GitHub took away the license debate. What’s left is the skills gap.

GitHub’s move to usage-based pricing makes investing in engineers the clearest route to agentic engineering.
I am not a developer. For a decade, my focus centered on software as the product—scaling it, selling it, and driving customer value. Today, I am tackling a different challenge: people as the product.
I have always been fascinated by human behavior: why some teams eagerly adopt new workflows while others resist, or why habits persist long after their original purpose expires. Right now, software engineers live that dynamic every day as AI reshapes their roles faster than ever.
Since joining Andela, that human element has framed most of my conversations on enterprise AI adoption—specifically, how people keep pace with evolving tools.
And last month, one specific announcement triggered more of those discussions than almost anything else.
What GitHub actually changed
On Oct. 28, GitHub announced that GitHub Enterprise Cloud customers working with a GitHub or Microsoft account team can now give Copilot to every licensed user, without a separate Copilot seat. You pay for the AI your teams actually use, from the first request.
On paper it’s a pricing update. In practice it changes the question every engineering leader asks.
From “who gets a seat?” to “how well are we using it?”
It’s a bit like moving from a fixed mobile contract to pay-as-you-go. Fixed bundles obscure actual consumption; pay-as-you-go exposes every detail.
Under seat-based licensing, teams debated who deserved access. With platform-wide availability, the focus shifts to ROI.
This transition converts Copilot spend from a fixed overhead cost into a dynamic monthly usage budget, creating an ideal inflection point to align finance and engineering.
And it’s landing just as the frontier speeds up. GitHub’s Octoverse report says more than 40% of pull requests on GitHub started with an agent’s first draft in September, up from 17% in March. As Jay Parikh, Microsoft’s EVP of CoreAI, put it: “Completed work, not more code, is the output that matters.”
Value now follows usage. And usage follows skill.
The gap is human
Every company I work with is at a different stage, depending on industry, size, and growth rate. But one thing is almost universal: adoption across tech teams is uneven.
That is no surprise. People differ in tenure, experience, confidence, and mindset. Most teams have a handful of power users racing ahead and a much larger group still finding their feet.
The metric I hear most often is that 60% to 80% of engineers still use AI mainly for autocomplete and chat, rather than handing real work to agents. The tools are ready; habits are still catching up.
It is not for lack of trying. Hackathons and lunchtime workshops spark interest, but they rarely shift daily habits on their own. And top-down mandates did not last long. (Remember the token leaderboard phase? It tracked who was busy, not who was producing value.)
Take a European professional services firm I have been speaking with. They have a long-tenured team maintaining a legacy codebase. Fittingly, that is where agents can help most—explaining old code and writing the tests nobody had time for. But that only works once developers trust the tools enough to try.
Under seat pricing, this gap sat quietly inside a fixed line in the budget. Now it stands fully visible. GitHub’s impact dashboard shows how deeply teams are adopting Copilot, and Microsoft’s new Pro Badge for GitHub Copilot is earned through real usage rather than an exam. Google’s DORA research calls AI “an amplifier” of whatever is already there.
Forward-thinking leaders use this telemetry as a map to target support, not a scoreboard to track developers.
Why getting it right matters
Many experienced engineers will tell you code is a liability. The more code you maintain, the more production risk you carry.
AI agents can now generate thousands of lines in minutes. While guardrails exist, teams that lack proper enablement face endless review hours and flawed code in production.
That is why the technology leaders I speak with agree that enablement must ground itself in core software engineering principles:
- Breaking work into clear, well-scoped tasks
- Providing agents with precise context and instructions
- Rigorously testing and reviewing generated output
- Enforcing guardrails across code reviews and CI/CD pipelines
- Executing clean handoffs across engineering teams
Mastering these fundamentals enables teams to orchestrate AI agents safely across the entire development lifecycle at scale—a major milestone for any engineering organization.
What actually closes the gap
Many engineering and L&D leaders admit self-directed learning carries teams only so far. Online documentation works for the motivated few. For the majority, shifting habits is difficult, dedicated time rarely materializes between sprints, and watching product demo videos feels like a chore.
Here is what works in practice:
- Train by role and baseline. The most effective organizations map each developer's current capability against specific role requirements. Without that baseline, enablement ends up too complex, too basic, or disconnected from daily work.
- Answer the "so what?" Highlight immediate wins that simplify daily workloads. Allow engineers to adopt agentic workflows on their own terms, starting with tasks they enjoy least: first-pass code reviews, test generation, and documentation.
- Leverage power users. Champion models—where internal enthusiasts guide their peers—build genuine momentum. Ensure every team participates, not just the early adopters.
- Prioritize hands-on execution. Most engineers learn best by building on active projects, seeing the impact on their own workflows in real time.
- Treat enablement as change management. Define clear objectives, track progress, and allocate dedicated time for deliberate practice.
If your codebase does not reside on GitHub Enterprise yet, establishing the underlying platform is step one—unifying your tools, data, and telemetry. Just remember that migration is the baseline, not the finish line.
Better together
At GitHub Universe I was energized to see how much GitHub leans on partners for exactly this part. GitHub builds the platform. Partners help people grow into it.
That’s what we’re focused on at Andela. We believe strongly in hands-on, instructor-led learning built around the real capability of each team and engineer. That includes capstone projects: building something real in their own codebase, so the impact shows up quickly.
We started with a large energy company on a pilot of 20 engineers. Within two months, weekly active AI usage rose from 60% to 95%, and the share of engineers working independently with AI nearly tripled, from 19% to 53%. They’re now rolling it out to 600.
So, who’s using it well?
GitHub has made it easier than ever to put Copilot in every developer's hands. Soon, every team will have access to the same baseline tools. The competitive edge belongs to organizations that invest in their people to master them.
Where does your team sit on that journey today, and what is holding back your next step?
































