Artificial intelligence (AI) is now part of most every organization. When you look at your organization’s strategic roadmap for AI, you should have a clear sense for how the technology can be used to help you leverage your valuable content. The right AI-infused content management system can bridge the gap between highly trained data scientists and non-technical business users, making it one of the most effective tools possible for content analysis.
To optimize this opportunity, it’s not enough to simply check the box as far as incorporating AI into your organization’s informational workflows. Instead, your systems must understand content and its associated data as well as a knowledgeable human – and human equivalency requires accuracy above 85-90%. What’s more, they need to do so at scale.
Indeed, AI-enabled content management technologies – most notably, content services platforms (CSPs) — offer the ability to operate at extreme scale. These technologies also offer extremely short training times. Take, for example, recent proofs-of-concept (POCs) that we have conducted for product identification, talent identification (which models or celebrities appear in photos and images) and even cross-product identification. With a relatively small training set – even fewer than 2,000 images with corresponding metadata – the right CSP can train a machine learning (ML) model that delivers 95%+ accuracy. Even better, these models can be trained incredibly quickly. In just a few short hours, we’ve been able to train models that produce better than human-level accuracy.
Think about how long it would take to train a human to accurately identify hundreds or even thousands of different products as well as dozens different models. Now think about how long it would take a human to go through thousands of images to pick out specific products, talent, etc. And machines can do this in just hours. This is the real promise of AI.
In addition to delivering powerful, business-specific outcomes, your AI-powered CSP must also be enterprise-ready. This means delivering a cloud service that will scale and perform as large, enterprise organizations need it to. On another level, “enterprise ready” also means enabling proper governance for AI services.
The right enterprise AI tool can distinguish machine-generated data from human-generated data. It must version and save each “content bot,” as well as store training data sets so that organizations can demonstrate how a particular model was trained and why it is producing the results it produces.
Fundamentally, an enterprise AI offering should be fail-proof and fully auditable. If a bot becomes corrupted, or begins to show bias, or if the performance of a particular model begins to degrade over time, you should have the ability to not only roll back to a previous version of the model, but also to roll back all of the data values the corrupted model produced. And, should a regulator or auditor ever inquire as to the source of data or how a model was trained, an enterprise AI service should give you the tools to provide an informed response.
Bridging the gap
We mentioned earlier the value of bridging the gap between data scientists and business users. It’s important to recognize that not all organizations are equally savvy with AI, so it’s helpful to have tools that makes it easy for them to both train and administer their own custom ML models. By employing a “point-and-click” or wizard-driven experience, even casual business users can configure and train new models.
Dashboards also help enable less-technical users, at a glance, to determine how their models are performing over time. This way, they can quickly and easily identify models that either aren’t working or are beginning to degrade. Related intelligence that guides less-technical users through the process of identifying the right training set will also help produce the results they seek. After all, machine-learning models can’t guess, and they can’t deliver values that they haven’t been properly trained to produce.
Some AI services have placed the power of fully trained machine-learning models in the hands of enterprise users. By “bridging the gap” between data scientists and business users, these services make it easier than ever for organizations to fully leverage their content as critical corporate information.
Combining with workflow services
AI and workflow technologies are converging in valuable ways. Think about the way in which information gets ingested and subsequently consumed in a business process. Many processes are initiated by the receipt of new content — perhaps, in this example, an insurance application for underwriting. The first thing we may need to determine is whether the form has been completed correctly, with the necessary information and requisite signatures. Then, the form needs to be processed and the customer data needs to be extracted and passed on to other key business systems to initiate the underwriting process. Content also facilitates the workflow. Perhaps, to extend the example, the workflow has been suspended, awaiting the results of a medical exam. Again, once the proper medical documentation has been received and processed, then the workflow can proceed and, ultimately, a new policy can be issued. This is the value that AI can add to a workflow, automatically ingesting critical new information into the process to enable proper decision making.
AI can also help to automate how information is delivered to a knowledge worker. Take case management, for example — assembling a correlated set of content and data before assigning it to a work performer for a decision or resolution. Over time, AI could even empower us to make certain decisions automatically, further automating workflows. We can also perform intelligent exception management, identifying problematic cases or work items early in the process and automatically rerouting them for resolution by a knowledge worker.
And even further out, we envision deep-learning-driven processes that can make automated decisions. Think about straight-through processing, or even workflows that dynamically adjust to changing business conditions. As a simple example, if a specific type of work item is routinely delegated by a work performer (perhaps to a specialist for processing), we can make this association, amend the workflow, and then automatically reroute this type of work item in the future, all without human involvement. The objective here is simple: workflows that “learn” over time to increase process efficiency and reduce costs.
AI should be more than just a helpful new part of your enterprise, it should be front and center in your organization’s informational workflows. Indeed, the first step in smarter workflows is understanding how AI can optimize your content. Today’s modern CSPs can train an ML model extremely quickly, and at scale to produce near-instantaneous results, and with comparable or greater accuracy than a human. Think about what that means for productivity, innovation and the future of work.
The disconnects that occur between your data science teams and other less technical members of your organization no longer threaten progress toward organizational objectives — they are now being bridged with intelligent, enterprise AI services. That’s what smarter workflows are about, and they start with content and AI.