# How to sell AI agents to businesses: a playbook for agencies

> How to sell AI agents to businesses without promising what AI can't deliver: who to sell to first, a 30-minute discovery, the demo, the proposal and honest answers to objections.

- **HTML version:** https://tolky.to/en/blog/how-to-sell-ai-agents-to-businesses
- **Author:** [Marlos Carmo](https://tolky.to/en/author/marlos), Co-founder of Tolky
- **Published:** November 20, 2026
- **Reviewed:** November 20, 2026
- **Language:** en

## Summary

**Selling AI agents to businesses** means finding where conversations stall sales or support, demoing the agent with the client's own data and closing a proposal with scope, timeline and what the AI won't do. Agencies close more when they start with clients who have heavy WhatsApp volume, expensive leads or after-hours demand, and when they promise AI working with people, not AI replacing them.

The demo lands. The agent answers fast, understands the question and even names the right product. Your agency's client leaves the meeting excited, and three weeks later the proposal is still unanswered. When the client finally comes back, it's with a question you didn't expect: "what if the AI says something stupid to my customer?"

For you, running the agency, that stalled deal is expensive: presales hours, a hand-built demo and, worse, the sense that the agency sold a promise. This guide shows how to sell AI agents to businesses in a way that closes: who to sell to first, how to run discovery in 30 minutes, how to demo, what goes in the proposal and how to answer objections honestly.

*Updated in November 2026.*

## What does selling AI agents to businesses mean?

**Selling AI agents to businesses means offering a client software that talks to their customers, understands intent, answers from what the company taught it, takes actions and hands off to a person when the case calls for it, with scope, timeline and accountability defined.** The definition of an [AI agent](https://tolky.to/en/glossary/ai-agent) is in the glossary; what the agency really sells is the outcome of that conversation done well: a lead answered, a question resolved, a visit booked.

The agent is easy to show because any language model answers a generic question well. It's hard to sell because the buyer isn't buying a nice answer. They're buying the assurance that the AI won't make up a price, won't promise a deadline the company can't meet and will pass the conversation to the team at the right moment. The deal closes when the proposal answers those three fears before the client asks.

## Who should you sell to first?

**Sell first to clients who already have lots of WhatsApp conversations, pay a lot per lead or miss conversations after hours, because for them the problem exists today and is easy to measure.** An AI agent for a business that gets ten messages a week is a solution looking for a problem.

Three signals tell you the account is ready:

- **Conversation volume.** The team answers the same questions all day, and the queue grows during campaigns or seasonal peaks.
- **Expensive leads.** The client spends on media, the lead lands on WhatsApp and waits hours for a first reply. The guide on [lead loss on WhatsApp](https://tolky.to/en/blog/lead-loss-on-whatsapp) shows where leads slip away.
- **After-hours demand.** Messages at night and on weekends wait until Monday, when interest has already cooled.

For the agency, the paid media book is the best place to start. You already know what the client pays per lead, and you've already heard the complaint that "the leads are bad." Often the lead is fine and the response is what's slow.

## The 30-minute discovery: which questions qualify a client?

**A 30-minute discovery, with closed questions about volume, response time, repeated questions, systems and who handles exceptions, separates the client ready for an agent from the one who still needs to fix their support process.** Run it before any demo, with the people who operate WhatsApp, not just the owner.

| Discovery question | Answer that signals fit | Answer that calls for caution |
| :--- | :--- | :--- |
| How many WhatsApp conversations per week? | Hundreds, with spikes during campaigns | A few dozen: the gain is small |
| How long does the first reply take? | Hours or the next day | Minutes, with an idle team |
| Which questions repeat? | Price, delivery time, hours, status, booking | Every conversation is a unique technical case |
| Where do the right answers live? | Website, catalog, written policy | In one person's head |
| Which systems does support check? | Calendar, orders or CRM with an API | A loose spreadsheet or nothing |
| Who handles what the AI can't solve? | A team with hours and a queue | Nobody assigned |
| Do they use the official WhatsApp API? | Yes, or willing to migrate | A number on an unofficial tool |

The last row matters more than it looks. A business AI agent on WhatsApp runs on the official API; a number on an unofficial tool can be banned, and the problem lands on whoever sold it. The guide on [how to sell WhatsApp chatbots without breaking Meta's rules](https://tolky.to/en/blog/how-to-sell-whatsapp-chatbots) covers that risk.

Leave discovery with three things written down: the ten most common questions, the moment a conversation should go to a person and the number the client wants to move (response time, leads answered after hours, bookings).

## How do you demo with the client's own data?

**The demo that sells uses the client's real questions, their catalog and a handoff to their team, and shows the agent declining what it shouldn't answer.** A demo with generic data proves the AI can talk; a demo with the client's data proves it can talk about their business.

Build it in three steps:

1. **A lean knowledge base.** Use the website, the price list and the ten questions from discovery. The guide to [training the agent with a knowledge base](https://tolky.to/en/blog/train-ai-agent-knowledge-base) shows how to organize it.
2. **Test the hard questions.** Ask what the agent doesn't know and check that it admits it and hands off. A [conversation simulation](https://tolky.to/en/glossary/conversation-simulation) before the meeting avoids live surprises.
3. **A visible handoff.** Show the conversation reaching a person with its history. That's when the fear of "AI talking on its own" fades; the rules are in [AI to human handoff](https://tolky.to/en/blog/ai-to-human-handoff-7-rules).

In the meeting, let the client type. When they try to "break" the agent and it stays in scope or passes the conversation on, the main objection disappears without you arguing.

## What goes into an AI agent proposal?

**An AI agent proposal needs a closed scope, a timeline per stage, what the AI does and doesn't do, who owns what after go-live and how results will be measured.** A proposal that promises "100% automated support" closes fast and cancels in month three.

| Proposal section | What to write | Why it protects the agency |
| :--- | :--- | :--- |
| Scope | Channels, agent missions (support, sales, booking), allowed topics | Avoids new requests without an amendment |
| Out of scope | What the AI won't answer and when it hands off | Avoids promises the AI can't keep |
| Stages and timeline | Discovery, knowledge base, testing, go-live, follow-up | Gives dates you can hold the client to as well |
| Client responsibilities | Content, approvals, a team for handoffs | Shares the outcome with whoever operates |
| Measurement | Baseline and agreed metrics | Shows results without argument |
| Separate costs | Platform, Meta's per-message fee, out-of-scope changes | Avoids surprise invoices |

The separate costs row is where most proposals go wrong. Meta charges per message on the official API, and that cost isn't the agency's. The guide on [WhatsApp AI agent pricing](https://tolky.to/en/blog/whatsapp-ai-agent-pricing) shows the components, and [how much to charge for a WhatsApp AI agent](https://tolky.to/en/blog/how-much-to-charge-for-whatsapp-ai-agent) shows how to build the agency's price.

For measurement, agree on the baseline before go-live: without it, nobody can prove improvement. The guide to [AI automation ROI](https://tolky.to/en/blog/roi-of-ai-automation-measure-intelligent-agents) lists what to measure before and after.

## Common objections and honest answers

**The most common objections when selling AI agents are fear of mistakes, fear of losing the human touch, doubts about cost and doubts about how much work is left for the team; an honest answer to each sells better than a promise.**

### "What if the AI says something wrong?"

It can make mistakes, and the proposal should say so. What reduces errors is a closed scope, a reviewed knowledge base, testing before publishing and handoff to people. Show how answers are tested and reviewed, as on [trustworthy AI](https://tolky.to/en/trustworthy-ai), instead of promising the AI never errs.

### "My customers want to talk to a person"

Many want a fast answer, and some want a person. The agent handles the simple questions and passes the rest along with the history, so nobody repeats their story. Say which cases always go to a person.

### "What will this cost at the end of the month?"

Show the math: platform, Meta's fee and the agency's service, listed separately. A client who understands how the price is built accepts adjustments when volume grows.

### "Will my team have more work?"

At first, yes: someone has to review conversations and approve answers. Later, the team stops answering the same question and focuses on conversations that need a person. Put that curve in the proposal.

## What does this change for your agency?

**Selling AI agents turns the agency from a media vendor into the team accountable for what happens after the click, which opens recurring revenue and protects the account.** When the lead your media generates gets an answer in minutes, the report no longer ends at cost per lead.

- **A new service for your current book.** Start with paid media clients, where you already know the cost per lead.
- **Implementation and operation as separate lines.** Bill implementation as a project and operation (conversation review, adjustments, reporting) as a monthly retainer.
- **Retention.** An agent running every day, with history in the CRM, builds a relationship a media retainer alone doesn't.
- **A sales team that's ready.** Whoever sells needs to know what the AI doesn't do. Train sales before opening the demo calendar.
- **Business model.** If AI becomes the agency's main offer, the guide to [starting an AI automation agency](https://tolky.to/en/blog/ai-automation-agency-what-it-is-how-to-start) shows how to set up the operation.

## Where does Tolky fit?

**Tolky is an AI conversational CRM: AI agents configured by mission, its own CRM, campaigns and human support in one platform, on the official WhatsApp API, plus webchat, email and Microsoft Teams.** For the agency, that solves the part of the sale that stalls most: showing control.

![Real screenshot of Tolky's Conversations screen, with All, Unread and Paused filters and the list of WhatsApp chats showing each contact's name, time and a summary of the conversation](https://tolky.to/screenshots/conversas-whatsapp-en.png)

- **A demo with no platform cost.** The Individual plan is free and includes webchat with AI, which is enough to show the agent answering with the client's data; WhatsApp starts on the Starter plan. See the [plans](https://tolky.to/en/pricing).
- **Testing before publishing.** Simulations and versions let you test the hard questions before go-live, as the [trustworthy AI](https://tolky.to/en/trustworthy-ai) page shows.
- **Handoff with context.** The conversation goes to the team with its history, and the contact is logged in the CRM, with a pipeline and automatic Follow Up. See the [AI chatbot for WhatsApp](https://tolky.to/en/features/whatsapp-chatbot).

**What Tolky doesn't do:** it doesn't support Instagram or SMS yet (both coming soon), and it doesn't link the originating ad to the sale on its own; that link needs the client's process or a custom integration, to be confirmed in a demo.

## Frequently asked questions

### How do you sell AI to businesses without a technical background?

Start with discovery and a demo built on the client's data, and leave configuration to whoever implements. What sells is understanding the client's problem and promising only what the AI delivers. The technical part can go to a partner or a trained person on your team.

### What's the best niche for selling AI agents?

The best niche is one that already has lots of WhatsApp conversations and expensive leads: clinics, real estate, education, services and retail tend to fit. More important than the industry is the discovery signal: volume, slow replies and repeated questions.

### How long does it take to implement an AI agent?

It depends on scope and integrations. A support agent with a ready knowledge base goes live faster; one that checks a calendar, orders or a CRM takes longer because of testing. Put the timeline per stage in the proposal, not a single deadline.

### Do I need the official WhatsApp API to sell an AI agent?

Yes, to sell safely. The official API is the path Meta allows for businesses using AI in customer service, and a number on an unofficial tool risks being banned. The step-by-step is in [how to get the official WhatsApp Business API](https://tolky.to/en/blog/how-to-get-official-whatsapp-business-api).

### How do you show results after selling the agent?

Agree on a baseline before go-live and track the same metrics every month: first response time, conversations resolved by the AI, handoffs and leads answered after hours. Measured results are what renew the contract.

---

## Conclusion

Selling AI agents to businesses is less about wowing in the demo and more about good discovery, showing the agent with the client's data and writing a proposal that says what the AI won't do. Agencies that sell this way close projects that last and add a recurring revenue line next to media.

[Read how to choose a WhatsApp AI agent](https://tolky.to/en/blog/how-to-choose-ai-agent-for-whatsapp-checklist), [see the plans](https://tolky.to/en/pricing) or [talk to the team →](https://tolky.to/en/contact).

## Your agency can deliver this with Tolky

**The Tolky Partner Program is for agencies that want to bring AI agents, CRM and WhatsApp campaigns to their own client book, with recurring earnings.** On the Solution track, your agency sells, implements and supports clients under the Tolky brand and bills its own implementation and operation services, with the Tolky Sales certification for the people who present and sell; the other tracks are referring Tolky or running the platform under your own brand. Training and certification are free for companies in the program, and the Tolky team supports every initial project in 1:1 sessions.

[Explore the Partner Program →](https://tolky.to/en/partners) Let's talk?

---

This is the Markdown version of the page. The full version, with images and navigation, is at https://tolky.to/en/blog/how-to-sell-ai-agents-to-businesses

Tolky: https://tolky.to
