Here’s an interesting statistic from Microsoft regarding how long certain technologies took to reach 100 million users: Mobile Phones took 16 years, the Internet – 7 years, and Facebook - 4.5 years. So how long did ChatGPT take to reach an audience of 100 million? The answer – just 3 months.
What is Microsoft Copilot (and what does it mean for my business)?
And so to the question “What is Microsoft Copilot?”…
At its heart, Microsoft Copilot is Microsoft’s incarnation of ChatGPT. It’s the result of Microsoft’s ongoing collaboration with ChatGPT’s founders, OpenAI, with Microsoft using the OpenAI source code to develop their own Generative AI solution called ‘Copilot’.
Both ChatGPT and Copilot are a form of ‘Generative AI’, which is a subset of the many forms of AI and operates on a model which utilises a Large Language Model (LLM), and is trained to recognise patterns in data based on pre-learning its subject area. In the case of ChatGPT, the subject area is pretty much the content that can be found on the Internet, whilst in certain forms of Microsoft Copilot, the subject area is the data held within your Microsoft tenant.
Generative AI technologies are designed to respond to a prompt (mostly in the form of text, but see below for where this is heading), which then combines with the LLM to generate unique content on a predictive basis. Launched only 18 months ago (in November 2022), ChatGPT’s phenomenal rise to ‘fame’ has largely been driven by its millions of users discovering that it is capable of creating unique responses to requests to generate CVs, school homework, job adverts and so on.
Generative AI is also capable of creating unique images, again in response to a prompt. OpenAI’s Dall-E is perhaps the best known of this form of AI, with its uncanny ability to render ‘realist’ images from simple prompts such as “Please generate a picture of a goldfish driving a tractor in Monaco” (try it for yourself).
As already mentioned, Microsoft Copilot is a form of Generative AI based upon ChatGPT’s source code; however Microsoft have designed Copilot to operate in an embedded manner within various aspects of a client’s Microsoft environment. At time of writing, there are dozens of variants of Copilot, including: Microsoft Copilot for Microsoft 365, Copilot for Azure, Security Copilot, and Copilot for Power Platform. In each case, Copilot is designed to empower users by providing answers to specific contextual questions, or to generate new content or information based upon the data in their Microsoft environment.
The above may all sound interesting from a technological point of view, but to answer the question posed by this article: “What does it mean for my business?”
Given the variety of Copilot technologies available, the simplest way to answer this question is to discuss Microsoft’s flagship technology, Microsoft Copilot for Microsoft 365 (M365). Licensed on a per-user basis, Copilot for M365 can be provided to your users on a case-by-case basis, although we’re finding that the real power of Copilot comes from a wider, organisational adoption.
When first implemented in your environment, Copilot for M365 will take a little time to complete what is known as ‘semantic indexing’. This is the process where Copilot processes all of the data within your Microsoft 365 Tenant in order to build a map of the information and knowledge stored within your environment. This essentially becomes a Large Language Model, trained on your data. Subsequent prompts to Copilot via your Office 365 apps will use this index to form responses, and to command functionality with various Microsoft Apps, such as Word, Excel, and PowerPoint.
Firstly, it is important to state that, whilst Copilot can ‘see’ all of your data, it still only operates using the credentials of the calling user. Quite simply, this means that Copilot will not disclose information to a user who would not normally be able to access that information using any other Microsoft product (e.g. SharePoint, OneDrive). Secondly, It must be stated that the index, or LLM, that it creates, remain strictly within your Microsoft Tenant, with no sensitive data being stores elsewhere, or used to train other Copilots outside of your environment.
From a user’s perspective, Microsoft Copilot can boost productivity by providing real-time assistance, improving writing, automating repetitive tasks, and providing quick access to information. It integrates with Microsoft 365, making it easy to use and enhancing workflow automation.
Consider how the following might improve efficiency in your business:
01
HR
How your HR team might save time with the ability to generate job adverts, respond to applicants, or produce draft contracts in mere seconds.
02
PA
Have Copilot attend a Teams meeting (with or without you) and generate a full transcript of the meeting in a manner that you can interrogate with simple questions such as “Where there any actions for me?”, or “What were the key highlights of the meeting?”
03
Marketing
Use Copilot to create a full PowerPoint presentation, with speaker notes, by simply providing it with a template, and pointing it to a product user guide.
Finally, it should always be remembered that the responses generated by Copilot (as with any generative AI) are not ‘intelligent’. They are merely the construct of a series of algorithmic decisions made based upon probabilities. The outputs should be thoroughly checked before sharing with other colleagues, or external parties.
As for the future; we are currently working with clients to develop their own, business-specific Copilot by training the AI using data from across their business to allow bespoke responses to be produced. Using Microsoft Copilot Studio, we are able to combine ‘knowledge sources’ from a variety of locations to allow Virtual Agents, underpinned with Natural Language speech recognition, to respond to specific questions such as “How many 15mm roller bearings are currently in stock?”, or “How much annual leave do I have remaining?”
Further into the future which, according to rumours currently doing the circuit, means 2027; the era of Artificial General Intelligence (AGI) will arrive courtesy of OpenAI. AGI is considered a level of AI that matches, or surpasses, that of a human being across a range of areas. With this, comes the exciting (or otherwise) prospect of being able to create a subject matter expert in a particular field, and to then spawn hundreds, or thousands of copies in minutes, and have them all concentrate on finding a solution to that one problem. It should be interesting!
Given its capabilities, a colleague asked me if I’d used Copilot to produce this article in order to save time. I’ll leave it to you decide if I did or didn’t…