Salesforce was among the first big tech companies to enter the agentic AI arena with its Agentforce platform, which aimed to revolutionize digital labor by creating powerful, autonomous AI agents. However, being a first mover has brought its own set of challenges. Salesforce now grapples with high research and development costs and the pressure of maintaining a competitive edge. Despite these hurdles, Salesforce continues to leverage its early entry to refine and enhance Agentforce, striving to stay ahead in the rapidly evolving AI landscape.
With Salesforce’s recent naming as the best software company for 2025, there is little doubt that Agentforce remains a strong contender in the agentic arms race. Salesforce’s view is that by embedding AI directly into the CRM, AI is woven into the natural workflow of its users, maximising value, reducing adoption cost, and improving overall efficiency. There are already enough real-world implementations of Agentforce to say that this claim is valid.
However, since Dreamforce 2024, new entrants to the agentic race have begun attacking Salesforce on its home turf – third-party AI tools are building agents to act as admins, developers, and virtual support staff within your Salesforce org. In this article, we explore how Agentforce compares against platform-agnostic agents, along with their pros and cons.
Agentforce: Deeply Embedded AI Rooted in Trust and Security
Salesforce makes a point of its number one value being trust. Trust in its platforms, its features, data stored within it, and now, in AI used by the system. Salesforce employs advanced security measures, including data encryption, access controls, and threat detection, to safeguard sensitive information.
Additionally, Agentforce incorporates topic-based limitations and harm/toxicity detection to prevent inappropriate interactions. This commitment to security and trust not only differentiates Agentforce but also reinforces Salesforce’s reputation as a reliable and secure AI solution provider.
In addition to trust and security, Salesforce has another ace up its sleeve – its position as the market leader for CRM systems. With over a decade of dominance in the CRM market, Salesforce has built a robust ecosystem that seamlessly integrates AI with comprehensive customer relationship management.
This allows Agentforce to leverage rich, contextual customer data, providing more accurate and actionable insights. Additionally, Salesforce’s extensive experience and established trust in the CRM space enhance Agentforce’s reliability and appeal, making it a preferred choice for businesses seeking a unified, data-driven AI solution.
Finally, a key factor in the usefulness of agentic AI is the extent to which it can be embedded into users’ natural workflow. When a separate login or tab for the AI tool is needed, usability and adoption suffer. A study in the US found that user adoption of AI has stalled significantly in 2024 compared to the previous year. The main reason for this is a lack of training; 70% of workers have had less than five hours of training to use AI in their work.
Agentforce taps into this issue by embedding AI directly into existing processes, avoiding the need for extensive retraining or workflow disruptions. This way, companies gain the benefits of agentic AI with a faster and less expensive user onboarding.
Pros of Agentforce
- Salesforce Focus: Agentforce is deeply integrated with the Salesforce ecosystem, allowing for smooth data flow and enhanced functionality. Agentforce extends beyond the CRM to Slack, Tableau, Google Workspace, and integrated platforms via MuleSoft.
- Efficient Workflow: Agentforce embeds agents into users’ natural workflow within Salesforce platforms, allowing companies to get the most out of AI with minimal effort required from the user.
- Enhanced Security: Salesforce places a strong emphasis on security and trust, ensuring robust protection of customer data. Einstein Trust Layer guards against toxicity, bias, and data leakage, so you can ground LLMs without worry.
- User Familiarity: Users already familiar with Salesforce’s interface and tools can quickly adapt to Agentforce, reducing the learning curve. Salesforce admins can spin up new agents without having to code, and existing Flows and Apex can be retrofitted for agents to use as actions.
- Grounded in Rich Customer Data: Natively working with Salesforce’s extensive customer data allows Agentforce to provide more accurate and contextually relevant insights.
- Active Ecosystem: Salesforce’s partner ecosystem is one of the world’s most active, meaning there are countless certified implementation partners and independent software vendors to extend the capabilities of Agentforce.
Cons of Agentforce
- External System Connectivity: Being tied to the Salesforce ecosystem limits agent functionalities against external systems. Integrations do work with Agentforce but require custom development and maintenance.
- Higher Usage Costs: The $2 per conversation price tag is hefty compared to many platform-agnostic services. Special care must be taken on use case design to ensure Agentforce agents break even.
- Risk of Vendor Lock-In: Businesses become more reliant on Salesforce, which can put them at a disadvantage if they wish to diversify their tech stack.
- Complex Configuration: Agentforce agents are highly customizable but require a great deal of configuration to work reliably. The use of Data Cloud features adds to this complexity.
Third-Party Tools: Centralised Command of Platform-Agnostic Agents
As of the beginning of 2025, there are two kinds of platform-agnostic AI agent platforms. One type is integration platforms (iPaaS) that have recently ventured into the agentic space. Examples include Tray.ai, Workato, and Zapier, with their established low-code integration platforms now augmented with natural language processing and LLMs.
Using their existing integration workflow engine together with new features for instructions and guardrails, you can set up agents that act across several systems based on user prompts. These agents operate in a similar way to Agentforce, although they aren’t centered around any particular SaaS platform.
The allure of iPaaS-based agents is clear, and there is already intense competition in that space. Being able to centrally orchestrate agentic workflows is a powerful asset that should not be underestimated. But neither should the risks of something going awry. As with any integrated solution, there are challenges in scoping the agent’s access and permissions, as well as in testing and release management.
These can, of course, be mitigated with proper data governance and integration architecture, but this is considerably harder to do than with a platform-native agent builder like Agentforce.
In addition to iPaaS vendors, LLM providers themselves have begun offering agents to act on your behalf. Examples like OpenAI’s ChatGPT Operator and Anthropic’s Claude Sonnet employ what are dubbed computer-using agents (CUA) to handle multi-step workflows across different websites and platforms.
Instead of API-based integrations, these agents are browser-based, using image and natural language processing to understand their context and jobs to be done, and then clicking their way through websites like a human would.
Evidenced by OpenAI’s collaboration with big names like DoorDash and OpenTable, there is significant potential in CUAs but their viability as an enterprise-level tool remains to be seen.
Pros of Third-Party Tools
- Flexibility: Platform-agnostic solutions can integrate with a wide range of systems and tools, offering greater flexibility for businesses with diverse tech stacks.
- Cost Efficiency: These solutions can be more cost-effective, especially for businesses that do not need the full suite of features offered by platform-specific tools.
- Scalability: Platform-agnostic tools are designed to be scalable and can handle large datasets and complex models without compromising performance.
- Future-Proofing: They can easily adapt to new technologies and models, ensuring long-term viability and reducing the risk of obsolescence.
Cons of Third-Party Tools
- Integration Complexity: Integrating with multiple platforms can be complex and may require additional development resources.
- Security Concerns: Ensuring consistent security across various platforms can be challenging and may pose risks. These may be blockers, especially for enterprise-level customers requiring robust data isolation and security procedures.
- Lack of Specialized Features: These solutions may lack the specialized features and deep integration that platform-specific tools like Agentforce offer.
Final Thoughts: Who Will Win?
The classic consultant answer is “it depends”. In the race between Salesforce’s Agentforce and platform-agnostic AI solutions, the winner will be determined by the specific needs and priorities of your business. Agentforce stands out with its roots firmly set in the Salesforce ecosystem, offering security, trust, and seamless workflow embedding. This makes it an ideal choice for companies heavily invested in the Salesforce ecosystem and seeking a unified, data-driven AI solution.
On the other hand, platform-agnostic solutions like Tray.ai provide greater flexibility, cost efficiency, and scalability, making them attractive for businesses with diverse tech stacks and a need for broad integration capabilities. Ultimately, the choice hinges on whether a business values the specialized, secure, and integrated approach of Agentforce or the versatile and adaptable nature of platform-agnostic AI solutions.
Let us know who you think the victor is in the comments below.
