Why "Agentic Sales" Is Marketing

An agent is a system that can decide what to do next without being told. Most "agentic sales" tools currently sold under that label are macros with better prose.

The distinction is not pedantic. It is the entire question. A workflow automation with an LLM narration layer, triggered by a rule and executing a predefined sequence, is not an agent. It is a macro. It runs when a condition is met. It does what the macro was built to do. The LLM layer writes the email or summarizes the output in natural language, but the decision of what to do is hard-coded at the workflow level. No reasoning occurs inside the loop. No alternate paths are considered. No judgment is exercised about whether the action should happen at all.

This is worth naming carefully because the category is being prematurely branded. Every revenue technology vendor who had AI features in 2024 has added the word "agentic" to their 2026 marketing. The implication is that the software now exhibits agent-like properties. For most of it, that implication is false.

The defensible definition of an agent involves three properties. First, the system selects actions from a space of possible actions, rather than executing a predefined sequence. Second, the system's selection process involves reasoning over the context, not just matching on triggers. Third, the system can choose not to act, or to act differently than the surface task would imply, when the context suggests a different action is warranted.

Almost nothing currently sold as agentic sales software meets all three tests. Some meet the first. A system that composes an email drawing from multiple templates based on context is selecting from a space. Some meet the second loosely. A system that summarizes a call and flags specific risks is reasoning over context. Almost none meet the third. The ability to not act, or to act differently than expected, is what distinguishes a true agent from a sophisticated automation. It is also the property that would make vendors nervous to deploy.

A true sales agent, deployed at scale, would occasionally decide not to send the email, not to book the meeting, not to advance the deal stage. It would do so because its reasoning about the context told it that the action would be counterproductive. Any vendor who deployed a true agent at scale would find customers complaining about missed outreach and skipped follow-ups. The product management answer to that complaint would be to reduce the agent's latitude. Reducing the latitude turns the agent back into a macro. The economic pressure flows toward macros, because macros are predictable, and macros with LLM layers can be marketed as agents without the operational risk of actual agency.

This is how the category is being hollowed out in real time. The thing being sold is not the thing being named. The category language implies autonomous reasoning. The product behavior is mostly conditional automation with a language model writing the output copy. The gap between the two will not be closed, because the economics of product management push against closing it.

The operational implication for CROs is that the agentic sales software they are purchasing in 2026 is, in most cases, an incremental improvement on the conditional automation they purchased in 2023. It will produce modest productivity gains. It will not produce structural change. The compounding advantage that agent-native revenue motions could theoretically produce will not be produced by this software.

True agent-native revenue motions are emerging, but they are being built inside a small number of operating companies, not sold by vendors. The companies building them are doing so with custom infrastructure. They are integrating frontier model APIs directly. They are writing their own orchestration layers. They are running their own evals. They are accepting the operational cost of genuine agency, including the cost of occasional miscalibrated actions, because they believe the compounding advantage over the medium term justifies the short-term operational overhead.

This is the work we would bet on. It is not visible in any vendor marketing. It is not purchasable as a SKU. It is built. The CROs who are building it tend to have engineering backgrounds or the willingness to lead cross-functional work with engineering. They are not the majority of the market. They will not be the majority of the market for several more years. When they are, the language of agentic sales will have moved on. The vendors currently using the term will still be selling conditional automation, under a different name.

The category is being prematurely branded because there is commercial pressure to associate every new AI feature with the frontier of the technology. The frontier, in late 2025 and early 2026, is agent-based systems. Every product manager at every revenue technology vendor has been briefed to find the agentic story in their roadmap. Some have found real agent-based features to point to. Most have found conditional automations with LLM narration and named them agents.

The buyer's defense against this is structural. A sales or RevOps leader evaluating agentic software should ask one question and accept no hedge on the answer. Can this system decide not to act? If the answer is no, or if the answer is yes but with such narrow latitude that the decision never occurs in practice, the system is a macro. It may be a very good macro. It may produce measurable value. It is not an agent, and the buyer should not pay agent-class prices or grant agent-class expectations.

By 2028, the language will have evolved. The vendors that deployed real agents will have different products than the vendors that deployed automations. The productivity gap will be visible in operator time reclaimed, in motion coherence, and in the specific kinds of outreach that only agent-based systems produce. Until then, the discipline is linguistic. Call the thing what it is. The macros are useful. The agents, when they arrive, will be transformational. Treating the former as the latter is how categories get hollowed out and how buyers get disappointed.