Changing the Culture: Shifting from “Static Implementation” to “Continuous Discovery”
To thrive with AI, leaders must foster continuous discovery, robust evaluation, and agile experimentation, moving away from static processes and empowering teams to adapt quickly in an ever-changing landscape.
Read the previous blog on this topic.
Durga Charan Panda
Continuous Discovery
Getting the right technology is only the beginning. As a leader, one of the toughest challenges with AI is helping your team adjust to new ways of working. To kick off this shift, schedule a short discussion this week with your engineering leads about how to integrate regular AI feature exploration into team routines. In addition, identify one current project in which you can pilot expanded evaluation processes or sandbox testing practices.
Transforming culture is an ongoing journey, not a checklist.
Focus on these three key team cultural areas:
Shifting Team Mindsets
In the past, enterprise teams would finish their architecture and consider the code done. With AI, teams need to view their architecture as always changing. Encourage engineers to regularly try out new features in AI libraries. Managers can support this by recognizing team members who look into and share new AI library features. This could be a shout-out in meetings or an opportunity to present their findings. If teams stop exploring improvem, technical debt can pile up fast.
If teams stop exploring improvements, technical debt can pile up fast.
Building Robust Evaluation Processes
In the past, QA and testing happened at the end of a sprint. Now, with agentic engineering, LLM evaluation (Evals) is a key part of the architecture. Teams should setup strong, behavior-based evaluation processes to see how agents handle uncertainty and tool routing. For example, a good evaluation process could include defining real user scenarios, setting clear behaviors for each, and testing agent responses in different situations, like unclear user input or tool failures.
Teams can run these tests automatically in CI, score responses based on metrics like accuracy, safety, and proper tool use, and review failures often to improve prompts or code. If teams do not treat Evals as carefully as unit tests, they will miss out on the full benefits of autonomous Canary builds.
Fostering a Culture of Experimentation
Engineers should get credit for managing complexity well. Create a culture where building small, temporary test apps is seen as smart work, not just extra tasks. The quicker a team can set up a sandbox to test a new library version, the better your production apps will be. To help large teams or business units use this agile approach, offer a central place for sandbox templates, share best practices in regular meetings, and choose champions in each group to help others. This lets engineers across the company experiment and use sandboxes as a normal part of their work.
Final Thought
Transforming culture is an ongoing journey, not a checklist. By making continuous discovery, robust evaluation, and agile experimentation part of day-to-day engineering, leaders give their teams the resilience and creativity needed to thrive in an AI-driven landscape.
