How UI and UX design can bridge the gap to an AI future in content management
April 23 2019
by Conner Forrest
No matter what segment of enterprise technology you focus on, there is likely to be a vendor working on some sort of machine learning (ML) or artificial intelligence (AI) solution for that space. This is especially true within workforce productivity products – business leaders want to improve employees' ability to add value. We believe that content management stands to gain dramatic usability improvements from ML and AI investments, eventually leading to an intelligent workspace that serves up contextual content and relevant tasks for each employee. However, our ML- and AI-powered content future is still some ways off. In the interim, 451 Research believes that content management vendors should focus on 'dumb' manual automation enablement and adding context through enhanced user-interface and -experience (UI and UX) design, especially during initial service implementation and setup.
The 451 Take
Given the ubiquity of knowledge work that revolves around enterprise content and digital assets, the potential impact of AI and ML on content management is great, but it is far from a reality at this point. In the meantime, content management vendors should focus on redesigning their UI and UX to facilitate contextual work, automated workflows and content relevancy. This should lead to productivity gains in the short term, help users determine which processes need to be culled and help prove the value of automation for future investments in ML and AI (when they are enterprise-ready).
Why don't we just jump straight to AI- and ML-powered content management? It's complicated. The technologies are there (conceptually and technically), but there are several other hurdles that need to be overcome before AI and ML can deliver on their promises. The most glaring of these is metadata.
Most of the traditional methods for adding metadata or tags to a given piece of content are manual, leading to challenges typically tied to human error, like incomplete metadata or taxonomy issues. Some vendors have begun automating parts of the process, but it is often difficult to do so in a very robust way with a broad identification tree. The irony of this is that automating the generation of metadata is a perfect use case for machine learning, but without an initial data set to go off of, it cannot be done at scale. Metadata is the ultimate chicken-and-egg problem for ML in content management, and until it is solved, will hold the space back from its embrace of AI technologies.
Another challenge is the misunderstanding of what ML and AI are actually capable of. Modern media portrayals of artificially intelligent systems have many knowledge workers believing that they will have a digital assistant that will act as a technological secretary on steroids to help them stay on top of their work. What we have come to realize is that most of the work done by AI and ML will be happening in the background, and the productivity gains, while potentially significant, might not be immediately felt by all users.
Another misconception is that AI and ML should be feared because they will displace anyone that uses these technologies from their job. Automation has been eliminating certain positions for decades, but the risk is much lower for knowledge workers in content-centric roles. What will likely happen is that these technologies will be working as a complement to support the work of the human employee, not a replacement.
Technologies like AI and ML have the potential to help users draw conclusions from connections made between disparate data sources that reside in different systems. For content management, this will eventually mean a seamless integration of communication, sales, marketing and HR data with asset creation and asset management tools. Today, though, disparate systems hold ML tools back from making those connections. Content management vendors need deeper partner integrations and stronger partner ecosystems so that the true value of these technologies can be realized.
Despite the challenges that ML and AI face in content management, there are still opportunities for vendors to embrace automation and contextualization in various ways – for example, 'dumb' automation, or automation that is implemented with the help of manual processes and human choices. In many use cases, ML automates parts of a workflow based on patterns it notices in a given dataset. However, even without true machine learning, individual knowledge workers should be able to automate aspects of their daily work and make their content experience more relevant.
One opportunity for dumb automation is through vendor approaches to UI and UX, especially when first onboarding a new user. Many of us are familiar with the setup wizard interface of old that used to walk us through the installation of legacy software. Within that paradigm, there are lessons that content management vendors can learn to help them create a more meaningful product.
As a new user is onboarded, one of the first things they could do would be to select their role, a simple act that could – in conjunction with a series of 'if, then' processes – immediately add more context to their content management experience based on what content that role is primarily focused on. For example, if a user identified as a member of a legal team, that could alert the content management system to automatically add weight to PDF files and files with 'proposal' or 'contract' in the name when a search is conducted.
Alternatively, users could even be prompted to list their most-used file types or work types as a means of personalizing their content search and management. Much like how email providers such as Google Gmail and Microsoft Outlook allow for filters to be created for specific types of messages, knowledge workers need a filtering system for their content management tools to create tags for high-value role-specific, project-specific or department-specific content. Users could be walked through the creation of these filters during onboarding or could be provided a separate setup wizard to create or modify them later. Some vendors offer similar functionality, but it's not as obvious of a feature as it needs to be.
Content management vendors – at least those with a workspace or content launcher product – can also learn from the advertising industry. Far too often, folks are served with ads that, upon hiding, task users with explaining why they hid them by clicking an option such as 'this is not relevant to me.' On a much smaller scale, that type of experience can be applied to the content management space to, once again, automate processes and improve context. Imagine end users uploading a certain file to a shared folder for a marketing project. By asking the user if there is any other content to be uploaded to the folder, a dumb automation process could be then enacted for future content. The user could say that yes, there is additional content to be shared, prompting the following dialogue:
Current content, future content or both? Future.
Okay, sure. File with names containing what keyword? MarketingProjectX.
Okay. All files with MarketingProjectX in the file name will automatically be uploaded to selected folder.
This is a gross oversimplification of the engineering that must go into a content management product to enable such a feature. The point of this is merely to illustrate that there is a lot of value left on the table in current content management deployments that could be leveraged to impact business outcomes.
What's the point of all this dumb automation? Why don't companies just wait until ML and AI tools are ready for primetime? For starters, you need to ease your end users into the idea of automation and more contextual work.
Despite the lauding of those in the tech industry, ML and AI are still intimidating concepts for many front-line knowledge workers. Dumb automation and manual workflow triggers offer a more comfortable on-ramp to full ML-powered automation because it gives the user more control over the automation. Once comfort is established, transitioning a userbase to a fully AI-enabled solution won't be as daunting (for the users or the admins).
Dumb automation is still automation, and successful use cases help prove the value of automation as a whole. This earns buy-in from end users, as noted, but also has the potential to earn buy-in from key decision-makers, especially if the implementor can get anecdotal and quantitative data on potential productivity improvements. Once again, this greases the wheels for an ML/AI future and makes it easier to get the go-ahead for a SaaS upgrade once AI and ML content management tools are production-ready.
Productivity impacts the bottom line in a variety of ways, and productivity-focused tools can make companies more attractive to potential job candidates. 451 Research asked the respondents to our September 2018 Corporate Software Survey the following question: Imagine you were in the job market looking for a new employer. How important would it be for a new employer to offer devices, applications and other productivity tools to help get your work done?
Of the 1,100-plus people surveyed, 43% said it would be very important and 38% said it would be somewhat important. That is 81% of our respondents that believe productivity tools are potential differentiators when job hunting.
AI and ML show great promise in content management, and will likely be game-changing technologies in terms of their impact on the space. With that said, 451 Research believes there is a greenfield opportunity for content management vendors to improve productivity and efficiency within their tools by leveraging dumb automation through available tools and UI/UX improvements. These vendors can help their customers get more work done now and more easily bridge the gap to our AI and ML content management future.