Field notes on how agentic products talk. Same test every time:
- Who is the hero?
- What does the first sentence sell?
- Is draft vs post visible?
These are not reviews. They are first-line audits.
Read the arguments first:
Horizontal Agents
Salesforce Agentforce
A catalog of job-ready agents. The hero is the agent at work. Trust is platform gravity.
ServiceNow AI Agents
Workflow OS plus agents. Same move as Agentforce: execution across the instance, control as the enterprise wrapper.
UiPath
Computer-use with a bot pedigree. The sentence is still “it drives the UI.” Governance is a later slide.
Glean Tau
Work graph plus agents. Knowledge-work beachhead, not a posting decision.
Anthropic Claude Projects / Agents
Strong model, weak Office of the CFO sentence. Projects are context. They are not a policy surface.
Harvey
Legal work with a trust buyer. Closer to “who owns the miss” than most horizontal tools — still not a ledger.
Operator / MultiOn / Adept
One sentence in three wrappers: the agent uses the computer for you. Fine for a ride. Fatal as the first line into the ledger.
Finance Wrappers
Ramp
Spend and close-adjacent automation. Speed and savings lead. The miss is a footnote under the card.
Brex
Card-plus-software. Autonomy and intelligence lead. Controller language is downstream of the spend product.
NetSuite Next / SuiteAgents
Right beachhead: agents in the ERP you already own. Wrong hero: “on behalf of users,” with roles and gates as the asterisk.
The Test
They lead with executes / faster / included.
They should lead with draft vs post / who owns the miss / then the agent.