Is Github Copilot’s biggest productivity threat bad project management?

The hype around GitHub CoPilot this past week was at an all-time high with the release of GitHub’s first 500-person Copilot impact survey, which outlined real, tangible results from developers’ usage of the tool (see below).

One theme clearly stands out: productivity. It’s no doubt that productivity is the most theorized and intended benefit of the AI boom. But as far as developer productivity goes, is “faster coding” really the answer to efficient development? I propose a challenge – it’s not. Or, at the very least, it’s not the whole picture.

Developers only spend 43% of their time coding

Last October, Zenhub conducted the Developer Happiness Report, which partly examined how developers were spending their time at work. One key finding? Their whole day isn’t coding–that’s just 43% of it. So, what does this mean for CoPilot? It means that, while clearly effective at accelerating coding, CoPilot alone likely won’t put a serious dent in inefficiency. Or, perhaps, even if it does, it has impending limitations. 

Beyond GitHub Copilot: what’s really keeping devs from productivity?

Feedback loops, project planning, tracking, status updating, and anything involving two heads instead of one, slows devs down a lot. That’s not to say they don’t love collaborating–we know they do–but according to the report, 70% of developers are spending at least 3 hours per week just waiting on feedback from colleagues. 

This is a concerning metric for GitHub Copilot, considering the fact that the main reported drawback was that AI-generated code resulted in increased code review time. The takeaway? “Coding faster” doesn’t necessarily mean positive things for the Software Development Lifecycle (SDLC) as a whole. This is why for teams investing in tools like Copilot, it’s more important than ever to pay attention to productivity metrics like Cycle Time and to keep an eye on a Cumulative Flow chart to identify when pipelines like QA are starting to overflow with work. 

AI solutions for developer productivity primarily focus on the tactical aspects of software development, but I believe there are larger factors at play. The issue is, in the SDLC, things tend to fall apart at the seams–that is, the work holding everything together. Think: project management, agile events, and meetings, categorizing and prioritizing tasks, etc.

Because software development is a team sport, trying to get on the same page as your team is one of the biggest time-drainers. In fact, 44% of devs say they spend too much time in meetings, and 38% said they spend too much time on admin tasks (like project management).  

Conversely, most developers stated they “wish they had more time to code.” For most developers, it seems carving out more time to code is the goal of productive ways of working, not being able to code more productively. After all, without enough time to code, how is one ever expected to get the full benefit of Copilot? 

AI project management + GitHub Copilot = superhuman developers?

This is not to discourage you from using CoPilot–quite the opposite–when we gathered GitHub CoPilot reviews for our own research, the productivity benefits were congruent with GitHub’s survey results. What I want to emphasize is the immense untapped opportunity for accelerated software development that engineering leaders are overlooking: AI-driven project management. 

This opportunity not only has the potential to enhance Copilot’s effectiveness but also to improve the overall developer experience (as we discovered in the Happiness Report).

Let’s look at some of the ways I predict AI could be applied to some of the issues we’ve discussed above in the coming years and even months: 

• Reducing time spent in agile meetings through AI-generated retrospectives, stand-ups, story point estimates, and backlog grooming. 
• Improving the categorization of work, e.g., “unplanned work,” “technical debt,” etc., through AI-generated suggestions.
• Reducing time spent drafting status updates through auto-generated summaries and reports.
• Improving a team’s ability to learn from past sprints through AI-generated performance recommendations.

Optimize for more than just 43% of a developer’s day

Copilot really is incredible. But, when developers are only spending 43% of their day coding, it might not make the impact you’re hoping it will. Luckily, AI is also taking project management by storm–Zenhub is starting to roll out some game-changing ways software teams will be able to leverage AI to improve processes and project organization. If you’re as excited as we are, you can signup here to be notified of new AI launches. 

In the meantime, for folks testing out Copilot who want to put their productivity to the test, we also recommend taking advantage of Zenhub’s auto-generated reports, particularly the GitHub Insights report. These should tell you just how impactful Copilot and other AI tools have been on your team’s productivity over time. 

Want more insight into how to manage productivity? Read our guide on Productivity Management for Developers.