Sellium · Agentforce

Agentforce on WhatsApp:what it handles and when to add a sales agent

If your company already runs on Salesforce, Agentforce is the natural platform for your AI agents. Here you will see how it connects to WhatsApp, what it takes to build an agent that sells, and when to add a specialized one that runs in the same org.

The short answer

How does Agentforce work on WhatsApp?

Agentforce, Salesforce's AI agent platform, connects to WhatsApp through a Service Cloud Messaging channel: an Omni-Channel flow routes each conversation to the agent and hands it off to a person when needed. To sell on WhatsApp with a catalog, customer-specific pricing and orders that land in the ERP, it often makes sense to add a specialized sales agent in the same Salesforce.

The two do not compete. Agentforce gives you the platform and Salesforce's own agents; Sellium is a WhatsApp sales agent, delivered fully configured, that runs inside your org, reads and writes your objects and integrates with your ERP. One possible split: Agentforce for service and Sellium for sales, with a clear handoff rule between them.

The platform

What Agentforce is and what it handles on WhatsApp

Agentforce is Salesforce's AI agent platform. It includes prebuilt agents for service and for sales, such as one that nurtures and qualifies inbound leads, answers questions and books meetings, plus a tool to build your own: Agentforce Builder, where you define the topics the agent covers, its instructions and the actions it can run through flows, Apex or prompt templates.

On WhatsApp, the path Salesforce documents in detail is the service agent: an agent that answers inquiries, resolves what it can and passes the conversation to someone on your team when needed. To get it running, your org needs these pieces:

  • WhatsApp channel: an enhanced Service Cloud Messaging channel, linked to a WhatsApp Business account created from the Salesforce setup flow.
  • Omni-Channel: an inbound flow that sends each conversation to the agent and, when appropriate, to a person or a queue.
  • AI foundation: Einstein turned on and Data Cloud (Data 360) provisioned, which Salesforce's documentation lists as a requirement for generative AI and its trust layer.
  • Licenses: Salesforce defines and prices them for each agent and each channel, so this guide does not include those prices. Confirm them with your account executive before you build the plan.
The project

What building a sales agent on Agentforce involves

Agentforce gives you the building blocks. Using them to build an agent that sells on WhatsApp end to end is a project with these stages, and each one takes time from your team or from a partner:

  • Topics and actions: which inquiries the agent handles, which instructions it follows and which actions it runs: quoting, checking inventory, building and confirming the order.
  • Data: catalog, customer-specific price lists, discounts and payment terms have to live in Salesforce or be within the agent's reach.
  • Integration: prices, inventory and account balances usually live in SAP, Oracle, Tango, Odoo or NetSuite, and the order has to end up there.
  • Tone and language: regional Spanish, voice notes, product shorthand and orders typed in a hurry.
  • Testing: conversations with real cases and edge cases, such as an out-of-stock product, a customer with no credit left or a price that does not match.
  • Maintenance: the catalog changes, price lists get updated and new rules appear. Someone has to own the agent after go-live.

This is not an Agentforce limitation: it is the work any serious sales agent requires, on any platform. The real decision is who does it: your team, a partner that builds it to order, or a specialized agent that already has the sales side solved.

The alternative

When it makes sense to add a specialized sales agent

Sellium is an AI sales agent for WhatsApp that is delivered configured and ready to run, and it runs inside your Salesforce: it reads and writes your objects, respects your profiles and permissions, and connects to your ERP. It is not a platform to build on: it arrives with WhatsApp selling already designed. It is usually the shortest path when these conditions apply:

  • Catalog selling: your customers ask for prices and availability, and want to place the order in the same chat.
  • Customer-specific pricing: account price lists, volume discounts or payment terms the agent has to honor when it quotes automatically.
  • Orders into the ERP: the agent takes the order on WhatsApp and enters it in your system, not in a spreadsheet.
  • Local language: your customers write the way they talk, in text and voice notes, with regional Spanish and the jargon of each trade.
  • Turnkey delivery: you need the agent running, not a construction project. Process mapping, training, integration and adjustments are part of the implementation.
Working together

How Agentforce and Sellium work together in the same org

Vantegrate is a Salesforce ISV and Consulting Partner, and Sellium is designed to join the ecosystem you already have, not to run outside it. Both agents read and write the same Salesforce: accounts, contacts, leads, opportunities and orders. What needs to be designed is how conversations are split.

Here is one way to split the work for companies that sell and support customers on WhatsApp: the Agentforce service agent handles post-sale inquiries and Sellium handles sales. Everything Sellium does on the sales side is on the Sellium page.

ConversationWho handles itWhat stays in Salesforce
A price or inventory question, or a new orderSellium, on the sales numberLead or opportunity, quote and order, also entered in the ERP
Follow-up on a delivered order, or a complaintThe Agentforce service agent or your service teamA Service Cloud case with the conversation history
Negotiating a large accountThe assigned sales rep, with the chat summaryActivity and next step on the account
A customer asking for support in the sales chatSellium hands it off according to the agreed ruleThe conversation and its summary, visible to whoever takes it

Illustrative example. The actual split (which number each agent answers, which topics it covers and at what hours) is defined during process mapping.

The golden rule: every type of conversation has a single owner and a written handoff path. That way a customer never gets two different answers to the same question.

To decide

Questions to choose the right path

Before choosing between building the sales agent on Agentforce, adding a specialized one or combining both, answer these questions with your sales leadership and your Salesforce admin. The answers frame the decision better than any feature list.

  • What does it need to do? The agent can handle service, sales or both. If it is sales, how far: qualify, quote, take the order, collect payment?
  • Where do your prices live? If prices and inventory live in the ERP, integration is a core part of the project, not a detail.
  • Who builds it? Building it on the platform takes admins and developers with time for topics, actions, testing and adjustments, including after go-live.
  • What is turned on? To build on Agentforce, check Service Cloud Messaging, Einstein, Data Cloud and the Agentforce licenses that apply.
  • By when? The timeline changes the math between building and adding something ready-made.
  • What exactly should it cover? Write it down before talking to vendors: the AI sales agent checklist helps you list the processes, systems and rules the agent has to handle.
Step by step

How Sellium joins a Salesforce org that runs Agentforce

The implementation starts from what you already have running in your org and ends with both agents sharing the work without stepping on each other.

1

Process mapping

We map how you sell on WhatsApp, what Agentforce already handles, which numbers you use and which systems hold prices, inventory and orders.

2

Handoff rule

We define which conversations each agent handles, at what hours and when they go to a person, with the context of what was said.

3

Integration

Sellium runs in your org with the permissions you configure and connects to your ERP to read prices and inventory, and to enter the order.

4

Training and testing

The agent learns your catalog, your price lists and the way you talk. We test it with real conversations before opening it to customers.

5

Go-live and tuning

Every conversation is logged in Salesforce, and with that data we tune answers, rules and handoffs during the first weeks.

Benchmarks

Data to size up the decision

Figures with a published source on sales reps' time, response speed and how AI projects perform. None of them is a Vantegrate result.

70%

Of sales reps' time goes to non-selling tasks, such as administrative work and meeting prep

Source: Salesforce, State of Sales (2024)

7x

More likely to qualify a lead when contacting it within the first hour than one hour later

Source: Harvard Business Review (2011)

67%

Success rate of AI bought from specialized vendors; internal builds reach one third of that rate

Source: MIT NANDA, The GenAI Divide (2025)

Caveats: the Salesforce figure comes from a survey of 5,500 sales professionals in 27 countries. The Harvard Business Review figure measures web leads contacted by phone or email, not WhatsApp, and is used as a reference for response speed. The MIT NANDA figure compares generative AI projects in companies, not specific platforms.

Frequently asked questions

Frequently asked questions about Agentforce on WhatsApp

What Salesforce admins and sales leadership usually ask before adding a sales agent.

Does Agentforce work on WhatsApp?

Yes. Salesforce documents how to connect an Agentforce service agent to a WhatsApp channel in Service Cloud Messaging, with an Omni-Channel flow that routes each conversation and hands it off to a person when needed. Which licenses and settings your org needs is defined by Salesforce for each account.

Does Sellium replace Agentforce?

No, they work together. Agentforce is Salesforce's agent platform and Sellium is an agent specialized in WhatsApp selling that runs inside your Salesforce. One possible split: Agentforce for service and Sellium for sales, both writing to the same CRM.

Do I need Data Cloud to use Sellium?

Sellium runs inside your Salesforce and works on your objects: leads, accounts, opportunities and orders. What your specific org needs (edition, objects, ERP integrations) is assessed during process mapping, before the quote, so there are no surprises halfway through the project.

Can Agentforce and Sellium answer the same WhatsApp number?

It depends on how your number is connected to the official API and how you want to split conversations. What is not negotiable is a clear handoff rule: which topics each agent covers, who answers first and when the conversation goes to a person. A simple option is to use separate numbers, one for sales and one for service. We define it during process mapping.

Can I start with Sellium and add Agentforce later, or the other way around?

Yes. Since both work on the same Salesforce, the order does not change your data or your history: what gets adjusted when you add the second one is how conversations are split and the handoff rule between them.

Does Vantegrate work with Agentforce?

Yes. Vantegrate is a Salesforce ISV and Consulting Partner, so we work with the platform's AI roadmap, not against it. When we add Sellium to an org that already uses Agentforce, process mapping covers how the two share the work. More on our Salesforce page.

Add a sales agent to your Salesforce without touching what already works

Tell us what Agentforce handles in your org, how you sell on WhatsApp and which ERP you use. We will show you how Sellium joins the ecosystem, with the handoff rule already worked out. If you want the number, request a quote.