For years, business units in organizations have been making data-driven decisions (DDDM) — personalized product recommendations, user behavior analysis, credit prediction models, and countless other applications. Honestly, despite always wanting to be data-driven, we’ve been pretty stuck at Kirkpatrick Level 2. In this space, AI has done for learning analytics in corporate L&D departments what COVID did for digital learning. Since the democratization of generative AI, learning development teams have effectively gained a digital analyst on staff — for $20 a month.
Let me speak from personal experience. I work at one of Israel’s largest banks, a recognized leader in data and analytics. But it’s clear that when limited project resources are being allocated and decision-makers are weighing a customer-facing project against a training data analysis initiative — one whose impact isn’t always directly measurable — they make the right call and invest the marginal dollar in the process that touches the customer, improves service, or drives sales. Honestly, I’d probably make the same decision.
Our goal is to create learning interventions that impact business performance. To understand the problems, we need to analyze the numbers and let them tell us the story. For example: is there a gap between the rate at which credit card offers are made and the rate at which those cards are actually activated and used? Where does that gap exist, and what’s driving it?
In a business environment defined by targets and constant competition for employees’ attention, there’s no time to guess or invest in lengthy programs that don’t address the most impactful pain points. That’s done in two steps:
- Data analysis — How many sales opportunities does an employee encounter? How many of those opportunities turn into a pitch? How many pitches advance to a sale? And so on. You get the idea. This can point to gaps in employees’ ability to close deals, but it can also reveal process issues — for example, requests that stall for too long and ultimately fall through.
- Failure analysis — This step is primarily qualitative research: conversations with stakeholders, examination of workflow stages and potential failure points, and identification of perceptual, behavioral, and other human factors.
What Tasks Do We Assign to Our Digital Analyst?
Business Data Analysis
First and foremost, in a business organization — and especially in a bank — significant effort goes into analyzing business data to identify opportunities, failures, conversation quality, and other metrics that help improve customer service and generate additional business. This data is an integral part of our workflow. Whether we’re initiating a project or responding to an incoming request, we examine what the data is telling us to determine where investment will have the greatest impact — in other words, where to put our marginal dollar.
Organizational Learning Data Analysis
The L&D department is responsible for cultural shifts that affect the business — especially in a world where the capabilities we need to develop are constantly evolving. This means we play a critical dual role: responding to immediate capability needs in real time, while also preparing the workforce for the skills that will be required in the medium to long term. The organization’s learning programs must reinforce cultural values of personal and managerial accountability for professional and personal development. We want to see people embracing entrepreneurship, curiosity, and continuous improvement — so we monitor and analyze learning metrics that reflect these behaviors using AI. For example: how much learning is organization-directed versus field-driven? Of the field-driven learning, how much comes from employees versus managers? And what is the weight of coaching and peer learning in the organization’s culture?
Data Analysis and Learning Program Development
As those responsible for designing and executing the organizational learning plan — including course and training development, and professional micro-learning as part of daily workflows — it’s important to us to get the most out of our budget. We track our projects precisely: how long it takes from the initial idea to delivery, cost-benefit metrics, budget allocation, and other indicators that help us ensure we’re executing our development plan effectively. Generative AI makes it easy to build a program management dashboard without spinning up a complex technology project.
In Summary
These are just three basic use cases we’re applying today. I’m confident that as we deepen our interface with these tools, more applications will emerge. But if I had to distill the key takeaway from this section, it’s this: think about the data challenges you’re facing right now that you don’t have a good answer for — and outsource them. Remember, generative AI gives each of us access to an analyst on the team for a relatively modest cost, so there’s no reason not to consult one.
Before the cybersecurity and information security professionals come after me — it’s of course essential to avoid working in environments that aren’t approved by your organization, to use versions that don’t share data for model training, to manage risk carefully, to assess which data is sensitive and which is safe to share with an AI model, to anonymize or replace data with fictional details when necessary, and to follow the other practices we use in our own work.