Indeed Scales AI Adoption to 97% via Structured Enablement
Indeed increased AI coding tool adoption from 25% to 97% and reduced coding time by 35% through direct training, community engagement, and a mandate-to-train strategy. The case study highlights the shift from train-the-trainer models to comprehensive enablement and the emergence of code review bottlenecks.
Indeed's engineering organization achieved a transformative leap in AI adoption, moving from 25% to 97% weekly active usage while reducing coding time by over 35% for trained cohorts. This success stemmed from a deliberate pivot away from ineffective "train-the-trainer" models and tool mandates toward comprehensive, direct enablement. The "AI Coding Essentials" program required training completion rather than tool usage, a nuance that fostered psychological safety and intrinsic motivation. By investing approximately $3-4 million in developer time for a five-week, tool-agnostic curriculum, Indeed ensured engineers could fluently navigate multiple agentic IDEs, including Cloud Code, Cursor, and Windsurf, without vendor lock-in. Teams frequently utilized three or more tools simultaneously, demonstrating that a multi-tool ecosystem enhances productivity.
Strategic Enablement Over Coercion
The initiative highlighted that enablement drives behavior change more effectively than mandates. Early attempts using "AI Coding Ambassadors" failed to sustain adoption among non-participants, revealing that peer influence alone was insufficient without direct skill acquisition. The subsequent direct training model, supported by leadership prioritization and private manager dashboards for completion tracking, achieved a 70% completion rate and measurable productivity gains. This approach balanced organizational goals with developer autonomy, proving that structured learning environments are critical for scaling complex tool adoption. Engineering leads explicitly carved out time for training, treating upskilling as a delivery priority.
Community and Continuous Learning
Sustaining adoption required robust community infrastructure. Indeed established a multi-modal engagement strategy featuring a 2,100-member Slack channel, monthly coding forums, and an "AI Showcase" recognition program offering monetary prizes for innovation. These mechanisms kept the conversation alive post-training, encouraging knowledge sharing across engineering and non-engineering R&D roles. The data showed that community engagement correlated with higher tool effectiveness, particularly as frontier models evolved, necessitating continuous upskilling in context engineering and MCP usage. Adoption spikes post-holidays further indicated that personal experimentation reinforced professional application.
Operational Outcomes and Emerging Challenges
The productivity gains were quantifiable: diffs per engineer increased by 40-50% year-over-year, and DORA metrics improved without compromising quality, maintaining a change failure rate below 4%. Coding time was rigorously measured from Jira ticket pickup to GitLab diff opening. However, the surge in coding velocity introduced new operational bottlenecks. Merge request resolution rates began to lag behind creation, indicating a strain on code review processes. Indeed is now addressing this by exploring AI-assisted review workflows and expanding enablement to the broader product development lifecycle, including product management and UX, to optimize the entire value chain. Future initiatives include "AI as Coach" to provide real-time guidance for untrained users.
Key insights
-
Structured, hands-on training for all engineers significantly outperforms train-the-trainer models in sustaining AI tool adoption across large organizations.
Impact: Prevents adoption decay post-program and ensures uniform skill levels, reducing variance in team productivity.
-
Mandating training completion rather than tool usage preserves psychological safety while driving higher intrinsic adoption and effective tool application.
Impact: Reduces resistance to new technologies and fosters a culture of voluntary experimentation and continuous learning.
-
Tool-agnostic enablement encourages multi-tool fluency, with teams leveraging three or more agentic tools simultaneously to optimize workflows.
Impact: Mitigates vendor lock-in risks and allows engineers to select the best tool for specific tasks, enhancing overall flexibility.
-
Rapid increases in coding velocity can expose downstream bottlenecks in code review and merge request resolution, requiring process adaptation.
Impact: Organizations must monitor end-to-end SDLC metrics to prevent velocity gains from stalling at integration points.
Action items
-
Evaluate current AI adoption strategies and shift from peer-led ambassador programs to direct, structured training for all engineering staff.
Impact: Ensures consistent skill acquisition and prevents adoption gaps between early adopters and the broader workforce.
-
Establish organizational targets for training completion rather than tool usage, supported by private manager dashboards to facilitate supportive conversations.
Impact: Drives high completion rates while maintaining psychological safety and avoiding counterproductive coercion.
-
Launch recognition programs and community channels that incentivize experimentation and knowledge sharing across both engineering and non-engineering R&D roles.
Impact: Sustains engagement long after formal training ends and accelerates cross-functional adoption of productivity tools.
-
Track end-to-end SDLC metrics, including merge request resolution times, to identify and address bottlenecks created by increased coding velocity.
Impact: Prevents productivity gains from being negated by downstream delays in code review and integration processes.
Quotes
“We never set a mandate that said you must use AI, right? We did, however, set a very suggested, you should complete this target for all of our engineers for the training itself.”
“What we realized was the key difference was actually having the hands-on training yourself, right? Not having somebody tell you, hey, here's this cool training that I took and here's the stuff I learned, but actually having people take that training.”
“We're going to try to reduce the size of that yellow piece of the pie... make the Pac-Man bigger and thereby double overall engineering product productivity over the next couple of years.”