Agentforce – Everything you need to know about costs, tools and integrations

“Our vision is bold: to empower one billion agents with Agentforce by the end of 2025. This is what AI is meant to be.” Marc Benioff, CEO of Salesforce, presented Agentforce with this vision in September 2024.

Agentforce is the autonomous AI application from Salesforce. Digital assistants, so-called AI agents, complete tasks independently. They answer questions, draw conclusions, make decisions and interpret correlations. In this way, they increase the productivity of companies.

Predictive AI has been analyzing historical data for decades to predict trends and behaviors. Generative AI goes further: it generates new content based on existing information. Tools such as ChatGPT have made it popular. Agents are the next evolutionary stage. They act independently, imitate human communication and create new things – proactively and with foresight.

Since fall 2024, Salesforce has been relying on Agentforce, a new, powerful AI tool. At the same time, the familiar Einstein remains in place. What is the difference between the two and which tool is the right one?

In short: there is no better or worse. Agentforce and Einstein pursue different approaches and fulfill different tasks.

Einstein is the central AI platform from Salesforce. Based on existing data, it predicts customer behaviour, optimizes dispatch times in campaigns and recommends content that suits the target group. In doing so, Einstein extends individual Salesforce products and processes structured CRM data. If you want to use it, you have to enter specific prompts.

Agentforce, on the other hand, acts independently, proactively proposes measures, implements them and makes decisions – for example when prioritizing tasks. It does this with almost no human intervention. Agentforce processes structured and unstructured data, for example from emails, conversations or other content in natural language. Each AI agent follows a defined task area. Unlike Einstein, Agentforce can be used not only for specific products, but across the entire Salesforce ecosystem.

While Einstein supports employees, Agentforce becomes a colleague.

Artificial intelligence only delivers strong results if the database is right. This requires reliable, high-quality data. One good measure is the Salesforce Data Cloud as data storage. It can be activated free of charge. However, the free version has potential sources of error.

Companies that do not use the data cloud should carefully review their data structure. Superfluous information should be deleted. Every data entry must be validated, otherwise it can be removed.

Experienced Salesforce users in particular often rely on workarounds or work with inconsistent data. This is a problem for AI tools such as Agentforce: they only work with clear, consistent data.

A planned Agentforce implementation can be the ideal occasion for companies to prioritize data cleansing and get rid of legacy issues.

Agentforce is deeply connected to Salesforce systems such as Sales Cloud, Service Cloud, Slack and Tableau and can use their data directly. However, the potential is not just limited to Salesforce. Agentforce can be flexibly connected to external systems such as ERP solutions, cloud services or collaboration tools – via APIs, connectors and integration platforms such as MuleSoft.

The MuleSoft Agentforce Connector creates the bridge between Salesforce and third-party systems. Agentforce can thus retrieve, use and process information from external sources. The MuleSoft Anypoint Platform and the Anypoint Code Builder provide the technological basis for seamlessly integrating the AI agents into existing IT landscapes so that they can interact autonomously with APIs.

MuleSoft guarantees a high level of data security. Data and interfaces can be centrally managed, controlled and secured. In addition, the Einstein Trust Layer protects sensitive information from unauthorized access, both in Salesforce and in external applications.

Thanks to the integration, companies can access third-party data directly from Agentforce. This includes customer, financial and operational data. Access is fast and efficient, even if the data is not stored in Salesforce.

Technically, the reverse should also be possible: Starting Agentforce directly in external systems. As of May 2025, however, there are still no details on practical implementation.

Flex Credits: The consumption-based pricing model

Salesforce is taking a new approach with Agentforce: instead of traditional licensing models, it is relying on a AI consumption-based pricing model for the AI platform. To this end, it has introduced so-called Flex Credits.

Flex Credits ensure that companies only pay for the actions that Agentforce actually carries out. An action can be the updating of customer data records, the automation of complex workflows or the processing of use cases. Each of these actions consumes 20 Flex Credits, which corresponds to 0.10 dollars.

An Agentforce action costs 0.10 dollars.

Companies can buy Flex Credits in packages of 100,000 credits at 500 dollars each. In the Salesforce Digital Wallet, they can then manage their credits and strategically assign them to individual use cases – and thus optimize the impact of their AI agents. The Flex Credit system is set to be expanded to include Agentforce user licenses and add-ons from summer 2025.

With the pricing model, Salesforce promises its customers control and profitability. For the idea of AI agents is to automate processes and reduce or even eliminate human intervention. So if you use Agentforce specifically where there are efficiency gains, you will achieve immediate added value. Ideally, the revenue generated by Agentforce will exceed the costs of use.

What is particularly attractive is that there are no major hurdles to getting started. Companies do not have to pay high license fees to gain initial experience or implement a proof of concept.

At the end of the day, there is a clear cost-benefit analysis. The AI agent takes over tasks that would otherwise tie up human resources – faster, cheaper and for a better customer experience. For this evaluation, Salesforce provides a free Agentforce ROI calculator on its website.

The conversation-based pricing model

The old conversation-based pricing model continues to apply to existing Agentforce contracts: each conversation with Agentforce costs two dollars. This means that as soon as an employee, partner or customer sends at least one message to the AI agent within 24 hours or selects a menu option that does not end the chat, a conversation is charged. Salesforce offers individual price agreements for companies with high conversation volumes.

Until now, developing an AI agent required a great deal of effort and programming knowledge. Agentforce fundamentally changes this: with natural language and just a few clicks, an agent can be created that is directly integrated into Salesforce. Coding knowledge is no longer necessary.

This is made possible by the Agent Builder, an AI-supported low-code tool. It manages data sources and allows you to create an Agentforce agent intuitively and in just a few steps. It also provides a library of actions that an agent can perform.

In the first step, users choose whether they want to adapt one of these existing actions or develop a completely new AI agent. They then define the scope of their agent's actions and give it clear instructions on how it should perform its tasks. To do this, they use ready-made workflows or enter their instructions in natural language on the Agent Builder screen.

A live preview immediately shows whether the agent is responding correctly and fulfilling the tasks.

Decision-makers must be able to understand the specific added value Agentforce offers for their use case. This is where Salesforce's Agentic Maturity Model comes in. It provides orientation, makes progress measurable and shows how far companies have already come in the use of agent-based AI. The framework defines five maturity levels for this purpose.

The framework shows how the agents work, where there is potential for optimization and where the ROI is. This makes it a strategic basis for companies to further develop and scale their agent solutions in a targeted manner.

Agentforce's AI agents are still in their infancy but are already showing their enormous potential. They noticeably increase efficiency, automate processes and ensure fast responses. The result is greater customer loyalty and lower costs while scaling at the same time.

Customers expect fast and uncomplicated communication, anytime and anywhere. If you want to stand out from the competition, you have to take this seriously and deliver. Agentforce offers the right solution, making it a powerful growth driver.