We are now entering a new era of AI, and it feels like there is very little slowing it down. As the hype of generative AI becomes commonplace, AI agents are now looking to pave the way for the next steps in this technological revolution.
As we know, Salesforce have planted their wheels firmly on the race track with Agentforce, but it sits next to other agentic platforms/tools, two of the most notable being Microsoft Copilot and ServiceNow’s AI agents. So, how does Salesforce compare to two of their biggest competitors? Let’s take a look.
The Fact Files
In order to assess Agentforce against Microsoft Colpilot’s agents and ServiceNow’s agents, it’s important to look at a number of factors. A lot of the fully-fledged tools and features that each platform/tool boasts are not yet generally available, so, for the purposes of this article, I will be comparing proposed capabilities and ease of use as well as the safety features and any guardrails.
Agentforce
Starting off strong with Agentforce which was formally announced at this year’s Dreamforce, it’s clear that Salesforce wants to stress how imperative AI agents are to the future of AI developments.
Agents take generative AI a step further by building upon the fundamentals of chatbots and excelling them with higher intelligence, empathy, and capabilities. Traditional bots are like pre-programmed assistants, so they can only answer questions or queries with answers that they have “stored” – agents use a blend of large language models (LLMs) and generative AI to understand the full context of customer messages. They use the wealth of data that your org has to offer more bespoke answers, providing more of a personal, humanistic service.
Agentforce is the platform that Salesforce’s AI agents are built and operate on, and the Atlas reasoning engine powers it. Salesforce states that there is an “Agentforce in every app”, which just means that Agentforce is available for all the clouds in their product suite, so that’s Service Agents, Sales Agents, Marketing Agents, Commerce Agents, and Platform Agents.
When pitching this new direction for AI development, Salesforce’s CEO Marc Benioff announced that they would be “making a quantum leap in AI” with Agentforce. It does feel like Agentforce is stepping in to fill the gaps that Einstein Copilot Studio fell flat on. Although ambitious, we’ve reported internally that reasons for its stunted growth could include a “lack of readiness for enterprises, the relatively rudimentary output of LLMs, and a lack of fleshed-out use cases.”
That being said, Salesforce have already teased the companies that they have seen success with, such as Royal Bank of Canada and OpenTable, stating that some of these are achieving 90% case resolution through Agentforce.
But how does it compare to two other giants in the market?
Microsoft Copilot
Microsoft Copilot was first announced at the company’s flagship conference Microsoft Ignite at the end of 2023 – the same year that Salesforce’s Einstein Copilot was unveiled. It first became generally available for enterprises on November 1st, 2023.
Fast forward to 2024, and on September 16th, Microsoft announced that they would be releasing their own Copilot agents in Microsoft 365 Copilot. Using the Copilot’s powerful functionality, customers were granted the ability to create and publish agents to a multitude of channels including Microsoft Teams, mobile apps, and websites, appearing right in the user’s workflow.
Very similar to Agentforce, Copilot’s agents allow users to enhance functionality by drawing from an organization’s data. The more (clean) data that these agents can draw from, the smarter and more tuned-in the responses and actions will be.
The potential for expansion should also definitely be noted; you can advance these agents by equipping them with new skills, such as sending emails, updating records, or creating support tickets. From here, you can integrate them into existing systems, allowing for in-the-moment agent work and assistance.
If trust is an area of concern, Microsoft have covered that too. Any data that has been used to ground agents remains securely within the Microsoft 365 trust boundary, adhering to your organization’s labeling and policy guidelines. If you want to extend beyond the Copilot, then your data is automatically protected by Microsoft Azure tenant boundary, with security layers in place to prevent invalid authentication and authorization.
According to Charles Lamanna, the Corporate Vice President of Business & Industry Copilot:
“(Autonomous agents) will fundamentally change business processes. There will be as many agents as there are documents or SharePoint sites in an organization.”
ServiceNow Agents
As a third focus organization in this trio, it can’t be denied that ServiceNow is a company that many people are keeping a watchful eye on.
Following a very similar pattern to both Salesforce and Microsoft, ServiceNow announced the release of their AI agents in September, getting their foot into the door of this transformational technology.
ServiceNow’s agents use advanced reasoning and are grounded in cross‑enterprise data through the Now Platform (ServiceNow’s version of the Salesforce platform, in super simple terms). Essentially, they evolve prompt‑based activity to deep contextual comprehension, “keeping people in the loop for robust oversight and governance”.
With the expectation to “reinvent workflows across the enterprise”, these agents benefit from unified, real-time access to a wide range of knowledge, tools, workflows, and data on the Now Platform, allowing them to unlock sophisticated comprehension skills while offering intelligent and personalized responses.
According to ServiceNow, they envision a future in generative AI “where people act as skilled administrators of an entire team of ServiceNow AI Agents that proactively manage multiple concurrent tasks and end‑to‑end workflows, supported by an agent‑AI interaction layer built into the Now Platform.”
This is why they’ve already spoken about development opportunities for the future, highlighting the possibility of ServiceNow’s agents pulling context from inputs like voice, video, and images while providing personalized, relevant responses, accelerating the rate of efficiency across IT, customer service, and more.
The Comparison
It must be noted that the full extent of these tools is not fully generally available at the moment, so adequate comparisons with opinions and reasoning cannot yet be made. This comparison section will just take a look at the differences between the tools’ capabilities, proposed ease of use, and dedication to safety.
Capabilities
Starting with capabilities, it does feel like Agentforce has the most advanced set of capabilities, which Salesforce did not shy away from demonstrating during Dreamforce.
With a keen emphasis placed on creating just short of magical AI tools with low-code tools, Salesforce has opened up their product to a large user base made up of varying levels of expertise. Not only that, but with the “Agentforce in every app” approach, these advanced AI capabilities are opened up to multiple clouds, multiple use cases, and multiple possibilities.
Microsoft Copilot’s agents have their own impressive set of capabilities. Analogously to Salesforce, Microsoft have developed a Copilot Studio experience designed to empower users of any skill level to build Copilot agents directly within BizChat and SharePoint, two of Microsoft’s most prominent platforms.
ServiceNow’s agents’ capabilities fall into the same realm of what Agentforce and Microsoft Copilot are pledging, however, if before the tools are made available, ServiceNow can provide more context on the possible use cases these agents could be involved in, then it’s likely that the excitement around them will increase.
Proposed Ease of Use
As we know, Salesforce has several Agentforce/AI resources on Trailhead. They have made their AI certifications free, and they even had an Agentforce ‘Launch Zone’ at Dreamforce where professionals could try using Agentforce for the first time.
Despite this, a lot of discussions of what the platform is capable of, how to use it, or even what it is, are leaning towards the skeptical side. Professionals have admitted that they’re still not sure of Agentforce’s potential use cases, with some individuals worrying that it will price out non-enterprise companies, or that it is no more than a “glorified chatbot”.
Salesforce has the right resources out there to get professionals up and running with Agentforce, but perhaps a more relevant sense of clarity needs to be established in the meantime.
Microsoft Copliot’s agents’ proposed ease of use appears to be similar to that of Agentforce but with some notable differences. Firstly, its use cases don’t appear to be as advanced as some of the known ones of Agentforce, and secondly, it appears that Microsoft is very aware that it needs to cater to all skill levels. That’s why they’ve developed a lightweight version of the tool, dedicated to professionals who need a bit more guidance when making their agents.
With ServiceNow, it’s evident that they have done all they can to make their agent platform as accessible as possible. The organization has their own version of Trailhead called Now Learning, which professionals can use to learn about generative AI for free. But perhaps more helpful, they’ve also released the new Now Assist Skills Kit, that enables organizations to build, test, and deploy custom generative AI skills on the ServiceNow platform.
Some of its key features include the ability to create custom GenAI skills for their specific business needs, and then build, test, and deploy these skills and their underlying prompts, select models, assigning these skills to applications.
Dedication to Safety
Data security sits at the heart of all online business processes, and in the realm of AI, this is no different. Salesforce, Microsoft, and ServiceNow have pledged and demonstrated awareness and dedication to safety through guardrails like the Einstein Trust Layer and the Microsoft Azure tenant boundary, but unfortunately, two out of three organizations have experienced some level of data leakage over the last two years.
In 2023, a Microsoft AI research team that uploaded training data on GitHub with the aim to offer other researchers open-source code and AI models for image recognition, inadvertently exposed 38TB of personal data. Within this data were passwords to Microsoft services, secret keys, and over 30,000 internal Teams messages from hundreds of employees.
Additionally, earlier this year it was revealed that thousands of companies are potentially leaking secrets from their internal knowledge base (KB) articles via ServiceNow misconfigurations.
Now, it must be said that data leaks and data breaches, however devastating their impacts are, are a fairly common issue for many software companies – in fact, there was a 20% increase in data breaches from 2022 to 2023. That being said, as the possibilities of generative AI become more advanced and agents begin to learn from more refined data sets, the prevention of data leaks and breaches will need to be paramount in continual development efforts.
Final Thoughts
All three platforms boast impressive capabilities that have yet to be explored fully and have their own tools and resources available to make the platforms as easy to understand and easy to use as possible. Guardrails and security layers protect each platform, but we’ve learned that consistently working on trust is an extremely important step to success.
The Salesforce Agentforce platform is one that has got heads turning, but it might not be as understood yet as Salesforce had probably hoped. If Salesforce works to provide information on Agentforce’s potential use cases as well as how its advanced capabilities can help a range of businesses from SMBs all the way to enterprises, then I feel it has the power to blow its competitors out of the water. We’ll have to see what’s next.