The Fastest AI Agent Isn't the One That Wins — Forrester's Four-Speed Framework for Enterprise Agents

Forrester VP Craig Le Clair has a new framework for the crowded enterprise AI-agent market, and it starts by admitting the obvious leaders aren’t the safe bet. His July 22 post sorts agent vendors into four speeds: “Hares” — fast-deploying, horizontal SaaS agents from Microsoft, Salesforce, ServiceNow, SAP, and Google, dominating early adoption on tasks like email summarization and content drafting, but running on largely the same foundation models and exposed to commoditization; “Sloths” — edge agents running near the data source (factories, medical devices, industrial equipment) on smaller models optimized for speed and low cost; “Tortoises” — custom-built agents running behind the firewall on private models, addressing an organization’s specific processes rather than generic tasks, which Le Clair argues “close the action gap” better than horizontal tools; and “Super-Tortoises” — coordinated multi-agent systems built for a specific industry or function, capable of running an entire business process end to end. His conclusion: “the fastest runner doesn’t always win,” and the organizations that come out ahead will run more than one speed rather than chasing whichever agent shipped first.

The framework gives language to something Forrester’s own research already pointed at a week earlier. Principal analyst David Mooter argued that an agent without integration is “merely an answer engine” — and found that organizations with genuinely high readiness to adopt agentic AI are spending their next 12 months on integration and change management, while low-readiness organizations are still shopping for model capability. Put next to Le Clair’s framework, that’s a tell: teams comparing agents on model quality are shopping for a Hare, while the orgs actually closing gaps are building Tortoise-style integration work most vendors don’t sell.

BCG’s guidance on regulated-industry agent deployments backs the pattern with a number: a platform approach — not a fragmented one — produced a 25% productivity gain across the software development life cycle and a 20-30% quality improvement, with one automotive client’s knowledge-assistant deployment cutting search time 50-75%. For anyone running an agent bake-off this quarter, the deciding question isn’t which vendor demoed fastest. It’s whether the Tortoise work — integration, governance, private-model deployment — is even on the evaluation scorecard.