Legacy Code Mess Hinders AI Adoption in Enterprises

As AI adoption accelerates in enterprises, companies are facing a new challenge: deploying agents into their workflows. While purchasing an agent is no longer a difficult task, getting it to work seamlessly with existing systems can be a daunting problem.

CMG Financial, a US mortgage company, encountered this issue firsthand when they tried to integrate 100 Agents into their Salesforce system. Although they had migrated part of their software development work to Claude Code and proved their engineering team's ability to adopt the latest AI tools, the project slowed down when trying to get Agents to access Salesforce and participate in real business processes.

The problem lies in the 'legacy code mess' that enterprises have accumulated over the years. Agents can write code and call APIs, but they lack understanding of data, permissions, and business processes, making it difficult to deploy them effectively.

A study by Christian & Timbers found that only about 2,000 Forward Deployed Engineers (FDE) in the US have the ability to deploy AI systems into enterprises and help customers obtain quantifiable returns. To address this shortage, large consulting and service companies are rapidly expanding their FDE teams.

Startup company June helped CMG find a solution by scanning their existing software and databases, identifying business processes, duplicate fields, data breaks, and permission conflicts, and generating an implementation roadmap.