Software development is one profession that can achieve a significant productivity boost from the use of generative artificial intelligence (GenAI). GenAI provides measurable gains in developer productivity by performing numerous repetitive tasks, freeing developers to fulfill duties better suited to humans, such as coding.
A company can maximize GenAI benefits by incorporating it into development workflows gradually and cost-effectively, continuously analyzing its performance, and adjusting the human/GenAI division of labor as needed. A boost in developer productivity often results in a substantial improvement in company profitability.
More Time Spent on Non-Coding Tasks
A typical software developer spends a disproportionate amount of time on non-coding tasks, which reduces their productivity. A study of Microsoft developers found that they devote more time to communication and meetings than coding (about 12% versus 11%) and only 9% to debugging code. On the other hand, 20% of these developers wanted to spend more time coding. According to Amazon Web Services, the company’s developers spend just one hour daily on actual coding and the remainder of the time on tasks such as fixing problems and hunting down vulnerabilities.
Software quality and profitability data indicate that performing these non-coding tasks is not suited for human developers. In 2022, the Consortium for Information & Software Quality estimated that the cost of poor software quality in the United States was at least $2.41 trillion, although cybercrime also plays a significant role. Additionally, the accumulated software technical debt (TD) was an estimated $1.52 trillion. As technologies evolve, these figures will continue to grow.
A Dramatic and Measurable Productivity Impact
Time is money. A Google Cloud State of Development Operations study revealed that, compared with low-performing developers, elite developers deliver 127 times shorter lead times, 182 times more deployments per year, an eight times lower change failure rate, and 2,293 times shorter failed deployment recovery times. GenAI tools, like internal chatbots, can improve developers’ productivity by automating repetitive tasks such as:
- Code generation and debugging support. Tools such as GitHub Copilot, Windsurf Editor and Supermaven can autocomplete code, generate entire functions and suggest relevant snippets, saving developers keystrokes and enabling them to code faster. Also, these tools guide fixes by analyzing error logs and accessing historical records of similar bug fixes.
- Code review assistance. Tools like GitHub Copilot and Qodo help to speed up code reviews with intelligent suggestions and automated insights.
- Automated testing. These tools help developers write test cases and check code quality by ensuring that developers follow the company’s established code-writing practices.
- Internal knowledge base access. GenAI chatbots can be trained on a company’s large language model (LLM) to search its proprietary internal documentation of historical coding. These tools provide the foundational coding structure for a current project, drastically reducing the amount of required coding “from scratch.”
According to a study by Microsoft, MIT and other university researchers, the impact of GenAI chatbots on developer productivity can be dramatic and measurable. The researchers found that access to GitHub Copilot increased output — the number of completed weekly tasks — by 26% overall, 27-39% among recent hires and junior developers, and 8-13% among senior developers.
How to Implement GenAI Profitably
The application of six best practices improves the odds of profitable GenAI implementation.
- Trust but verify the data. GenAI chatbots well-trained on a company’s LLM will generally generate accurate foundational code. Still, meticulous human validation provides the necessary quality control.
- Ensure proprietary data security. Establishing data security protocols and monitoring adherence to the protocols can ensure that the proprietary internal documentation used in GenAI models remains proprietary.
- Optimize LLM build costs. Building an internal LLM can be a significant investment. Staying focused on the cost can optimize the investment.
- Implement gradually and adjust as needed. “Crawl” is the appropriate stage in a crawl-walk-run implementation process for gathering feedback from developers on the impact of GenAI on workflows. This valuable feedback can inform adjustments before expanding the use of GenAI in development workflows.
- Make GenAI an optional, rather than mandatory, gatekeeper. Tools such as the Sonar plug-in offer developers the flexibility to use GenAI as an optional quality control gatekeeper. Developers can improve their static code analyses by adding a step in code review or test case writing that incorporates these tools without unduly disrupting their workflows.
- Analyze developers’ time and task allocations. After GenAI tools are incorporated into development workflows, it’s a good idea to evaluate developers’ task allocations. Time allocation key performance indicators (KPIs) include coding, deploying code, debugging and other repetitive tasks. If GenAI is used effectively, coding time should exceed the 11% allocation cited earlier. Eventually, setting realistic GenAI-aided developer productivity KPIs will follow suit.
Potential for More and Better Development Jobs
After benchmarking appropriate time allocations for human developers based on using GenAI in their workflows, organizations can then implement GenAI and focus on continuously improving developer productivity and work quality. With this approach, organizations change how developers do their jobs. At the same time, this evolution can result in higher worker and customer satisfaction, and a worthwhile return on investment. Just as the introduction of the personal computer created many new jobs rather than eliminating them, GenAI has tremendous potential to yield a net gain in business-critical jobs.
