Microsoft Project Cortex: An Overview
Announced in November 2019, Microsoft will shortly launch a new AI-powered service in Microsoft 365 called Project Cortex. Its ambitious goal is to finally solve knowledge related challenges affecting many organisations. Combining existing Microsoft technologies, like Microsoft Search and the Microsoft Graph, Azure Artificial Intelligence and Machine Learning, Project Cortex creates a “Knowledge Network” of people, information and skills within an organisation.
Potential Challenges
Many organisations are facing challenges about upskilling people and getting people up to speed quickly when they change roles. Also, many organisations have offices in multiple locations, sometimes located in different countries (sometimes continents!). Even within the same offices, different teams have multiple tools and repositories, making it hard for other people within the organisation to access the information they may need.
To summarise, most organisation face challenges in the following areas:
- Knowledge compartmentation;
- Knowledge sharing;
- Knowledge retention;
- Knowledge access.
Connecting information to people and people to information, Project Cortex is Microsoft’s answer to the challenges above.
In a nutshell…
Project Cortex employs AI for organising knowledge across people and systems. This knowledge is filled in the apps people use every day, allowing them to get information quickly and efficiently.
With Project Cortex, Microsoft is trying to expand its vision of knowledge within an organisation following three main themes: empower, organise, manage.
Empower
Empower people with in-time knowledge and expertise available from the app we use every day. Project Cortex collects and organises the information within the company into:
- Topic cards: AI-generated quick summary of the topic, along with people and resources (PDFs, Word documents, and other related files). Resources are people-specific: different between teams, based on users’ permissions.
- Topic pages: AI-generated wiki-pages providing all the relevant information on a topic. They also show, in a graphical format, the knowledge network and related topics. Q&As will be also shown on the pages and users will be able to enter them through yammer.
- Knowledge centres: users can access all the pages which are related to them, via the documents they worked on. Also, they can discover new topics they may be interested in. Knowledge centres are standard SharePoint pages users can customise and access for editing their knowledge: content, topics and so on.
It’s wiki-like: experts can add, delete or modify what they like. This way, employees can easily find the information they are looking for such as summaries, experts and relevant people within the organisation, related resources and topics.
Organise
Organise content coming into the company and make the content smarter with:
- Microsoft Graph to get data and content.
- Signals to get connection and collaboration around documents, meetings, and tasks
Probabilistic programming is employed to better understand the content, the main areas of a document, and connections to other documents. These are verified. Connections and signals can be shown on a topical map.
Organisations have control on the content deciding what websites or documents can be excluded/included. The same applies to topics: sensible topics can be excluded. A revision process (accept/reject) could be put in place. Finally, topics can be merged or archived.
Manage
Other than manually updating the content, three other methods are available:
- Automated Object Detection: Azure Computer Vision Services and Azure Search available in the SharePoint area to get content from photos, scanned images and so on. Handwritten documents are understood.
- Form processing: with AI Builder from Power Platform users can build AI Machine Learning (ML) models to get content from documents having a standardised template such as forms, claims, invoices, etc.
- File classification and entity extraction from unstructured documents (contracts, CVs, manuals, etc) trough Machine Teaching (MT) from Subject Matter Experts (SMEs) and content owners.
Content Centres
Machine Teaching (MT) can be done through Content Centres which also provide monitoring and analytics capabilities.
When the documents are uploaded, the relevant metadata is automatically produced. Per each document, columns such as title, notes, and file type are automatically generated but new columns can be added.
Similarly, when structured documents are uploaded, UI Builder in Power Apps can be used to automatically map key-value pairs from documents. Fields can be confirmed, deleted, added, modified etc. When the model is created, it will be used to detect the value from the trained fields within similarly structured documents.
Project Cortex is an attempt to move from information to knowledge, and, at the same time, from knowledge to learning. In Microsoft words: “The information I need, in the apps I use every day”.
Microsoft has planned to launch Project Cortex in the first half of 2020, so we can expect some announcements in the next few weeks.
