Serval’s Funding Makes a Headline; Its Automation Still Needs a Service Desk Test

Wowtale’s headline reports that Serval has raised a $150 million Series C at a $3 billion valuation, positioning its AI automation against ServiceNow. That is the extent of the supplied reporting: there is no excerpt describing the product, customers or implementation results. We can discuss the buying question, but we cannot turn a financing headline into evidence that the software performs a particular service desk task.

Our starting point would be the work, not the competitive framing. Choose a request that is frequent enough to matter and narrow enough to evaluate without exposing the organization to unnecessary risk. Document the current process, including the decisions an analyst makes and the situations that require escalation. Then ask Serval to demonstrate what it proposes to automate. We would distinguish between drafting a response, recommending a decision and taking an action, because those deserve different acceptance criteria.

Permissions should be part of that demonstration rather than a follow-up security discussion. Ask which identity performs an action, how authorization is checked and what happens when the request is incomplete or contradictory. Require the proposed design to identify where human approval remains necessary. We would also ask for a demonstration of an unsuccessful attempt, including what the analyst sees and how the work returns to a controlled queue. A tidy success path is not enough for a service desk evaluation.

The ServiceNow comparison needs similar discipline. Does the proposal involve replacing a system, adding automation around it or moving only selected workflows? The supplied headline does not answer that question. We would require an architecture and responsibility map before discussing migration savings. Identify the intended location of service records, the owner of each integration and the process for investigating inconsistent outcomes. Otherwise, a promised reduction in analyst work could simply move effort into administration and troubleshooting.

For the next step, write a pilot brief before accepting a vendor-led demonstration. Define the permitted actions, excluded cases, baseline workload and evidence needed to proceed. Track the amount of human correction as well as the work completed, and agree on a stop condition if the controls do not behave as expected. Serval’s reported financing is a reason to examine the proposition, not a reason to relax procurement standards. We would let a repeatable workflow test determine whether the product belongs on the shortlist.