Agentwashing Sounds Like a SaaS Product. It's Actually a Confession.

How many of your org's "agents" can actually complete a multi-step task without someone feeding it each step by hand? VentureBeat's Pulse Research asked that question directly after surveying 101 enterprises about their AI agent deployment. The answer was blunt: 71% admitted a quarter or fewer of what they called agents were real multi-step workflows, not single-prompt wrappers with new packaging. Only 10% had more than half their fleet crossing that bar. Claude leads as the primary platform at 40%, roughly double whatever follows. Microsoft lands at 18%, OpenAI at 13%. Could not independently verify those share figures beyond VentureBeat's own press release. The cost question is worse. 27% have no real-time visibility into what a running agent is spending. That means somewhere between 10 and 27 organizations were flying blind on runtime costs at the time of the survey. Compare that to other surveys from roughly the same period. Zapier found 72% of enterprises either deploying or testing agents. Writer's 2026 poll had 97% of executives claiming agent deployment. Same market. Wildly different numbers. Because none of these surveys actually measure the same thing, and nobody standardized on what "agent" means yet. Gartner coined a word for it: agentwashing. Companies call whatever they built an agent. The word stopped meaning anything specific enough to compare across organizations, which makes every adoption stat in this space almost worthless without reading the methodology first. Agent orchestration isn't meaningless where it was actually constructed that way. It just means the industry is still arguing about definitions while pretending to measure maturity. https://preview.redd.it/v7xtkw5oyodh1.png?width=4500&format=png&auto=webp&s=965c1b9efc3002dba03ae6aa4b260945b0ea5e9f