Contents

The context

Most companies are already trying

From hackathons and AI coffee talks to internal champions and LinkedIn Learning subscriptions, the majority of enterprises have made some move toward AI adoption. The intent is there. The results often aren't.

‍

According to 2025 MIT research (GenAI Divide: State of AI in Business), roughly 95% of AI pilots fail to deliver measurable P&L impact. The RAND Corporation puts AI initiative failure rates above 80%. The gap between "we have something" and "it's working" is where most organizations are stuck, and it almost always comes down to how capability was built, not whether it was attempted.

"Almost everywhere we went, enterprises were trying to build their own tool. But the data showed purchased solutions delivered more reliable results."
MIT GenAI Divide Research
2025

The decision

Internal programs vs. external specialists

There's no universal answer, but there are patterns. Understanding the real trade-offs helps teams make a more honest decision about where to invest.

Internal training
Building it yourself 		External specialists
Working with experts
Pros

+ Deep domain knowledge: your experts understand your products, processes, and culture

+ Lower upfront cost perception: no external vendor invoice

+ Champions build internal credibility and peer trust 		Pros

+ Purpose-built for teaching: methodology, structure, and habit formation are the core product

+ Practitioner instructors who use AI in production daily, not theory-first trainers

+ Built-in measurement: skill assessments, completion tracking, productivity metrics

+ Scales without adding headcount: 500-person cohorts are operational, not aspirational

+ Fully adapted to your approved toolset: no compliance risk, no approval detours
Cons

− Takes your best people away from core work: the person training is the person not shipping

− Domain experts are rarely training experts: knowing AI and teaching AI are different skills

− Hard to scale: what works for 20 people breaks at 200

− Completion rates typically below 30% for self-paced internal content

− No structured measurement: hard to show ROI to leadership

− Content gets stale fast as AI tools evolve: maintaining it is an ongoing cost 		Cons

− Requires active leadership sponsorship: without it, adoption stalls

− Budget line that needs justification: visible cost vs. hidden internal cost

− Quality varies widely: generic platforms won't deliver the same result as tailored programs
The hidden cost of internal training: When a senior engineer runs an internal AI workshop, the cost isn't zero. It's their time, multiplied by their hourly rate, multiplied by the sessions they need to run to reach the whole team. At scale, that often exceeds external program costs, without the structural benefit of a repeatable curriculum, measurable outcomes, or professional facilitation.

What we hear from teams

The challenge isn't awareness — it's adoption

Across conversations with companies in IT services, fintech, gaming, and manufacturing, the pattern is consistent: internal programs raise awareness but don't change behavior. The missing piece is practical, hands-on learning that connects to real daily tasks.

"Existing self-paced learning through LinkedIn Learning has not been effective because it lacks practical application and 'stickiness.'"
Director of L&D
Acosta Group
"Developers show significant resistance to AI adoption despite our CEO's strong support."
CloudLinux
"CFO and management want to see the efficiency gains. We did all the sessions. But how do we measure them?"
AI Engineer
Xsolla

The practical answer

Not either/or — but with clear roles

The most effective AI upskilling programs we've seen don't choose between internal and external, they combine them intentionally. Internal champions provide context, culture fit, and peer trust. External specialists provide the curriculum backbone, facilitation expertise, and measurement infrastructure that makes the program scalable and provable.

‍

The goal isn't to replace your internal knowledge. It's to make it teachable.

What this looks like for real teams

Two examples from companies that moved from internal-only programs to a structured, specialist-led model.

inDrive

From resistance to measurable productivity gains

Exness

From skepticism to a company-wide learning program

In their own words

"The tools presented in this chapter are amazing. They cater to different levels of learners, and the exercises are clearly stated. This is a very well-researched chapter."
"The course module was clear and interesting. It helped me understand the examples and learned how to use ChatGPT more carefully."
"Overall, I found the course module very informative and well-structured. The content was clear, engaging, and easy to follow. Great structuring from basics to advance."
"Most interesting module so far! Very useful tools and an actual demo at the end of each module."
"The lesson was clear and well-structured, with helpful explanations and examples. Lots of examples is a really good thing!"

How Nebius Academy approaches this

We build programs around your approved tools, your team's actual workflows, and your existing skill baseline, not a generic catalog. Our instructors are practitioners who work with AI in production. Every program includes a before/after assessment so you can show leadership what changed.

‍