AI coding assistants are everywhere, but are developers really using them?

AI coding assistants (like GitHub Copilot, Cursor, Windsurf, Devin, etc.) have gone from experimental to mainstream. 

Engineering leaders are betting on productivity gains, while the tools themselves are met with both hype and skepticism from the people on the ground. Terms like “vibe coding” are bandied about, as are some sober realizations about how Developers’ jobs will evolve with these new tools.

Behind all this noise, how widespread is the actual use of AI coding assistants among developers today?

I wanted to get the real story here, so I surveyed engineering leaders and Platform teams from dozens of tech companies, ranging from smaller startups to global giants.  I found out more about adoption, as well as the benefits they were chasing and the main things holding them back.

The AI coding tool landscape

Not surprisingly, nearly ALL tech companies already use AI to code

Our data reveals that an impressive 94% of companies surveyed have some teams actively using AI coding assistants. This widespread adoption showcases a strong industry consensus: AI tools are no longer just “coming”; they're here.

‍

What are they trying to get out of AI coding assistants?

These engineering teams were aiming to get a number of benefits out of these AI tools, but increased productivity and time-to-market dominated the answers, with 72% of respondents calling it out explicitly. In a future post, I can break down the various benefits teams are aiming towards, but for now, let’s dive deeper into how well they’re doing at the first step in achieving these goals: getting adoption of these tools.

GitHub Copilot dominates, but it’s not a monopoly

Within these organizations, GitHub Copilot clearly leads, with it being in use at approximately 88% of companies using AI coding tools. Yet, the landscape is diverse: Claude, Cursor, and ChatGPT also have significant footholds, reflecting varied use cases and preferences across teams. There’s a longer tail of other tools in use in more of a minority capacity, including Devin, Windsurf, etc.

Also extremely common was organizations experimenting with multiple tools simultaneously. It was pretty rare to see teams betting everything on one player. On average, organizations were using ~3 AI coding assistants each, showing that there is still a fair bit of experimentation happening across teams and use cases, and that certain tools are better at certain jobs.

One engineering leader I spoke to told me that he was trying to encourage adoption by letting his teams experiment with whatever tools they thought would help them - and so they had a lot of duplication, but that wasn’t a concern. (At least, at the moment.)

Who is actually using these tools?

Broad availability, shallow uptake

While almost every company surveyed has introduced AI coding assistants, only about a third have achieved majority developer adoption. That’s a critical detail here: the tools are present, but still far from widespread use.

Despite the ubiquity of AI tools at a company level, individual developer uptake remains surprisingly shallow:

• Only a third of companies using AI report that 50% or more of their developers regularly use AI assistants.
• A similarly significant percentage reports adoption below 25%.

While this shows encouraging momentum, it also confirms that most organizations are still in the early innings. AI tools are available, but many developers aren’t fully leveraging them yet.

This means that two-thirds of engineering organizations are still missing out on the productivity gains they are targeting.

Adoption lags on big and small teams alike

Interestingly, my survey data didn’t show a strong correlation between company size and adoption of AI coding tools. Larger companies typically face more friction, from legal/security overhead to general inertia, making sweeping tool adoption slower. But on average, smaller companies are seeing a similar lag in their adoption of AI tooling across their engineering organizations.

Small teams CAN get to full adoption sooner

Despite the above, we DID see smaller teams pull ahead in the highest segment of adoption (>75%). That makes sense: it’s easier to get buy-in across a 20-person engineering team than a 2,000-person one. Change can happen more completely when there’s less organizational weight.

Looking ahead

Expansion is not just likely, it’s inevitable

In the survey, I asked about the use of AI tools today and future plans to expand that use. The message from engineering leadership is overwhelmingly clear: AI assistant usage is set to expand significantly.

Well over 90% of respondents confirmed they either "definitely" or "likely" plan to expand their AI assistant footprint soon.

This near-universal intention to expand points to an inevitable future where AI coding tools become deeply embedded in everyday engineering workflows.

So, what is keeping usage low today?

From what we can tell, there is a strong hunger, at least from engineering leadership, for AI coding assistants to gain broader adoption within their organizations and, in turn, realize the productivity gains they’re targeting.

So, what is holding back this adoption? 

I’ll dive into this in my next couple of posts, where I’ll cover the benefits companies expect from these tools, the barriers currently limiting broader adoption, and bring some actionable strategies to boost AI coding assistant usage in your organization. To catch the updates, follow me on LinkedIn.

‍

John Laban is the CEO and co-founder of OpsLevel. After spending several years at Amazon, John joined PagerDuty as the first hire. John spent over a decade building and scaling engineering teams before starting OpsLevel with a mission to help teams move fast while increasing quality and ownership.