Draftt and the retailer’s platform team mapped 11 recurring workflows, analyzed the bottlenecks and identified eight as candidates for full execution by AI agents. They started with two workflows and a goal of reducing the maintenance backlog by 10% within 60 days. The engagement expanded across 20 teams and four divisions.
The challenge
A growing maintenance backlog increased cloud costs and compliance risk. Limited engineering capacity and uncertainty about the impact of production changes prevented teams from clearing the work.
Results after six months
12,000 verified production changes across four divisions.
| Measure | Result | Detail |
|---|---|---|
| Proactiveness | 90% | of planned work done proactively, up from 20% |
| Customer effort | 95% less | ~57,000 hours returned to other priorities |
| Delivery cost | $7.2M | lower per year; 80% lower cost per verified task |
| AI execution | 60% | 7,200 of 12,000 tasks executed by agents |
| Adoption | 20 of 20 | teams used the production service |
| Technical debt | −30% | vs. an initial 10% goal in 60 days |
How the work was delivered
Cluster and database upgrades, OS and image rollouts, runtime, dependency, certificate and configuration maintenance.
- Define. Map obligations and bottlenecks, pick priority workflows, set the baseline.
- Commit. Agree outcomes, change plan, ownership, approvals and acceptance checks.
- Deliver. Agents apply approved changes; engineers complete work needing intervention.
- Prove. Verify production state and health, retain evidence, measure progress.
The work with Draftt changed the mindset of the team. We made proactive maintenance a service capability: 12,000 tasks completed, 95% less demand on our engineers and $7.2 million lower annual delivery cost, with AI adoption across multiple divisions. We see it as a major part of our software factory initiative.
— SVP Platform, Fortune 500 retailer
