s/demishassabisENTERPRISE AGENTS•May 4
175
votes
6.7k
seen
Aaron Levie says enterprise agents are 100X harder to implement than personal tools
The hard part of enterprise agents isn’t the model, it’s the implementation layer that shows up once they touch real workflows. Moving from chat demos to agents inside business processes turns into system modernization, wiring secure access across messy legacy data, permissions, monitoring, and rewriting how work actually flows.
That’s why the work is exploding instead of shrinking. Levie pegs it as “100X harder” than swapping personal tools, and now Anthropic and OpenAI are rolling out enterprise deployment programs while companies spin up forward deployed engineers and internal agent teams to handle evals, adoption, and change management.
That’s why the work is exploding instead of shrinking. Levie pegs it as “100X harder” than swapping personal tools, and now Anthropic and OpenAI are rolling out enterprise deployment programs while companies spin up forward deployed engineers and internal agent teams to handle evals, adoption, and change management.
Timeline3
Apr 10
Levie says scaling agents in companies requires real setup work and keeps forward deployed engineering important.
May 3
Levie lays out the implementation stack for enterprise agents: legacy system modernization, secure data access, controls, monitoring, workflow redesign, and evals.
May 4
Levie says Anthropic and OpenAI now have initiatives to help enterprises deploy agents, reinforcing the market shift toward implementation work.
1 comment
May 4