How Engineering Teams Are Reimagining Work Through AI

Forbes

30-07-2025

Bratin Saha, Chief Product and Technology Officer at DigitalOcean, tech executive with 20 years of experience across AI and cloud computing.
AI is one of the most transformative technologies of our time. Given the rapid pace of AI innovation, it seems that practically every leader is thinking about how they can reimagine their work using AI. While this may be daunting and will take persistence, I believe the reward is well worth it.
Having worked on many projects using AI, I want to discuss the factors that are critical to ensuring the success of these projects—from an iterative mindset to new mechanisms and rigor in tracking metrics.
AI For Coding
AI is rapidly emerging as an indispensable tool for software engineers. As teams explore AI-powered coding assistants, it's essential to look beyond 'lines of code generated' and consider impact on code quality, security and long-term maintainability.
Many engineering teams are now experimenting with tools like GitHub Copilot, and we've personally seen up to 40% increases in code generation. But productivity isn't the only metric worth tracking—security, code quality and maintainability are just as critical. One helpful practice is to implement internal evaluation systems that compare AI-generated and human-authored code for defects, rollback frequency and overall impact on velocity. Our initial findings suggest that AI-generated code can match or even outperform human benchmarks in some areas, though consistent monitoring remains essential.
For leaders considering similar integrations, there are a few principles that can help guide responsible adoption. Define baseline metrics early, evaluate AI output with the same rigor applied to human-authored code and build feedback loops to inform ongoing tool selection. Thoughtful experimentation combined with clear evaluation criteria is key to realizing the value of AI without compromising quality or trust.
AI For Rootcausing Cloud Incidents
One way of figuring out where to add AI is to understand where your employees are spending the most time and then automating that activity with AI. Cloud engineers, for instance, typically spend over 20% of their time troubleshooting incidents. Additionally, high availability is critical for customers who rely on cloud services.
AI can play a powerful role in accelerating incident resolution, especially when engineers are spending a significant portion of time on root cause analysis. As an example, we developed a GenAI-powered site reliability engineer (SRE) agent that assists engineers by analyzing real-time logs and telemetry to find root causes autonomously during incidents. Engineers can ask follow-up questions and rate accuracy. By eliminating the need to assemble multiple engineering teams for incident triage and diagnosis, this approach can help reduce the time and effort required to resolve issues and restore the service faster.
Accuracy is one of the crucial measures of the agent's effectiveness, and achieving that was an iterative process. The agent needs to be trained on high-quality, representative data; it needs to be tightly integrated into incident response workflows with real-time access to observability systems; and it must have mechanisms to learn from new incidents and user feedback.
Besides encouraging engineer feedback, one thing we found beneficial is to incorporate the agent into the post-incident review (PIR) process. This retrospective analysis refined the agent's accuracy and functionality by clarifying incident causes and guiding engineers on prompt optimization.
Once the agent meets your predefined success metrics, you can then expand its role. In our case, we extended the agent role beyond reactive incident response. By embedding the agent earlier in the incident lifecycle to monitor system alerts, the agent can automatically assess alerts and propose root cause solutions, significantly reducing investigative time for engineers. These continuous and targeted improvements are key to building a successful GenAI agent.
AI For Server Maintenance
Another way to figure out how to use AI is to consider operations where data analysis can be used to avoid undesirable outcomes and help teams move from a reactive mode (fire fighting) to a proactive mode.
For example, server downtime in a data center is undesirable because it directly impacts the service uptime. Servers usually do not fail out of the blue; there is a pattern of malfunction that can be detected by closely monitoring server health with AI tools.
At DigitalOcean, we use AI to analyze logs in real time, providing a confident root cause to engineers. This analysis can help repair machines faster while reducing repeat outages.
We also collect messages that are emitted by the operating system or the out-of-band management controllers of servers, and perform a rules-based evaluation to trigger the removal of workloads from at-risk machines. If a stick of RAM issues a hardware warning or a disk array degrades in a production hypervisor, AI can automatically migrate customer and internal workloads to healthy machines.
Companies can use similar techniques that use AI to perform a real-time analysis of relevant metrics to drive operational improvements and use predictive rather than reactive operations.
In Conclusion
AI is already changing every aspect of how we work, and it is important for leaders to get in front of it. The most important part is to get started; identify some workflows that are ripe for automation, put together a tiger team and give them the latitude to experiment till they get the AI right. Even if the initial experiments do not work, the learnings are invaluable and set you up for success down the road.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?