The Learning Infrastructure AI Quietly Removed

Insights from the Scaling Business Summit 2026, Ho Chi Minh City.

Most AI conversations in 2025 were about what AI could do. Thijs Van Loon, Founder and Management Consultant at Betterworks Asia, came to SBS to talk about what AI was quietly undoing.

Thijs works with organizations on corporate training, coaching, and learning and development. His session was not about AI capability. It was about the side effect of rapid AI adoption that most leadership teams have not yet named: the systematic erosion of how junior staff learn to think.

The session followed the arc of his framework: Hands, Heads, Hearts, and delivered a practical case for why the companies that invest in human development infrastructure right now will have a structural advantage within three to five years over the ones that do not.

1. You Automated the Task. You Also Automated the Learning.

In December 2004, Thijs was at Google's European office. The day after the Thai tsunami, he was assigned to manually filter Dutch word strings in Gmail's ad system to remove contextually inappropriate ads from appearing in disaster-related emails. Six hours of the most painstaking, tedious work imaginable.

He learned more in those six hours than in weeks of structured training. He learned how Gmail's ad targeting worked from the inside. He learned how the system parsed language. He learned how to manipulate it, improve it, and advise clients about it. The manual labor was the education.

That path is now largely closed. AI handles the tedious work. And in doing so, it handles the incidental learning that the tedious work delivered. “When we implement AI at the speed that we are doing now, the chances are bigger and bigger that we are creating a firewall between our juniors and the rest of our organization.”

The BCG data quantifies the adoption gap already in play:

• 85% of AI users operate at a basic feature level.
• The 10-15% who are heavy users see approximately 10 additional hours of productivity per week.
• HubSpot made AI standard in their engineering onboarding and achieved 46% productivity gains with 98% adoption.

The upside of AI adoption is well-documented. The downside is less discussed.

• Only 12% of junior employees report understanding how AI supports their company's strategy.
• 69% of C-suite leaders feel AI-ready.

The gap between those two numbers is a pipeline problem that is accumulating quietly every quarter. “The pipeline of junior people you can effectively train inside your organization to bring them up, AI stole that example. It took out that deep-digging work that more experienced managers and educators went through.”

Lesson 1: AI productivity gains at the senior level are real. But the same automation that produces those gains is removing the learning infrastructure your juniors depend on to become seniors.

2. Hands: Give Juniors Safe Struggle on Purpose

The first layer of the solution is the most counterintuitive. If AI is removing the natural struggle that used to develop competency, the answer is to reintroduce struggle deliberately. Thijs calls this simulation training, and the mechanism is simple: break the environment on purpose.

Build training scenarios where the tools and setups are intentionally broken. Put juniors in a sandbox where the AI output is wrong, the code has errors, the model has been misconfigured. Require them to diagnose and fix the problem manually, without AI assistance. Not as punishment. As practice.

The goal is to recreate the conditions that used to produce deep system understanding organically: you cannot fix what you do not understand. Juniors who go through this kind of structured struggle develop the ability to evaluate AI output, not just consume it. They learn to ask whether the AI has put the nuts and bolts in the right place, not just whether the output looks plausible.

The practical question Thijs posed to the room: think of one manual simulation task your junior staff could start doing from today. Not a week-long program. One task. The answer to that question is the beginning of the hands layer.

Lesson 2: Junior capability is built through managed struggle, not smooth process. Design broken environments where people learn by fixing, not by watching AI succeed.

3. Heads: Go Slow to Go Fast

The second layer requires something from senior and mid-level leaders that is easy to deprioritize: explaining themselves. Not their decisions. The reasoning behind the decisions. The why that a junior will never extract from an AI output.

Thijs's specific tool is the narrated code review. A senior sits down with a junior, opens the code, and walks through it step by step: why this structure, why this choice at this junction, what would have broken if the decision had been different. The junior does not just receive the output. They receive the thinking.

The effect is twofold. Juniors build the critical thinking they are not developing through AI-assisted work. Seniors develop the habit of making their knowledge transferable rather than hoarding it as tacit expertise. When the senior eventually leaves, the knowledge stays.

About a quarter of the room had a formal coaching or mentoring program in place. Thijs acknowledged why the number is low: mentoring slows you down.

“We need to go slow to go fast. You need to build people's capabilities first before you can go faster in the implementation. If you struggle for one or two months but afterwards more people can pick it up thoroughly, that's when you do well.”

He also named a structural incentive that most organizations have not connected: make the ability to explain your reasoning a promotion criterion. If critical thinking is what differentiates a capable senior from a good AI user, then the promotion track should measure it.

Lesson 3: Senior knowledge does not transfer automatically. Build narrated review and mentoring into your operating rhythm. And measure whether your people can explain the why, not just deliver the output.

4. Hearts: The Gig Worker Feeling Is a Retention Problem

The third layer addresses something that AI adoption has accelerated but did not create: the experience of being peripheral. Juniors working remotely, checking AI outputs at a distance, completing tasks without context about the organization they belong to, many report feeling like gig workers, not employees.

The consequence is predictable. People who do not feel part of an organization do not invest in it. They do not learn beyond their immediate task. They do not stay long enough to become seniors. The pipeline breaks not just because of what AI takes away but because of what disconnection costs.

Thijs's prescription is intentional, customized onboarding. Not just role onboarding, here is your job, here is your tool but cultural onboarding: here is who we are, here is why we do what we do, here is why your presence in this company matters.

“That cultural onboarding where people feel connected to you is what creates long-term motivation and willingness to learn.”

The question he asked the room: what is the one element of your company culture that you would most want a new joiner to understand? The answer to that question is the foundation of your hearts layer.

Lesson 4: People who feel like contractors behave like contractors. Customized cultural onboarding is not a soft initiative, it is the mechanism that converts a hire into someone who wants to build something with you.

5. Build the Academy. Stop Buying the Headcount.

The talent economics are shifting in one direction. Senior hires are expensive and the supply is insufficient. Junior hires are cheaper but require more investment to develop. The companies that have built internal development infrastructure are in an increasingly strong position. The ones relying on lateral senior hiring are facing a market that keeps moving against them.

Thijs's framing was direct: “Experience hiring is dead. Capability relearning and retraining hiring is very much up.” The new hiring criterion is not a CV that proves what someone already knows. It is evidence that someone can unlearn quickly and relearn just as fast. That is the skill that makes an employee durable in a fast-moving environment.

The internal academy model addresses this. Build the pipeline yourself: bring juniors in, run them through structured hands-on development, create internal mobility paths that let them move through the organization as they grow. “Internal mobility is your supply chain. Build it, don't buy it.”

The subsidy landscape reinforces the economics. Vietnam's law on employment provides budget for companies that build qualifying development programs. The EU has committed $50 million to digital and green upskilling. The Philippines' EBIT Act allows organizations to train and receive accreditation credit. Governments are acknowledging what the private sector is slow to accept: the workforce development gap is structural, and organizations are the only ones with the infrastructure to address it at the required scale.

Lesson 5: The companies building internal academies now are acquiring a compounding structural advantage. The ones waiting to hire their way out of the talent gap will find the market more expensive and the talent more scarce every year.

The CEO Execution Playbook: What to Do Tomorrow

1. 1. Name one manual simulation task for your juniors this week. Identify one process where AI currently does the work but the junior never had to understand it. Create a broken version and require them to diagnose it without AI. Start there.
2. 2. Schedule a narrated review this month. Have one senior in your organization walk a junior through a piece of work step by step, explaining the reasoning at each decision point. Measure whether the junior could recreate the reasoning afterward.
3. 3. Audit your onboarding for culture, not just role. Map what a new joiner learns in their first 30 days. How much of it is tool and process? How much of it is who you are and why you do what you do? The ratio tells you whether you are building employees or contractors.
4. 4. Make critical thinking a promotion criterion. Add one assessment to your next promotion cycle: can this person explain the why behind the decisions they make? If the answer is consistently no, the capability gap is already costing you.
5. 5. Check your jurisdiction's upskilling subsidies. Before your next L&D budget conversation, find out what Vietnam's law on employment or relevant regional programs offer for qualifying development programs. The subsidy may partially fund the academy you have been deferring.

→ Watch the full session on YouTube