. In PowerPoint it is ok, but not super useful. In Edge it is valuable, as well as in Outlook although it could be better.
But Copilot in MS Teams is very valuable to Wolfgang for his daily practice. Wolfgang sees AI not so much as a way to hire fewer people, but more as an opportunity to stay ahead of the competition.
Copilot is not free. A license costs about 30 euros per employee per month, depending on the MS environment in the organization. For Wolfgang, it's worth the investment.
Wolfgang thinks the name Copilot is well chosen. He himself is the pilot, and Copilot is a tool that makes “flying” easier, but no more than that.
Getting started with Copilot
Wolfgang's advice is to start a pilot with a group of people that includes some AI critics, not just enthusiasts. Such a pilot should ideally last 6 months or perhaps longer. Consult regularly and share experiences. The key question is: Is Copilot useful and does it outweigh the investment? Also measure usage and remind people that the test group does need to keep testing.
Wolfgang regularly researches the purposes of using Copilot. This also provides interesting insights that he then discusses in the pilot sessions.
Comments