AI Success Hinges on Preparation, Not Speed
A recent study of over 2,000 leaders found that what separates successful organizations from those stuck in expensive pilots has little to do with speed. Instead, it comes down to what they did before launching AI agents: preparing the right data, defining where a person stays in the loop, and building guardrails beforehand.
Only 31% of organizations unified their data first, but this didn't result in significantly better returns on investment (ROI). In fact, those that deployed iteratively, starting with what they had and integrating more over time, still reached ROI in 8.2 months, a modest gap from the 7.3 months it took for those with fully unified data.
The study also found that perfect data isn't necessary; organizations can start with what they have and integrate more data as needed. This approach is reflected in Salesforce's own experience with Help Agent, an AI agent on their support platform that has handled nearly five million customer conversations across seven languages and three portals, resolving 68% without human intervention.