Bringing performance support and learning into the workflow is a much-discussed concept, as it aims to improve the efficiency and relevance of learning, ideally in the moment of need, and to boost employees’ performance. This fits into our series on delivering content to workers and learners as a means of boosting their performance and ensuring that training directly relates to specific business goals and outcomes.

Our series launched with L&D Must Resolve 'Content Chaos' to Meet Performance Goals, an overview of the biggest challenges organizational development and L&D face right now.

We continued the series with Can Skills Frameworks Close the Skills-to-Performance Gap?, where we took a closer look at the challenges and potential values and limits of skills and competency frameworks, including some practical advice and insights gained from working with a large variety of small, medium, and enterprise businesses.

In this article we dive into the role of AI in improving the accessibility of information and answers in the moment of need—in the “flow of work.”

Workflow learning: What’s old is new again


We may think “workflow learning” is a relatively new concept. But we must think again! Workflow learning has a rich history, and its origins can be traced back to the 1950s & 60s.

As part of our research we dug into this a little deeper, with the kind support of Guy W. Wallace, Roger Addison, Paul Harmon, and Lindsay Robinson.

In the early days—the 1960s—practitioners used flowcharts to analyze performance problems and help decide if the worker required a job aid. Tom Gilbert published early work as part of “Mathetics”; Mager and Pipe later used this concept in their guide on analyzing instructional problems.

In the early 1970s, Barry Boothe of Caterpillar spoke about job aids at the Detroit NSPI Chapter. Frequency of task and memory were the deciding factors: Job aids were less necessary for jobs done frequently, as it is easier and less expensive to provide training for these tasks.

Geary Rummler and Tom Gilbert then worked with Marty Mench to develop Programmed Instruction methods (information presented in controlled steps), with a focus on instruction and memorization of tasks. PI appeared in the late 1970s and early 80s, with the advent of computers and computer-generated flowcharts. Early pioneers Susan Markle, Stewart Margolis, Jim Evans, Donald Tosti, and Roger Kaufman used PI to begin looking at the performance of individuals at the team and organizational level.

In 1991 Gloria Gery published her book about EPSS (Electronic Performance Support Systems). She highlighted the dilemma we still face today, the lack of impact of formal learning on performance. She viewed performance support as “an integrated electronic environment incorporating knowledge, task support, data, tools and the opportunity to communicate, which enables people to learn while they conduct complex tasks in the workplace, with minimal support or solutions by others.”

Next on the scene, Jay Cross envisaged Workflow Learning in 2004 as a way to optimize “smart” software that would guide workers to do their jobs better by putting them in touch with the right information or expert. Cross likened it to service-oriented architecture.

His inspiration was the IDC white paper, “The Hidden Costs of Information Work,” citing the amount of time knowledge workers wasted looking for information and finding the right people.

And finally, we have Learning in the Flow of Work—the approach that makes “learning” accessible without disrupting normal working, with relevant content delivered in the moment of need. This drove the move toward content curation and learning libraries. You can listen to a CIPD podcast to learn more.

But is it learning?

While we might perceive workflow learning to be an aspiration, we argue against this being labeled as “learning” per se. Why? Because in practice its use is as a resource for reference and lookup at the point of need. Rather than providing someone with a full encyclopedia of information, we establish exactly what it is they need to know, right there and then.

As an example, let’s imagine an employee in software development. Yes, specialists might learn how to use a particular coding language through formal learning, and they might even continue to attend formal training and courses, generally.

But when involved in a project, how they make decisions, what to implement, how to implement it—all of this would be drawn from their existing knowledge and experience, as well as project-specific knowledge.

At the moment of need when they stumble across a challenge, it can likely be resolved by finding the answer to one, or just a few, very specific questions, in that moment. This is a challenge of accessing information and answers, not one of “training” or formal learning.

But how do we support performance in the moment of need? And how can technology support us in our endeavor to overcome this challenge?

Supporting performance on the job, in the moment of need

One of the biggest challenges for organizations right now—even more so for enterprise organizations—comes from the amount of information and content available across knowledge hubs, sharepoints, and LRS, LMS, and LXP systems, which encompass all resources, often from a range of providers. How can this information be used to full effect?

We know that hidden within the gigantic collection of available content are countless gems and nuggets that could provide answers to specific questions in the moment of need.

These “nuggets” therefore have immense potential value for employees, and thus for an organization’s performance and success. In addition, these resources will have come at some initial cost or be part of a paid-for subscription; this money and the time that have already been spent need to be utilized as effectively and efficiently as possible.

While this challenge is often coined “content overload” or “content chaos,” the good news is that all of this information and all these resources do exist in the first place. That is, if the person in need were indeed able to find the right information or resource, at the right time.

In our view, Lew Platt, former Hewlett-Packard CEO, famously summarized this challenge perfectly: “If HP knew what HP knows, we would be three times as profitable.”

Of course, Platt knew that all the needed expertise, knowledge, and skill are quite possibly already in existence within an organization—somewhere. If there only were a way to tap into all of this.

Co-author Markus Bernhardt’s recent article in Forbes, Organizational Development Content Overload: The Importance Of Finding The Right Tools, identifies two criteria that must be met in order for organizations and employees to stay on top of, as well as tap into, this wealth of knowledge and expertise:

  • Firstly, all of this content needs to be searchable, effectively and fast.
  • Secondly, all of this content needs to be organized, i.e., tagged and mapped, such that it can be managed and updated, as well as outdated content easily removed.

AI enables us to efficiently tap into vast amounts of content

Recent advances in AI technology now allow us to do both.

Contextualization engines and language models on the one hand allow us to organize, manage, and update huge volumes of digital content, at scale, while improved search functionality through AI allows us to search and find exactly what we are looking for—from paragraphs in documents right through to specific sections within audio and video resources.

Utilizing language models, searches in the moment of need may now be answered very specifically, with responses varying based on the complexity of the question.

How does this work? In the simplest of terms, we just need to imagine how differently an expert and a novice would word a question about a specific topic. An expert would be far more precise in their request and utilize more advanced terminology within their question, whereas a novice would word their questions in very basic and simple terms.

With such tools in place, we can ensure maximizing the value of information and existing digital assets. This huge step forward in terms of performance support is available 24/7, from almost any device.

But what about human interaction and support from colleagues?

Finding the experts

The experts hold greater tacit knowledge than our vast digital content libraries, but how do we tap into that knowledge?

Ask the question: “What do you find most useful to do your job?” and invariably, the most popular answer is “a colleague.” Yet we inadvertently add complexity to the search for the right expert.

Here’s how. Typically, we see in many organizations an ecosystem that includes Microsoft Teams or similar; an intranet; the likes of Yammer; and even a social element in the LMS/LXP. Emerging “Talent Marketplaces” may add yet another place where one can potentially find experts. While useful, we again find ourselves with a tech stack that needs keeping up to date and with different places for different things.

Not ideal.

However, across these ecosystems, independent of platform, portal, or app, improved search functionality utilizing AI will also bring SMEs and those requiring support closer together.

How would this work? Imagine, for example, user-generated content such as videos, previously recorded and stored somewhere on the system, sharing how to approach a specific challenge or do a certain task. It will be possible to find an employee-generated video via search—without its having been tagged or even given a title—simply due to its content, super-fast and agile.

Surfacing these assets is beneficial in its own right; however, it also enables employees seeking support and assistance to directly contact the author or relevant team within their organization that produced the asset.

From unmanageable … to unimaginable

We feel this is a milestone moment in an organization’s capacity and capability to optimize performance. Ever-growing content libraries and digital asset depositories, whether designed with learning in mind or not, have posed a challenge for many years, growing much faster in size and complexity than the tools have improved to help us stay on top of all of this content.

Filtering and sorting algorithms have provided little respite, and for large organizations their impact fades like a drop in the ocean. With the advent of powerful and agile AI engines and tools that allow us to surface content and answer questions, as well as organize and map content effectively, efficiently, and at scale, we are finally in a position to utilize digital assets and deliver answers and nuggets in the moment of need. What was once an unmanageable, partly out-of-date heap of data can now be organized, mapped, and utilized, to hitherto unimaginable effect and impact.

In our next piece, we will shift our focus to formal learning. We will explore how the future of performance support and learning will be shaped by an elegant combination of truly adaptive digital learning and in-person performance improvement and development.

Learn more

Dive into new paradigms and challenges at the winter Learning Leaders Online Forum March 15 & 16. Register today, and learn from the experts, network with learning leadership peers, and explore emerging issues. Our all-new format includes networking opportunities, micro-master classes, a collaborative case study—and much more. Don’t miss Markus Bernhardt’s opening session, “The Role of AI in the Future of L&D.”

The Online Forum is free to members of the Learning Leaders Alliance, a vendor-neutral global community for learning leaders who want to stay ahead of the curve. Join us to meet fellow leaders and aspiring leaders, build your skill set, and delve into trends and topics affecting your work.