Artificial Intelligence

Agentforce: Roadmap and Questions in 2025

By Andrew Nerney

Branded content with Titan

“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.

There is no doubt about it, 2025 is the year of the agent! Ever since Agentforce landed with a splash at Dreamforce 2024, it’s been the hot topic in our CRM world. This bold new era of a limitless digital workforce has created a lot of excitement but also stirred up some heated discussion. If Agentforce is to be adopted at scale over the coming year, are you and your organization ready?

In this article, we take a look at the road ahead as well as ask some tough questions. We will cover:

  • How Agentforce and AI work together.
  • Agentforce Roadmap, including highlights of upcoming features.
  • Agentforce and dirty data.
  • Agentforce and third-party integrations.

Understanding How Agentforce and AI Work Together

It’s important to understand where Agentforce fits into the bigger picture of AI innovation.

  • In the past, predictive AI was able to forecast trends and behaviors based on historical data – this technology has been around for decades. 
  • Generative AI took this a step further with the ability to create new content, for instance, texts and images. This AI hit the mainstream with ChatGPT. 
  • Autonomous AI revolutionizes things by providing a form of artificial intelligence that can engage in reasoning and make decisions on behalf of humans. Agentforce is an example of this next-gen technology.

Powered by Salesforce’s Atlas Reasoning Engine and integrated with Customer 360, agents are able to carry out tasks previously reserved for human intelligence. While there are still limits to this technology (and it certainly cannot replace more sophisticated human capabilities), agents can now take their place as digital employees.

An example of Agentforce acting as a virtual “employee” would be the reasoning it carries out based on the instructions you give it. For instance, when building an agent and configuring topics, an admin needs to explain the scope of the project in natural language and the range of actions the agent is authorized to carry out. 

Agents parse this information and deploy interpretive skills to figure out if they are authorized to perform certain actions. This is not so different from a human employee trying to work out what their boss really wants! 

At times, Agentforce’s reasoning is so advanced that it is not always clear what type of AI is powering specific actions. For instance, in the builder below, is it readily evident what each input is calling? And what type of AI is powering the instructions? 

Roadmap for Agentforce: Key Features

2025 will see several notable Agentforce features rolled out. Here are some of the ones you should look out for:

  • Agent Invocable Action: This feature was made generally available in February, and it deserves a special mention. It allows agents to take the leap from simply reacting to taking proactive engagement in workflows. Using this feature, you can set conditions on which invocable actions should take place. This can boost business efficiency and enable agents to make timely interventions. 
  • Custom Agent Template: This one is actually generally available as of March, but it’s still worth sharing. This feature enables your teams to build custom agents from scratch to suit your specific needs. It accelerates the process and empowers builders to weave natural language and manual configuration to create custom agents flexibly and efficiently.  
  • Agent Versioning: Another feature generally available as of March, Agent Versioning gives teams the ability to manage multiple agent versions more seamlessly. Relying on cloning and rollback processes enables teams to test versions without messing with your live environment. 
  • Bot Migration: In April, the Bot Migration Tool will be generally available. This feature will make it easier to migrate existing bots over to Agentforce Service Agent. 
  • Next-Gen Chat: Expected to be piloted in April, this feature is set to modernize the chat experience in websites and apps. This will include a more engaging redesign, giving agents the ability to take actions within apps and boosting agent chat performance.
  • Voice Capabilities: Now this is a good one! Agentforce currently doesn’t have voice functionality, but in May 2025, this is set to change – or we at least should have a pilot version. Agentforce will be offering a range of AI speech options, such as text-to-speech, speech-to-text, and translation. This will, of course, boost user experience and enable agents to operate more flexibly and across a wider range of channels. 
  • Agentforce Mobile SDK: In June, we are expecting to be able to pilot extending Agentforce to mobile apps. This includes both Salesforce and external applications, and will grant users the convenience of accessing Agentforce on the go. 

Agentforce and Dirty Data

We all know that AI is only as good as its data source. So, when looking to adopt Agentforce in 2025, ensuring that you are working with healthy data is crucial. One quick fix is connecting your Agentforce to a full implementation of Data Cloud, which we highly recommend. 

While Data Cloud does need to be enabled for Agentforce to function, Agentforce does not require full Data Cloud implementation (just the zero-dollar SKU that comes with Salesforce Enterprise). What this means is that it is possible to use Agentforce without the most robust functionalities of Data Cloud – which can leave room for error.

If you do not choose to purchase Data Cloud, here are three steps to take that will help you clean up your org:

  1. Do a general health check.
  2. Get rid of anything that is not used and archive redundant fields. 
  3. Check your data validations and identify records that don’t meet these requirements. 

Agentforce and Third-Party Integrations 

This is one of our biggest questions in 2025! While it is relatively straightforward to invoke third-party integrations from within Agentforce, how simple is it to invoke Agentforce from inside an external application? While technically this is meant to be possible, it is still not entirely clear how to carry this out. Ideally, you should be able to engage in conversation with agents from within your favorite third-party AppExchange tools.

Officially, Agentforce can be embedded in any external app using API, Flow, or Mulesoft, but currently, there isn’t a lot of available training on this or public use cases. Salesforce ISV packaging for topics and actions is generally available as of this March, but this seems to mainly address the ability of ISVs to create prompt templates and agent responses. 

One of the most eagerly anticipated innovations this year is the recent release of AgentExchange, which you can think of as a type of AppExchange for Agentforce. This new marketplace is set to see Salesforce partners contribute their own Agentforce solutions. While this is still at a nascent stage, we are keen to see what partners and ISVs will bring to Agentforce this year! 

Doc Gen Use Case: Agentforce and Titan

Speaking of Salesforce ISV innovation for Agentforce, let’s check out a use case from Titan, a no-code platform. Contract renewal is fast-tracked when Agentforce combines with real-time document generation. Given that neither Salesforce nor Agentforce has robust no-code doc gen capabilities, this showcases how ISVs can enhance Agentforce in 2025.

Let me outline a use case, using “Arthur McQueen” as an example.  

  • In Service Cloud, Arthur McQueen has a contract that’s expiring soon. He wants to renew without speaking to a sales representative. So Arthur asks Agentforce for help
  • Agentforce responds with contract details, saving Arthur time.
  • This is where Titan steps in, generating a link to Arthur’s contract. He authenticates via 2FA and is prompted to update his information in Titan. This syncs to Salesforce, where Arthur’s record is updated, and the contract is marked as a draft.
  • Agentforce provides a signature link to the Titan document so Arthur can sign the contract. All information is dynamically updated from Salesforce.
  • The signed contract is sent to Salesforce, where it is marked as complete. If Arthur chats with Agentforce, he will see the contract has been renewed.

If you’re intrigued, check out more Agentforce use cases here.

Final Thoughts

And that’s a wrap on some of the most exciting developments for Agentforce in 2025. We hope you found this article informative. 

Feel free to drop any comments below! 

The Author

Andrew Nerney

Andrew is a Senior Solution Engineer at TITAN.

Leave a Reply