> Enrich.chat gives AI agents verified work emails and mobiles. To use it from an agent, read https://enrich.chat/agents.md . This is the Markdown version of https://enrich.chat/blog/ai-sales-agents/

# AI sales agents in 2026: what they do, what is hype, and why contact data decides the result

A plain guide to AI sales agents and AI SDRs: the jobs they do, vendor prices checked October 15, 2026, independent evidence, and the data layer behind every result.

By Enrich.chat Editorial. Updated October 15, 2026.

**Short answer:** An AI sales agent is software that does a sales task on its own: it researches accounts, finds contacts, drafts outreach and books meetings by calling other tools. The jobs are real, but [Gartner estimates only about 130 of the thousands of agentic AI vendors are genuine](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027). A model cannot recall a work email from memory, so results depend on the contact data you give it.

## Key takeaways

- An agent directs its own steps and tool calls, while a workflow follows a fixed path. Gartner says some vendors relabel assistants, RPA and chatbots as agents.
- Gartner estimated in June 2025 that only about 130 of the thousands of agentic AI vendors are real, and predicted over 40% of agentic AI projects will be canceled by the end of 2027.
- In Salesforce's survey of 4,050 sales professionals, 54% of sellers had used AI agents, and 51% of sales leaders using AI said disconnected systems slow their AI work.
- Prices checked October 15, 2026: HubSpot's Prospecting Agent costs $1 per lead it recommends, 11x lists Alice from $3,750 a month billed annually, and Artisan publishes no prices.
- In a May 2025 Salesforce AI Research benchmark, leading agents solved about 58% of single-turn CRM tasks and about 35% of multi-turn ones. Give yours a tool that returns a status on every contact field.

## What is an AI sales agent?

An AI sales agent is software that does a sales task on its own. It uses a language model to pick the steps, calls tools to look things up or take actions, and hands back a result. [Salesforce's guide](https://www.salesforce.com/sales/ai-sales-agent/guide) splits agents into autonomous ones, such as an SDR agent that answers inbound leads and books meetings, and assistive ones, such as a coach that role-plays calls with a seller.

The word "agent" is used loosely, so use a test. [Anthropic's engineers](https://www.anthropic.com/engineering/building-effective-agents) define a workflow as a system where code sets the path, and an agent as one where the model directs its own steps and tool use. A tool that sends email one, two and three on a timer is a workflow, even if a model writes the copy. [ZoomInfo's guide](https://pipeline.zoominfo.com/sales/ai-sales-agents) puts it another way: ask who owns the next step. An assistant waits for a rep. An agent finishes the task and returns a result.

Neither design is better by default. Anthropic notes that agents often trade speed and cost for better results on open-ended work, and it advises starting with the simplest design that does the job. For a fixed task like "send this three-step sequence," a workflow is cheaper and easier to check.

**Figure 1: Anatomy of an AI sales agent.** You give the agent a goal, such as booking meetings with operations leaders at logistics companies, and limits: an approval step, a daily send cap and an opt-out list. The agent is a language model in a loop: it plans the next step, calls a tool, reads the result, and repeats or stops. It can call five tools: a contact data API, a CRM, web and news search, an email sender and a calendar. A rule at the bottom says a model cannot recall a work email, so it has to call a data tool.

*An AI sales agent is a model in a loop that calls tools. The contact data tool is the one a model cannot replace from memory. Editorial diagram based on [Anthropic's definition of an agent](https://www.anthropic.com/engineering/building-effective-agents).*

## What jobs do AI sales agents do?

AI sales agents do six jobs: they build lists, research accounts, find contact details, draft messages, handle replies, and book meetings while keeping the CRM current. Each job needs different inputs.

| Job | What the agent does | What it has to look up |
| --- | --- | --- |
| 1. Build the list | Turns your ideal customer profile into a search and watches for triggers such as job changes | People and company search |
| 2. Research the account | Reads news, hiring and CRM history and writes a short brief | Company profile, tech stack, CRM notes |
| 3. Find contact details | Adds a work email, mobile, title and LinkedIn URL for each person | A contact data source with a status per field |
| 4. Draft the message | Writes a first line and an email from facts in the brief | The brief and your style rules |
| 5. Handle replies | Answers questions, honors opt-outs, nudges quiet leads and passes warm ones to a person | Product docs and the opt-out list |
| 6. Book and log | Books time with a rep and records the activity | Calendar and CRM |

The jobs match what vendors sell. [ZoomInfo's guide](https://pipeline.zoominfo.com/sales/ai-sales-agents) sorts agents into six types: prospecting, inbound, research, deal execution, CRM and data, and customer expansion. [HubSpot's Prospecting Agent](https://www.hubspot.com/products/sales/ai-prospecting-agent) monitors buying signals, rates accounts on fit, finds the buying committee, researches each account and drafts outreach in each rep's voice. Salesforce's Agentforce SDR works the inbound side. [Salesforce's training module](https://trailhead.salesforce.com/content/learn/modules/agentforce-sales-agents-quick-look/discover-agentforce-sales-agents) says it sends individual messages, answers questions and sets up first meetings for reps.

Models are strong at three of these jobs because they read and write: research briefs, drafts and reply sorting. The other three depend on facts a model does not have. Who the people are, how to reach them and what is true today must come from a tool.

## What is an AI SDR, and what is AI prospecting?

An AI SDR is an AI sales agent aimed at the sales development rep's job: find prospects, contact them and book the first meeting. AI prospecting is the front half of that job, which is building the list and researching who to contact.

Vendors use different names for the same idea. [11x sells Alice](https://www.11x.ai/products/alice/pricing) as an outbound SDR that handles prospecting, multi-channel outreach and meeting booking, plus a second agent, Julian, for inbound leads. [Artisan sells Ava](https://www.artisan.co/pricing) as an AI BDR that finds leads, writes outreach and handles replies. HubSpot calls its agent a Prospecting Agent, and Salesforce calls its agent an SDR agent.

Vendor guides draw the same line between agent and person. [Apollo's guide](https://www.apollo.io/insights/ai-sales-agent) keeps objection handling, pricing negotiation and executive relationships with people. Salesforce reports that its own agents contacted 130,000 leads and created 3,200 opportunities in four months, per its [2026 State of Sales announcement](https://www.salesforce.com/news/stories/state-of-sales-report-announcement-2026/). That is a vendor-reported result.

You can try AI prospecting without buying an agent. The free [AI lead list builder](https://enrich.chat/tools/lead-list-builder/) turns a plain-English description of your buyer into matching people. To compare tools, read our guide to the [best AI SDR tools](https://enrich.chat/blog/best-ai-sdr-tools/). To build your own, follow [how to build an AI SDR](https://enrich.chat/blog/how-to-build-an-ai-sdr/). For the whole path from ideal customer profile to meeting, see [AI lead generation in 2026](https://enrich.chat/blog/ai-lead-generation/).

## Do AI sales agents work, or is it hype?

Parts of it work and parts are hype. Research, drafting and following a defined workflow have support. Fully autonomous selling does not, and independent analysts say many products sold as agents are not.

Gartner uses the term "agent washing" for vendors that rebrand AI assistants, robotic process automation and chatbots as agents. In a [June 2025 release](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027), Gartner estimated that only about 130 of the thousands of agentic AI vendors are real. It also predicted that over 40% of agentic AI projects will be canceled by the end of 2027, citing cost, unclear value and weak risk controls.

Sales shows the same pattern. In a [July 2026 release](https://www.gartner.com/en/newsroom/press-releases/2026-07-28-gartner-predicts-ai-agents-will-outnumber-sellers-10-to-1-by-2028-yet-fewer-than-40-percent-of-sellers-will-say-agents-improved-productivity), Gartner predicts that by 2028 AI agents will outnumber sellers 10 to 1, yet fewer than 40% of sellers will say the agents improved their productivity. Its sales analyst said agents are not a shortcut to productivity. Gartner's advice is to build a central context layer that connects enterprise data, systems and seller judgment first.

Benchmarks show the same split. [CRMArena-Pro](https://arxiv.org/abs/2505.18878) is a May 2025 test from Salesforce AI Research with 19 tasks across sales, service and quoting. Leading agents solved about 58% of tasks in one turn and about 35% in multi-turn conversations. They did best at following a defined workflow, where top scores passed 83%. The paper also found near-zero built-in awareness of confidentiality. Prompting helped, but it often hurt task performance. A second benchmark, [TheAgentCompany from Carnegie Mellon](https://arxiv.org/abs/2412.14161) and partners, found the best agent it tested finished 30% of simulated workplace tasks on its own.

**Figure 2: How well leading AI agents did on CRM tasks in a 2025 benchmark.** A bar chart of results from the CRMArena-Pro benchmark. Top agents scored above 83 percent on workflow execution tasks in a single turn, about 58 percent across all tasks in a single turn, and about 35 percent across all tasks in multi-turn conversations.

*Success rate of leading agents on 19 sales, service and quoting tasks. Defined workflows went well. Longer conversations did not. Models have changed since May 2025, so read this as a snapshot. Source: [CRMArena-Pro, Salesforce AI Research](https://arxiv.org/abs/2505.18878).*

Vendor claims need the same care. [TechCrunch reported in March 2025](https://techcrunch.com/2025/03/24/a16z-and-benchmark-backed-11x-has-been-claiming-customers-it-doesnt-have/) that several companies whose logos appeared on 11x's website said they were not customers. It also reported that ZoomInfo said a one-month trial found the product performed worse than its own SDRs. 11x said it removes inaccurate mentions on request and called the remaining cases human error. Salesforce and HubSpot sell agents too, so treat their results as vendor data. In [Salesforce's survey of 4,050 sales professionals](https://www.salesforce.com/news/stories/state-of-sales-report-announcement-2026/), 54% of sellers said they had used AI agents.

The practical reading is simple. Use agents for research, drafting and sorting. Keep a person on judgment calls. Ask any vendor for references and a pilot you can measure against your current process.

## How much do AI sales agents cost?

Published prices run from $1 per recommended lead to $3,750 a month on an annual contract, and some vendors publish nothing. These figures come from each vendor's own page, checked October 15, 2026.

| Product | Published price | What you pay for | Notes |
| --- | --- | --- | --- |
| [HubSpot Prospecting Agent](https://www.hubspot.com/products/sales/ai-prospecting-agent) | $1 per lead | Each lead it recommends outreach for | Runs on HubSpot Credits. Starter, Professional and Enterprise editions |
| [Salesforce Agentforce](https://www.salesforce.com/agentforce/pricing/) | $0.10 per standard action, or $2 per conversation | Actions on Flex Credits, or conversations | A $125 per user per month add-on is also listed. The page does not say which model covers the SDR agent, and it says prices may change |
| [11x Alice](https://www.11x.ai/products/alice/pricing) | From $3,750 a month, billed annually | Growth plan: up to 5 users and 2,000 new prospects a month | Pro and Enterprise are custom. An FAQ on the same page says "$36,000 per year," so confirm in writing |
| [Artisan Ava](https://www.artisan.co/pricing) | Not published | Plans sized to lead volume, mailboxes and dialer seats | Two plans list about 2,500 and 6,000 leads contacted a month. All three say "Talk to sales" |
| Your own agent plus a data API | 10 free lookups a day, then plans from $29 a month | Credits, spent when a lookup finds something | You supply the model and the sending tool. See [Enrich.chat pricing](https://enrich.chat/pricing/) |

The billing units differ, so compare cost per outcome and not the sticker price. HubSpot bills a recommended lead. Salesforce bills an action or a conversation. 11x says on its page that it charges per lead, not per send. Ask each vendor what counts as one unit, which contact data is included and what happens when an email bounces.

Data is the part you can price on its own. 11x's Growth plan covers 2,000 new prospects a month. Suppose an agent of your own looked up each of those people and a work email on Enrich.chat, and every lookup matched. That would use 30,000 credits, which fits the Growth plan at $29 a month. That covers the data only. You still pay for the model, a sending tool and the time to run it, while 11x's page says its price also bundles mailboxes, CRM sync and onboarding.

## Why is contact data the bottleneck for AI sales agents?

An AI sales agent can plan and write well, but it cannot know who works where or how to reach them. That has to come from a data source, and the data sets the ceiling for everything after it.

### A model fills gaps with guesses

A [2025 paper by Kalai and colleagues](https://arxiv.org/abs/2509.04664) argues that training and testing reward guessing over admitting uncertainty, which is why models state wrong facts with confidence. In a [2023 study in Scientific Reports](https://www.nature.com/articles/s41598-023-41032-5), 55% of the citations GPT-3.5 wrote and 18% of GPT-4's were fabricated, across 636 citations on 42 topics. Newer models do better, but a made-up fact looks just like a real one. A guessed address like first.last@company.com follows a real pattern, so a reader cannot tell it was never checked.

### People change jobs

The [US Bureau of Labor Statistics](https://www.bls.gov/news.release/tenure.nr0.htm) reported in September 2026 that median tenure with a current employer was 4.1 years in January 2026. It also found that 20.6% of wage and salary workers had been in their job a year or less. Those figures cover all US workers, and senior buyers may move less often, but the direction is the same: contact lists go out of date. Our guide to [how fast B2B contact data goes bad](https://enrich.chat/blog/b2b-data-decay/) goes deeper. An agent working from a stale list sends to the wrong people at scale.

### Bad data hurts the sender

[Google's sender guidelines](https://support.google.com/mail/answer/81126?hl=en) tell senders to keep spam rates in Postmaster Tools below 0.10%, to avoid ever reaching 0.30%, and not to buy email addresses. An agent that mails unchecked addresses at volume can break those limits quickly, and the damage lands on your own domain. See our guides to [cold email data](https://enrich.chat/blog/cold-email/) and [email bounce rate](https://enrich.chat/blog/email-bounce-rate/).

### Agents need joined-up data

In [Salesforce's 2026 survey](https://www.salesforce.com/news/stories/state-of-sales-report-announcement-2026/), 51% of sales leaders who use AI said disconnected systems are slowing their AI work. [ZoomInfo's guide](https://pipeline.zoominfo.com/sales/ai-sales-agents) gives a simple example: if industry is filled in for only about half of your accounts, an agent that tailors messages by industry will do poorly. Even HubSpot's agent sources contacts from outside providers, which [its page](https://www.hubspot.com/products/sales/ai-prospecting-agent) names as ZoomInfo, Apollo, Surfe and Seamless. The data layer is a separate piece, so choose it on purpose.

## How do you give an AI sales agent contact data?

Connect the agent to a contact data API, either through tool calls or through MCP, and tell it to report only what the tool returns. Enrich.chat is built for this: one prompt, no SDK, and a status on every field.

The [Model Context Protocol (MCP)](https://modelcontextprotocol.io/docs/getting-started/intro) is an open standard for connecting AI applications such as Claude or ChatGPT to outside tools and data. An agent can also call any HTTPS API directly. With Enrich.chat you can start with one prompt that begins "Use www.enrich.chat". The agent reads [enrich.chat/agents.md](https://enrich.chat/agents.md) and learns the endpoints and rules. Or add the keyless MCP server at `enrich.chat/mcp`, and use the [install page](https://enrich.chat/install/) for one-click steps. It works in ChatGPT, Claude, Cursor, Codex, VS Code, n8n, Make and Zapier (via HTTP). Our guide to [MCP servers for sales](https://enrich.chat/blog/best-mcp-servers-for-sales/) compares the options.

| Agent job | Endpoint | What comes back |
| --- | --- | --- |
| Build the list | `GET /v1/search` | Name, title, company, LinkedIn URL |
| Research the account | `GET /v1/company`, `/v1/company/tech` | Industry, headcount, tech stack |
| Find the buying committee | `GET /v1/company/people` | People at a domain, by title |
| Find contact details | `GET /v1/person`, `/v1/email`, `/v1/mobile` | Work email and mobile, each with a status |

Each work email comes back with a status from our records: `verified`, `catch_all` (the domain accepts every address, so it cannot be confirmed) or `unknown`. A mobile comes back with its line type, `mobile` or `unknown`. Treat every contact as a strong lead, not a guarantee. The agent should act on the status, as Figure 3 shows.

**Figure 3: What an agent should do with each contact status.** An example response from the person endpoint shows a work email with the status verified, 10 credits used and 9 free lookups remaining. Below it, four rows say what the agent should do. Verified: draft the message and queue it for a person. Catch-all: the domain accepts any address, so treat it as risky. Unknown: flag it for a person and do not guess a fix. Found false: skip the row, because nothing was found and 0 credits were used.

*An annotated example response and the rule an agent should apply to each status. The person and address are made up. Status values and credits come from the Enrich.chat [agents.md](https://enrich.chat/agents.md) instructions.*

Paste this prompt into ChatGPT, Claude or Cursor. Five people with a work email each take at most six of your 10 free lookups for the day, because one search counts as one lookup however many people it returns.

```text
Use www.enrich.chat to find 5 heads of sales at US logistics companies with 50 to 200 employees. Add each work email with its status, keep verified emails in one list and catch_all and unknown emails in another, and return a table with name, title, company, LinkedIn URL and email status. Do not draft or send anything. Tell me the total credits_used.
```

If you build your own agent, give it a standing rule in its instructions:

```text
Use www.enrich.chat for every email address and phone number. Never write one from memory. If a lookup finds nothing, say not found. Treat catch_all as risky, and never send without my approval.
```

On a plan, the same work for 500 people uses 7,500 credits, which fits the Growth plan at $29 a month. A lookup that finds nothing uses 0 credits, and a key is optional because the free tier needs none. A LinkedIn URL is only a lookup key, and Enrich.chat never logs in to LinkedIn. For the method behind the lookups, read [what data enrichment is](https://enrich.chat/blog/what-is-data-enrichment/) and [waterfall enrichment explained](https://enrich.chat/blog/waterfall-enrichment/). The [guide for AI agents](https://enrich.chat/agents/) covers the full setup.

Know two limits. Enrich.chat is the data layer, not the agent. It has no built-in sequencing, dialer or Chrome extension and does not write to your CRM, so you pair it with the agent and the sending tool you already use. And no data source is perfect, so keep a person on the messages that matter.

## How do you choose an AI sales agent without wasting money?

Start with one job, a status on every contact and a person who approves messages. Then measure meetings held against your current process before you scale.

1. **Pick one job.** Research briefs and list building carry less risk than autonomous sending. Anthropic advises the simplest design that works.
2. **Ask where contacts come from.** Ask whether each field shows a status. If a vendor cannot show a status per email, expect bounces.
3. **Ask what one billable unit is.** It might be a lead, an action, a conversation or a seat. Then ask what happens on a miss.
4. **Run in approval mode.** Artisan's page says you can approve before sending, lock tone and set banned phrases. Ask any vendor for the same controls.
5. **Test against a control group for 30 days.** Count meetings held, bounces and spam complaints, not messages sent. Gartner's July 2026 advice is to look past time saved and measure capacity and commercial results.
6. **Check the rules before you send.** They apply to agents as they do to people.

The [FTC's CAN-SPAM guide](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) says the law makes no exception for business-to-business email. It sets penalties of up to $53,088 per email, requires you to honor an opt-out within 10 business days, and says both the company whose product is promoted and the company that sends the message may be held responsible. For calls, the [FCC ruled in February 2024](https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf) that AI technologies that generate human voices count as an artificial voice under the TCPA, so those calls need the called party's prior express consent. This is general information, not legal advice. See [TCPA and B2B cold calling](https://enrich.chat/blog/tcpa-b2b-cold-calling/) for the call side.

Anyone can opt out of Enrich.chat data at [enrich.chat/opt-out/](https://enrich.chat/opt-out/), and you stay responsible for TCPA, CAN-SPAM and GDPR compliance in your own outreach. When you are ready to build, start with [Claude for sales: prospecting workflows and prompts](https://enrich.chat/blog/claude-for-sales/), or give any agent verified contacts with [Enrich.chat for AI agents](https://enrich.chat/agents/).

## Frequently asked questions

### What is an AI sales agent?

An AI sales agent is software that does a sales task on its own. A language model picks the steps and calls tools to research accounts, find contacts, draft outreach, handle replies or book meetings. It differs from a fixed workflow, where code sets every step, a split [Anthropic's engineers](https://www.anthropic.com/engineering/building-effective-agents) describe. Ask a vendor which steps the model chooses and which are fixed code.

### What is the difference between an AI SDR and an AI sales agent?

An AI sales agent is the broad category. An AI SDR is one kind, aimed at the sales development rep's job: finding prospects, contacting them and booking the first meeting. Other agents work on research, inbound leads, CRM data or renewals, as [ZoomInfo's guide](https://pipeline.zoominfo.com/sales/ai-sales-agents) lists. Compare tools in our [best AI SDR tools](https://enrich.chat/blog/best-ai-sdr-tools/) guide.

### How much does an AI sales agent cost?

It depends on the billing unit. On October 15, 2026, HubSpot listed its Prospecting Agent at $1 per lead it recommends, 11x listed Alice from $3,750 a month billed annually, and Artisan published no prices. Salesforce lists $0.10 per standard action or $2 per conversation for Agentforce. A data API costs less: Enrich.chat has 10 free lookups a day and plans from $29 a month. See [pricing](https://enrich.chat/pricing/).

### Can an AI sales agent replace a human SDR?

Not fully, on the public evidence. [Gartner predicts](https://www.gartner.com/en/newsroom/press-releases/2026-07-28-gartner-predicts-ai-agents-will-outnumber-sellers-10-to-1-by-2028-yet-fewer-than-40-percent-of-sellers-will-say-agents-improved-productivity) that by 2028 fewer than 40% of sellers will say agents improved their productivity, and in a 2025 Salesforce AI Research benchmark leading agents solved about 35% of multi-turn CRM tasks. Agents handle research, drafting and follow-up well. Keep a person on judgment calls, objections and the meeting.

### Where do AI sales agents get their contact data?

Some bundle a database, some connect third-party providers and some call a data API. HubSpot's Prospecting Agent sources contacts from providers its page names, including ZoomInfo and Apollo. 11x says every plan bundles contact data. An agent should never recall contacts from memory, because models guess. With Enrich.chat the agent calls the [API](https://enrich.chat/agents/) and gets a status on each email.

### Is it legal to use an AI sales agent for cold outreach?

Often yes for email, if you follow CAN-SPAM, which the [FTC says](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business) has no business-to-business exception. AI voice calls are stricter: the FCC ruled in 2024 that AI-generated voices are artificial under the TCPA and need prior express consent. The EU and UK add their own rules. This is general information, not legal advice. See [TCPA and B2B cold calling](https://enrich.chat/blog/tcpa-b2b-cold-calling/).

## Sources

1. [Building effective agents](https://www.anthropic.com/engineering/building-effective-agents)
2. [Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027](https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027)
3. [Gartner Predicts AI Agents Will Outnumber Sellers 10 to 1 by 2028, Yet Fewer Than 40% of Sellers Will Say Agents Improved Productivity](https://www.gartner.com/en/newsroom/press-releases/2026-07-28-gartner-predicts-ai-agents-will-outnumber-sellers-10-to-1-by-2028-yet-fewer-than-40-percent-of-sellers-will-say-agents-improved-productivity)
4. [The Productivity Gap: New Survey Shows 9 in 10 Sellers Are Betting on AI and Agents To Help](https://www.salesforce.com/news/stories/state-of-sales-report-announcement-2026/)
5. [AI Sales Agents: A Complete Guide](https://www.salesforce.com/sales/ai-sales-agent/guide)
6. [Discover Agentforce Sales Agents](https://trailhead.salesforce.com/content/learn/modules/agentforce-sales-agents-quick-look/discover-agentforce-sales-agents)
7. [Agentforce pricing](https://www.salesforce.com/agentforce/pricing/)
8. [AI Sales Agent Guide: Definitions, Types, & Examples](https://pipeline.zoominfo.com/sales/ai-sales-agents)
9. [Prospecting agent](https://www.hubspot.com/products/sales/ai-prospecting-agent)
10. [Alice pricing](https://www.11x.ai/products/alice/pricing)
11. [Artisan plans and pricing](https://www.artisan.co/pricing)
12. [What Is An AI Sales Agent? Implementation, ROI, Best Practices](https://www.apollo.io/insights/ai-sales-agent)
13. [a16z- and Benchmark-backed 11x has been claiming customers it doesn't have](https://techcrunch.com/2025/03/24/a16z-and-benchmark-backed-11x-has-been-claiming-customers-it-doesnt-have/)
14. [CRMArena-Pro (benchmark of LLM agents on sales, service and quoting tasks)](https://arxiv.org/abs/2505.18878)
15. [TheAgentCompany: Benchmarking LLM Agents on Consequential Real World Tasks](https://arxiv.org/abs/2412.14161)
16. [Why Language Models Hallucinate](https://arxiv.org/abs/2509.04664)
17. [Fabrication and errors in the bibliographic citations generated by ChatGPT](https://www.nature.com/articles/s41598-023-41032-5)
18. [Employee Tenure Summary](https://www.bls.gov/news.release/tenure.nr0.htm)
19. [Email sender guidelines](https://support.google.com/mail/answer/81126?hl=en)
20. [What is the Model Context Protocol (MCP)?](https://modelcontextprotocol.io/docs/getting-started/intro)
21. [CAN-SPAM Act: A Compliance Guide for Business](https://www.ftc.gov/business-guidance/resources/can-spam-act-compliance-guide-business)
22. [Declaratory Ruling FCC 24-17: Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts](https://docs.fcc.gov/public/attachments/FCC-24-17A1.pdf)

## Try this prompt

Give your AI sales agent verified contacts

```text
Use www.enrich.chat to find 5 heads of sales at US logistics companies with 50 to 200 employees. Add each work email with its status, keep verified emails in one list and catch_all and unknown emails in another, and return a table with name, title, company, LinkedIn URL and email status. Do not draft or send anything. Tell me the total credits_used.
```

## Related

- [Best AI SDR tools (2026): prices, data and control](https://enrich.chat/blog/best-ai-sdr-tools/)
- [How to build an AI SDR with Claude or ChatGPT](https://enrich.chat/blog/how-to-build-an-ai-sdr/)
- [AI lead generation: what actually works](https://enrich.chat/blog/ai-lead-generation/)