I’ve been thinking a lot lately about the economics of learning. Not the theory, but the actual reason most organizations still default to the same format they’ve been using since the 1970s. A recent piece by Dr. Philippa Hardman put it in sharp focus, and I think every accountant and finance professional should be paying attention to where this is heading.
The corporate training course has been the default unit of professional learning for 50 years. Not because it’s the best tool, but because it was the only thing that scaled. It solved three problems neatly: auditability, repeatability, and a visible response to “we need training on this.” Nobody could argue with a 92% completion rate on a compliance module.
The problem is the data has been uncomfortable for decades. We just had no economically viable alternative.
What the Research Has Always Said
Lombardo and Eichinger’s 70:20:10 research from 1996 found that people attribute roughly 10% of their professional development to formal learning. The other 90% comes from doing real work, stretch assignments, and peer mentoring. That’s a 30 year old finding.
Baldwin and Ford’s transfer-of-training research from 1988 is even harder to ignore: only 10–20% of formal training results in sustained on-the-job behavior change. That means for every dollar spent on a corporate course, roughly 80–90 cents produces no lasting change in how people actually work.
And here’s the compliance paradox that really drives it home: internal studies at large tech firms have found that employees who passed compliance assessments at 90%+ were still only following about one-third of role-critical protocols in live environments. The course taught people how to pass the course. That’s a different skill than doing the job.
Despite all of this, organizations globally spend over $400 billion a year on corporate training. And 74% of senior leaders still say their companies lack the skills they need to compete. Spend has grown. Confidence in actual skills has not.
Why AI Is the Forcing Function
Three things are shifting simultaneously, and AI is at the center of all three.
First: the delegation test. In 2025, agentic AI systems started completing asynchronous e-learning courses from start to finish. Clicking through content, passing quizzes, and earning certificates with no human learner in the loop. If a bot can finish your compliance training and earn the badge, you have to ask: was that ever really a learning experience, or was it record-keeping dressed up as learning?
Second: diagnosis got cheap. Historically, figuring out whether a performance problem was actually a learning problem required surveys, interviews, and political capital. Most of the time, it wasn’t a learning problem. It was a process or clarity issue. But diagnosis was expensive, so a course got built anyway. AI compresses that diagnostic step, making it possible to continuously surface where gaps actually live without a quarterly survey cycle.
Third: the better alternatives now scale. Worked examples, job aids, spaced repetition, coaching conversations, and feedback on real work were always pedagogically stronger than click-through courses. The blocker was economics. You couldn’t build and maintain them across dozens of roles at scale.
A Real-World Example Worth Knowing
One concrete case from Hardman’s research stands out for anyone in a regulated industry. A global charity replaced its 60–90 minute annual compliance course with weekly short-form scenario modules — “Compliance Bites” — delivered directly in employees’ normal work channels. The result: audit outcomes were maintained or improved, with meaningfully less time pulled away from actual work.
That’s the principle of spaced repetition in practice. Rather than one high-stakes course event per year, learning is distributed across smaller, more frequent touchpoints — timed to when knowledge is actually needed, not scheduled for administrative convenience. The same pattern is showing up across regulated industries, including financial services.
What It Means for Accountants and Finance Professionals
CPE compliance frameworks are the most directly exposed. The vast majority of CPE credit hours are earned through the exact model under pressure — click through content, pass the quiz, earn the credit. NASBA and state board standards are built around seat time and structured curricula. The regulatory framework hasn’t caught up yet, but the pressure is building.
For firm leaders, Bersin’s latest research identifies roughly 5% of companies running AI-native, workflow-embedded learning programs. Those firms are 28 times more likely to unlock employee potential and 6 times more likely to exceed financial targets than peers still running static course models. That’s a gap worth paying attention to.
For individual professionals, how you choose to fulfill your CPE requirements is a learning architecture decision. The research consistently shows that how learning is distributed over time matters as much as the total volume. Frequent, short, timely exposure produces better retention than the same hours consumed in a single sitting — even if the credit totals are identical.
Key Takeaways
- The corporate course survived 50 years as the default because it solved scalability and auditability — not because it was the most effective way to build capability.
- Only 10–20% of formal training changes on-the-job behavior. AI is now removing the economic excuses for ignoring that.
- The delegation test: if an AI agent can complete your training and pass, it was probably record-keeping, not learning.
- Microlearning and nudge-based systems — short lessons, frequent delivery, timed to real work — are backed by spaced repetition science and already producing measurably better compliance outcomes.
- CPE frameworks will face pressure to evolve. In the meantime, how learning is structured over time matters — not just the credit total.
Want the CPE credit? Take the full lesson on EverydayCPE and earn 0.2 CPE credits: [lesson link]
Leave a Reply