The Agentic AI Overlay:  Modernizing IT Service Without “Rip and Replace”

Agentic AI is upending the legacy IT service model. Agents can understand employee requests, interact with enterprise systems, and take actions.  Often, there is no need for a ticket.

Salesforce built its own IT Service with this shift in mind. Unlike legacy providers, which have bolted agentic AI on their platforms, we have built ours natively. This means that our platform is highly adaptable, much easier to use, and gets faster resolutions.

The response from the market has been strong. In just a year, Salesforce’s IT Service has signed more than 1,500 customers.  

But enterprises are hesitant to rip and replace an existing ITSM system. That’s understandable. 

Large-scale migrations have historically involved hefty costs, prolonged implementations and organizational disruptions. According to a survey from KPMG, only 14% of the respondents considered their organizations as “top performers” for technology transformations.

There is also growing evidence that agentic AI initiatives may present even more challenges. A report from Gartner forecasts that more than 40% of agentic AI projects will be cancelled by the end of 2027. Some of the reasons include high costs, unclear business value, and insufficient risk controls.

So how do enterprises adopt agentic AI without disrupting what they’ve already built? At Salesforce, we meet customers where they are. A key part of this approach is our agentic AI overlay. This allows IT Service to coexist with other ITSM platforms rather than replacing them. Enterprises can preserve their investment in their systems, data and workflows while leveraging AI agents to improve the service experience. You start with a high-value use case, prove it in weeks, and expand from there. AI-native speed, without betting the whole stack on day one.

Modernize without the overhaul

Change management is often the biggest obstacle to the future of ITSM. After all, an enterprise will likely have spent many years creating workflows, implementing integrations, and building a data model. Replacing all of that is daunting.

But an agentic AI overlay changes the equation. Instead of overhauling your system of record, you keep most of your current technology in place. AI becomes a new way for employees to get help and for service teams to resolve issues.

This can dramatically reduce the scope of the initial project, with much less data to migrate and fewer processes to rebuild. It can also shorten the time-to-value as the focus is only on a small number of use cases. 

The conversation shifts from replacement to value: how do we help you get more out of the systems you already run? That starts with an architecture designed for agents, not retrofitted for them. Employees interact with AI agents directly in the channels where they already work: Slack, Teams, Claude, and more.

As for the back end, AIforce provides a sophisticated system to orchestrate agents to carry out tasks. This can be done using MCP, APIs, skills, plug-ins and developer tools. 

Claudeforce adds another critical layer by combining Claude’s advanced reasoning with Salesforce data, agentic ITSM workflows, business logic and governance. These capabilities allow agents to determine the appropriate next steps and take action across enterprise systems.

How Paige layers on top

Paige, our Employee Service Agent, can sit on top of your existing ITSM system, not in place of it. When an employee asks for help in Slack or Teams, Paige understands the request, pulls what it needs from your current systems, and takes action, whether that’s resetting a password, provisioning access, or kicking off an onboarding workflow. Your system of record stays exactly where it is. Paige handles the employee conversation and the resolution on top of it. Service teams keep their existing tools. Employees get answers without knowing, or caring, which system sits underneath.

Start where the value is fastest

With an agentic AI overly, a good place to start is with a high-volume, tier-1 request where an agent can deliver measurable value quickly. Some common examples include: password resets, application access, employee onboarding or offboarding, account updates, or returns/refunds. These are usually on one channel and involve OOTB (Out-of-the-Box) agents.  

Meanwhile, an organization needs to measure progress. With Salesforce, you can trace the effectiveness of an implementation by using agent telemetry reports. They measure deflection and how agents are performing. There’s also Agentforce Studio, which detects gaps with the current state of an organization’s infrastructure, such as the knowledge or data that support the agents.

This visibility is what helps create a snowball effect. Once customers see measurable results, it becomes much easier to justify the next capability to implement, such as new requests or channels. Over time, new workflows can be introduced, older processes can be retired and more of the service experience can move to the agentic layer.

The result is a platform that’s more adaptable, easier to use, and faster to resolve issues.

The fastest movers start small

The benefits of agentic AI are real: more efficiency, higher productivity, better experiences. But implementing it well usually takes an incremental approach.

An agentic AI overlay lets you keep the systems, data, and workflows you already rely on while introducing agents where they’ll have the greatest impact. Start with a focused use case, measure the results, and expand as confidence grows. You lower the risk of adoption and modernize at your own pace, without tying your AI transformation to a massive rip-and-replace project.

The enterprises that move fastest aren’t the ones attempting the biggest transformation on day one. They’re the ones that start with a clear business problem, prove the value, and expand from there. Salesforce’s IT Service is built to help you do exactly that.