Microsoft’s Copilot: A force multiplier for KM

Generative AI (GenAI) applications will increasingly transform organizations’ IT platforms. Companies of any size that opt to create robust apps on their own, however, are in for a protracted, complex, and expensive experience.

There’s a better way: Buy into what I call a GenAI ecosystem from a vendor in whose tech you are already invested. These ecosystems are comprised of the sum of services customers mostly need to build and launch robust apps.

For those on a Microsoft platform, that’s a good thing. Its Copilot GenAI ecosystem is an elegant “single experience” of GenAI across its core products—Windows 11, Microsoft 365, Edge, Bing, and Azure. It also plays well with GitHub (remember that Microsoft owns GitHub), so programmers can automatically create and complete code. Users always have Bing Search in their workspace so that, when they query, Bing responds in a co-pane variously with answers or summaries of extended texts, such as online book chapters and long documents. Follow-up questions prompt more targeted searches for more exact responses that can include graphics, videos, blog posts, and other organizational information.

For sales, marketing, and customer service, Copilot analyzes content from customers’ company data as well as from the web. It functions, in essence, as a virtual assistant, which also becomes a KM assistant, wherever employees are in Word, PowerPoint, Excel, or Teams.

The process for developing and deploying a GenAI app is as follows: Organizations choose use cases > large language models (LLMs) > develop data pipelines into LLMs > train LLMs > experiment with prompt types > evaluate pilot performance (say, for relevance of data derived from different prompts) > evaluate pilot’s safety (say, to ensure privacy) > refine/scale to operationalize > monitor/update across time.

To achieve this, organizations leverage Copilot services common to most other IT and social media competitors in the GenAI ecosystem space.

Partners. GenAI is proving its value in horizontal use cases such as marketing and customer service and in vertical ones such as advertising, finance, and retail. Microsoft offers Copilot-enhanced Azure AI solutions in all of these. However, it sells into 25-plus verticals, and Copilot will eventually play in them all.

LLMs. GenAI ecosystem vendors provide their own LLMs, partner for LLMs, or enable environments for customers to create their own. Microsoft invested early and big in OpenAI and uses its GPT-3 family of LLMs and partners for Codex, DALL-E, and Whisper. LLMs come “as is,” but customers mostly train them by ingesting their own datasets via pipelines and by experimenting with prompts.

Search. Bing via Edge integrated across Windows 11, Microsoft 365, and Azure is robust, various, and rich. Using natural language (NL) queries, image, and other prompt types, users generate responses in all forms mentioned above—plus speech, semantic feedback, and sentiment analysis. Sentiment analysis, for instance, when used in a call center GenAI app, confirms a customer’s emotional state to appropriately respond to her. Bing AI also personalizes responses based on users’ search histories. As these activities are core to KM, Bing AI is a manifold KM tool that queries, derives insights from, and creates novel content for any knowledgebase within an organization or on the net.