What is an AI sales agent?

An AI sales agent is an autonomous software system that handles one or more stages of the sales process: prospecting, outreach, qualification, follow-up, or booking. Unlike a simple chatbot, an AI sales agent can reason about context, take multi-step actions (research prospect, draft email, update CRM, book meeting), and operate without per-interaction human approval.

What are the main types of AI sales agents?

Three main types: (1) Outbound SDR agents — find prospects, write personalised emails, send sequences, handle replies, book meetings. Examples: Reply.io Jason AI, 11x Alice. (2) Inbound qualification agents — respond to inbound leads via chat or email, qualify against ICP, route to human. (3) Voice SDR agents — make or receive calls, qualify via voice, book calendar appointments. Examples: Vapi, Retell AI-powered agents.

What is TCPA and does it apply to AI sales agents?

TCPA (Telephone Consumer Protection Act) applies to automated telephone calls and text messages in the US. AI voice agents that make outbound calls are subject to TCPA. The FCC issued a ruling (FCC 24-17) affirming that AI-generated voice calls require prior express written consent from recipients. Violations carry $500–$1,500 per call in statutory damages. Always obtain proper consent before running AI voice outbound campaigns.

What is California AB 3030 and how does it affect AI sales agents?

California AB 3030 (effective January 1, 2025) requires that AI-generated communications in health care contexts clearly disclose they are AI-generated. California AB 489 (proposed) extends similar disclosure requirements to other commercial AI communications. While AB 3030 is health-care-specific, disclosure obligations for AI sales agents are expanding. Consult legal counsel on current California law before deploying AI sales agents to California residents.

How do I evaluate an AI sales agent before buying?

Use a 6-dimension scoring framework: (1) Outreach quality — personalisation depth, reply handling accuracy. (2) CRM integration — bidirectional sync, field mapping. (3) Compliance tools — opt-out handling, consent management, DNC list scrubbing. (4) Reporting — reply rates, meeting booked rates, pipeline influenced. (5) Guardrails — how the agent handles objections, discloses AI identity when asked. (6) Cost — total cost per meeting booked including tool cost + AI API cost.

What evidence labels should I apply to AI sales agent claims?

Vendor claims about AI sales agent performance should be categorised: [Vendor Claim] — unverified assertion from the vendor. [Case Study] — result reported by a specific named customer. [Independent Test] — result measured by an independent third party. [Regulatory Finding] — compliance status confirmed by regulator. Treat [Vendor Claim] with significant scepticism; ask for [Case Study] evidence before buying.

Buyer Framework — June 2026

Best Way to Choose AI Sales Agents (2026)

A 6-dimension scoring framework, evidence label system for evaluating vendor claims, and compliance obligations including TCPA FCC 24-17 and California AB 3030.

Reviewed by Jordan M. Reyes — Updated 2026-06-13

Three Types of AI Sales Agents

Type	What It Does	Channel	Examples
Outbound SDR agent	Prospects, writes email, handles replies, books meetings	Email, LinkedIn	Reply.io Jason AI, 11x Alice
Inbound qualification agent	Responds to inbound leads, qualifies, routes to human	Chat, email	Drift AI, Intercom Fin
Voice SDR agent	Makes/receives calls, qualifies via voice, books appointments	Phone	Vapi, Retell AI-powered agents

6-Dimension Scoring Framework

Score each AI sales agent candidate on a 1\u20135 scale across these six dimensions. Weight them by what matters most to your use case.

Dimension	What to Evaluate	Red Flags
1. Outreach quality	Personalisation depth, reply classification accuracy, multi-touch sequence logic	Generic templates, no reply handling
2. CRM integration	Bidirectional sync, field mapping, pipeline stage updates	One-way export only, manual field mapping
3. Compliance tools	Opt-out handling, consent management, DNC list scrubbing, geographic restrictions	No opt-out mechanism, no DNC scrubbing
4. Reporting	Reply rates, meetings booked, pipeline influenced, cost per meeting	Only vanity metrics (sends, opens)
5. Guardrails	AI identity disclosure when asked, objection handling, escalation to human	No identity disclosure, no escalation path
6. Total cost	Platform fee + AI API cost + email infrastructure + meetings booked	Platform fee quoted without API cost

Evidence Labels for Vendor Claims

AI sales agent vendors make aggressive performance claims. Apply evidence labels before acting on any claim:

[Vendor Claim] — Unverified assertion from the vendor or its marketing. Treat with significant scepticism. These are aspirational, often tested in ideal conditions.
[Case Study] — Result reported by a specific named customer. More credible, but check industry match, team size, and ICP similarity to yours.
[Independent Test] — Result measured by an independent third party. Highest credibility. Rare in this industry.
[Regulatory Finding] — Compliance status confirmed by regulator. Use for compliance claims (e.g., TCPA safe harbour, GDPR adequacy).

Compliance: TCPA, AB 3030, and FTC

TCPA / FCC 24-17

FCC ruling 24-17 affirmed that AI-generated voice calls (even those with natural-sounding voices) require prior express written consent from recipients. This applies to AI outbound SDR calls. Do not confuse with general telemarketing rules — AI voice has specific consent requirements.

California AB 3030 and AB 489

AB 3030 (effective January 1, 2025) requires disclosure of AI-generated content in healthcare contexts. AB 489 (proposed) would extend AI disclosure requirements more broadly. Sales AI deployments targeting California residents should include AI identity disclosure regardless of current law status, as the regulatory trend is clear.

FTC Impersonation Rule

The FTC’s impersonation rule prohibits AI from impersonating businesses or government agencies in ways that deceive consumers. AI sales agents that falsely represent themselves as human when directly asked violate this rule.

Frequently Asked Questions

What is an AI sales agent?: An AI sales agent is an autonomous software system that handles one or more stages of the sales process: prospecting, outreach, qualification, follow-up, or booking. Unlike a simple chatbot, an AI sales agent can reason about context, take multi-step actions (research prospect, draft email, update CRM, book meeting), and operate without per-interaction human approval.
What are the main types of AI sales agents?: Three main types: (1) Outbound SDR agents — find prospects, write personalised emails, send sequences, handle replies, book meetings. Examples: Reply.io Jason AI, 11x Alice. (2) Inbound qualification agents — respond to inbound leads via chat or email, qualify against ICP, route to human. (3) Voice SDR agents — make or receive calls, qualify via voice, book calendar appointments. Examples: Vapi, Retell AI-powered agents.
What is TCPA and does it apply to AI sales agents?: TCPA (Telephone Consumer Protection Act) applies to automated telephone calls and text messages in the US. AI voice agents that make outbound calls are subject to TCPA. The FCC issued a ruling (FCC 24-17) affirming that AI-generated voice calls require prior express written consent from recipients. Violations carry $500–$1,500 per call in statutory damages. Always obtain proper consent before running AI voice outbound campaigns.
What is California AB 3030 and how does it affect AI sales agents?: California AB 3030 (effective January 1, 2025) requires that AI-generated communications in health care contexts clearly disclose they are AI-generated. California AB 489 (proposed) extends similar disclosure requirements to other commercial AI communications. While AB 3030 is health-care-specific, disclosure obligations for AI sales agents are expanding. Consult legal counsel on current California law before deploying AI sales agents to California residents.
How do I evaluate an AI sales agent before buying?: Use a 6-dimension scoring framework: (1) Outreach quality — personalisation depth, reply handling accuracy. (2) CRM integration — bidirectional sync, field mapping. (3) Compliance tools — opt-out handling, consent management, DNC list scrubbing. (4) Reporting — reply rates, meeting booked rates, pipeline influenced. (5) Guardrails — how the agent handles objections, discloses AI identity when asked. (6) Cost — total cost per meeting booked including tool cost + AI API cost.
What evidence labels should I apply to AI sales agent claims?: Vendor claims about AI sales agent performance should be categorised: [Vendor Claim] — unverified assertion from the vendor. [Case Study] — result reported by a specific named customer. [Independent Test] — result measured by an independent third party. [Regulatory Finding] — compliance status confirmed by regulator. Treat [Vendor Claim] with significant scepticism; ask for [Case Study] evidence before buying.