Whatever happened to Project Cortex? Microsoft introduced Cortex with great fanfare at its 2019 Ignite Conference, with subsequent posts reaffirming general availability by mid-2020. As late as May 2020, general availability was still being touted for the first half of 2020. After missing the delivery date, Microsoft deferred Cortex to later in the year, stating in July: “you can expect additional details about general availability at Microsoft Ignite 2020.”
So when Microsoft CEO Satya Nadella took the stage at Ignite 2020 a few weeks ago, expectations were high.
Well, we didn’t get Project Cortex. But we did get SharePoint Syntex. Microsoft is releasing Syntex as the first available service on the road to realizing the Cortex vision.
What happened? Simple. Auto-classifying Microsoft 365 content turns out to be a lot more complicated than originally anticipated. Content is messy. It is hard to classify documents and emails in ways that make sense to everyone who needs access, with or without artificial intelligence involved.
I predicted last year the ambitious release date wasn’t realistic. As a long-time proponent of topic computing, it seemed obvious that the monumental organizational and technology hurdles could not be surmounted in a single year. Turning the Microsoft 365 information infrastructure into a semi-automated knowledge system is a herculean task, regardless of the resources invested.
The fact that Cortex is a more a vision than a product was eventually understood at Microsoft as well. Dan Holme, Microsoft director of marketing told me “the idea of Cortex was to drive customer awareness of where Microsoft was going; such a significant investment … can be best brought to market and adopted by customers as a couple of major deliverables. The first was the foundational capabilities of metadata and new taxonomy services that was released July and is available to every customer. This provides a great foundation for knowledge management, search, discovery and content processing to unlock knowledge from content and simultaneously capture knowledge that's in people's heads … The idea is to accelerate people’s work and to leverage one person's expertise and take the knowledge captured in a model and scale it to other workers in the organization.”
Welcome SharePoint Syntex
So what exactly is SharePoint Syntex? The official Syntex announcement doesn’t provide much insight: “SharePoint Syntex mainstreams machine teaching, enabling your experts to capture their knowledge about content in AI models they can build with no code. Your experts train SharePoint Syntex to understand content like they do, to recognize key information, and to tag content automatically.”
Metadata Is the Key
Let’s break it down. Metadata is the construct in SharePoint for classifying content. To classify a document, you assign it metadata properties. For example, a document can be defined as ‘draft,’ a ‘contract,’ and associated with the ‘Keystone Project.’ But successfully using metadata has always been a challenge, for several reasons.
First, information architects need to agree on how to classify a document, i.e. which properties are important to facilitate future retrieval. For example, is it more important to organize documents by project, department or document type? Maybe all three are important. These are not trivial decisions — they cause countless info-turfs wars. A CIO at an organization running a knowledge management project once told me how difficult this can be. She said that after two years of negotiation, their panel of experts could only agree on four metadata properties, two of which were the document title and the date the document was created.
Second, even when metadata is well-defined, users routinely revolt. Knowledge workers chafe at having to assign metadata every time they add a document to SharePoint, so they find ways to circumvent the process. Even when metadata is applied, it's rarely applied consistently or uniformly. SharePoint becomes a document graveyard as a result, a place where people struggle to find documents.
Enter SharePoint Syntex. With Syntex, content subject experts define metadata for documents and create examples of proper classification. Then Microsoft AI, natural language processing and machine learning are unleashed to learn from these experts and automatically make similar metadata assignments on unclassified documents. The subject matter experts then audit the automated selections and correct mistakes, thereby improving the AI models for autoclassification. This probably works best for structured or semi-structured documents like contracts and invoices, which explains why these two document types appeared in the Ignite 2020 Syntex presentation.
With Syntex, a small effort exerted by a team of motivated knowledge managers could allow large teams to reap the benefits of metadata tagging without incurring the overhead. Syntex could be a big deal for organizations for whom managing documents is important. For these organizations, being able to find content quickly is more than just a productivity issue, it is a business enabler.
Related Article: The Secret Sauce Behind Project Cortex: Good Metadata
Does Syntex Work?
The big question is whether Syntex actually works as advertised and how much effort is needed to accurately train the AI models. Of course, there is no single answer. A lot depends on how structured and similar the documents are, how well and consistently they are trained by the subject matter experts, and how well Microsoft AI works.
According to Microsoft, dozens of early adopters are already using Cortex AI upon which Syntex is based, so we can assume the AI will continue to improve over time. A lot of Syntex success will depend, as always, on the quality of the user experience for defining metadata, as well as for training and correcting the AI models.
Do You Need a Knowledge Manager?
In a recent training session, Microsoft advocated for hiring a knowledge manager in organizations preparing to deploy Cortex. This person is responsible for defining the information to be shared. They select resources to be mined, as well as the structure and visibility of knowledge. These individuals work together with subject matter experts to train the AI models by applying ‘correct’ and complete metadata, and with content services admins who assign and manage access rights and service requests.
With Cortex slimming down to SharePoint Syntex, do organizations still need a knowledge manager? Naomi Moneypenny, Microsoft director of product development for Project Cortex told me, “it depends on your company … if you have a culture of knowledge sharing … the good news is with the capabilities that we have in Syntex, a subject matter expert basically can create an AI model to work on the business process that's important to them.” Having worked with 60 or so customers over the last year, Moneypenny said she has seen a wide spectrum of customers, some who already have a very structured environment around knowledge inside of their organizations, while others are not as prepared.
Related Article: Microsoft's Project Cortex Wants You to Hire a Knowledge Manager
Are Topics Synonymous with Metadata?
The essence of Project Cortex is to automatically classify and connect content by topics so knowledge workers can easily find related information. With Syntex's emphasis on metadata, I asked Moneypenny whether metadata will replace topics as the main construct for classifying information. Moneypenny said that metadata will be important but “it’s not everything that a topic will be, so we will have both options” going forward.
How Much Does Syntex Cost?
According to Holme, the US list price for Syntex is “$5 per user per month and that is on top of either Office 365 E3 or E5 enterprise licensing, so it's available to customers at both levels of our suite and then there are regional and account based discounts.” That adds about 14% to 20% to the price of a user subscription. Holme added that Syntex is now available in all cloud regions and it will be available in government clouds in the future.
Related Article: Microsoft's Flywheel Kicks Into Gear at Ignite 2020
What’s Next for Project Cortex?
Holme and Moneypenny both noted that Syntex is just the first of future capabilities that will be part of the Cortex vision. "We are looking at expertise and competency development inside of an organization because we bring a lot of capabilities ... We've got lots of great capabilities inside of Microsoft Search that we want to build on top of, and then bring some of the newest technology to bear in a new way," said Moneypenny.
Topics are coming. Holme said, "the topic experiences are still coming; those will come before the end of the year." He pointed out that the first appearance of topics is in Search. “When you do a search … the first answer you get is the topic card … which shows key resources. When you click through, you get the topic page, which becomes a summary answer for that search.”
In fact, the concept of topics remains at the core of Project Cortex. Moneypenny said, “our goal is to bring the topic experience into everywhere in Microsoft 365. We won't get there all the way by December, but we will certainly have preview experiences within Teams and Outlook and probably in some of the office applications as well.”
Stay tuned. Because while it will still take considerable time before topics become practical for most organizations, it’s going to get a lot more exciting as new components roll out.