The New Developer Adoption Gap
Tom Williams, CEO
AI has dramatically increased the speed at which software can be created. But rapid acceleration tends to produce pressure elsewhere. The challenge is no longer simply producing more software, but ensuring that what gets built can be validated, maintained, understood, and successfully adopted by end users.
Two recent reports show different sides of the same problem. CircleCI’s 2026 State of Software Delivery examines more than 28 million CI workflows, while Instruqt’s 2026 State of Developer Adoption looks at how products are marketed and sold to developers. One focuses on the delivery pipeline; the other on what happens once products reach the market. Together, they point to the next major AI bottleneck: turning what we build into something that can be shipped, maintained, understood, and adopted.
More code is not the same as more progress
CircleCI found that average development throughput rose 59% year over year in 2026, which sounds like the productivity dividend companies were promised. But the gains are uneven: the top 5% of teams nearly doubled throughput, the median team improved just 4%, and the bottom quarter saw no measurable gain. AI amplifies teams with strong delivery systems already in place; it doesn’t lift everyone equally.
More importantly, much of that activity isn’t reaching production. Median-team throughput on feature branches rose 15%, while main-branch activity fell nearly 7%. Developers are generating more code within their organizations, but not necessarily shipping more of it.
The constraint has moved downstream: Code still needs review, testing, integration, and maintenance; main-branch workflows now fail almost 30% of the time, with a 72-minute average recovery. AI can create more change. It can’t ensure an organization is ready to absorb it.
The same bottleneck appears after launch
CircleCI found that average development throughput rose 59% year over year in 2026, which sounds like the productivity dividend companies were promised. But the gains are uneven: the top 5% of teams nearly doubled throughput, the median team improved just 4%, and the bottom quarter saw no measurable gain. AI amplifies teams with strong delivery systems already in place; it doesn’t lift everyone equally.
More importantly, much of that activity isn’t reaching production. Median-team throughput on
feature branches rose 15%, while main-branch activity fell nearly 7%. Developers are generating more code without their organizations, but not necessarily shipping more of it.
The constraint has moved downstream: Code still needs review, testing, integration, and maintenance; main-branch workflows now fail almost 30% of the time, with a 72-minute average recovery. AI can create more change. It can’t ensure an organization is ready to absorb it.
The same bottleneck appears after launch
Instruqt’s research shows the same pattern from the other direction: companies can produce new products and features faster than developers can adopt them. More than half of practitioners say full adoption takes one to three months; almost a quarter say it takes longer. Among smaller vendors, 65% say developers need more than two months.
This is not developers being slow. Adoption depends on whether they can find a product, understand its value, prove it works, integrate it safely, and maintain it. None of that gets easier simply because the product shipped faster. In fact, faster development can make adoption harder: more features mean more documentation, more use cases, and more decisions to navigate. With 92% of practitioners reporting at least one significant adoption challenge, most commonly team misalignment, the barrier is not activity but coherence.
Documentation is not the whole experience
Documentation matters, but Instruqt found a gap between what companies use most and what actually works. Written docs are the most common adoption practice, yet rank below community spaces, proofs of concept, demos, and interactive labs in perceived effectiveness.
Documentation tells a developer what they could do; a hands-on experience proves they can do it. Among organizations using hands-on labs, 63% saw developers productive within two months, versus 42% without them, with the strongest labs mirroring production environments and connecting to real tools and APIs. Developers need a credible route in, not just more content about the destination.
Developer experience is one system
Engineering builds. Marketing creates demand. DevRel engages community. Education trains. Sales demonstrates. Customer success supports adoption. But the developer experiences one journey, and a failure anywhere can break it: a great feature nobody can find, a strong promise lacking a realistic environment in which to test it, or a smooth onboarding that falls apart at integration.
The full system must support discovery, evaluation, experimentation, integration, maintenance, and continued learning, while connecting the teams responsible for each stage.
Instruqt’s data shows how easily these teams fragment. Marketing rates communities highest, sales rates demos and labs, education rates guided learning. Each view is reasonable. The problem starts when every function optimizes its own piece without agreeing on the outcome.
AI raises the value of the surrounding system
There’s a temptation to answer the AI adoption gap with more AI: faster docs, a chatbot,
an in-product copilot, an automated learning journey. Some of this will help, but the lesson from both reports is that acceleration only creates value when the surrounding system can keep up.
For software delivery, that means stronger validation, testing, integration, and recovery. For developer adoption, it means clearer positioning, realistic evaluation, structured onboarding, and a deliberate path to first value. AI makes execution faster. It doesn’t remove the need for structure; instead, it increases it in many cases.
The competitive advantage is absorption
The first phase of the AI race was about access: who had the tools, who adopted copilots, and who could generate more code or content. That advantage is already becoming less distinctive.
The next advantage is absorption: turning generated code into stable, maintainable software, and faster release cycles into products developers can actually understand, adopt, and sustain. The winners won’t be the companies producing the most, but those with the strongest systems for helping people ship, maintain, understand, and adopt what they produce.
What Developer Marketing Teams Need to Do Now
Closing this gap means treating absorption, not just output, as the thing you’re optimizing for. In practice, it means:
Audit where work stalls after code ships. Look at review, testing, integration, and maintenance before adding more AI-generation capacity.
Invest in hands-on proof over more documentation. Build production-like sandboxes connected to real tools and APIs; the data shows this can cut time-to-productivity by a third.
Unify the journey across functions. Align marketing, DevRel, education, sales, and customer success around one adoption outcome, rather than separate metrics for separate stages.
Measure adoption speed, not content volume. Track time-to-first-value and time-to-full adoption as core KPIs, not just docs published or demos run.
Point AI at the bottleneck, not just the output. Use it to strengthen validation, onboarding, and support systems, not only to generate more code or content faster.
AI has accelerated the supply of software. The winners will be those who make more of it usable.
For more information read the full Developer Marketing Insights Report.
