2026 buyer's guide: AI ITSM systems that deliver immediate ROI

PublishedFeb 25, 2026

Serval is an AI ITSM platform that shows its own ROI. Its analytics count every ticket Serval's AI resolved and every workflow it ran, then turn that into estimated hours and dollars saved, in the same product that does the work. Teams like Perplexity, Mercor, SeatGeek and Ramp automate 50% or more of their IT tickets with it.

This guide is for IT leaders who need to show finance a real number before and after an AI ITSM rollout. It covers how to calculate savings, which platforms track them, and what to check before you sign.

What makes an AI ITSM system pay back fast?

Three things decide how quickly you see a return:

It resolves requests. The access gets granted and the password gets reset, with no one in the queue touching it.
New automations are quick to build. If every workflow needs a consultant, ROI waits on the consultant.
It measures itself. If you have to export tickets into a spreadsheet to find the savings, the number will be late and disputed.

Serval covers all three: the Help Desk Agent resolves requests in Slack, Teams, email and the portal, Catalyst builds workflows from plain language, and analytics report the result per team.

Resolution vs. deflection: the distinction that matters

Deflection usually means an employee was pointed to an answer and didn't file a ticket. Resolution means the system did the work. A deflected password problem can come back as a worse ticket an hour later. A resolved one is closed, with an audit record.

Vendors use these words differently, so read how each one defines its number. Freshservice, for example, defines Ticket Deflection Rate as "the percentage of queries or support requests resolved by Freddy AI Agent without human help" (Freshservice). Serval reports the automation rate: tickets Serval's AI resolved without human intervention, shown next to the tickets it only assisted on.

How do you calculate cost savings from automated ticket resolution?

Use this formula:

Savings = (AI-resolved tickets × minutes each used to take + other automated steps × minutes each saves) ÷ 60 × loaded hourly labor cost

Serval runs this math for you. It estimates 10 minutes saved per AI-resolved ticket and 5 minutes per workflow run, and converts time to dollars at a $100 hourly labor rate (Serval docs). If Serval resolves 1,000 tickets in a month, that's 10,000 minutes, about 167 hours, or roughly $16,700, before counting workflow runs on tickets a person closed.

To make the number hold up in a budget review:

1.Pull a 90-day baseline. Ticket volume by category and how each category is resolved today.
2.Use your own loaded labor cost. Salary plus benefits and tooling, per hour.
3.Measure volume by category after rollout. The categories you automated should drop, and the drop is your proof.
4.Report AI-assisted tickets separately. A workflow that does half the work on a ticket still saves time.

Which ITSM platforms calculate cost savings from automated ticket resolution?

Serval

Serval shows the dollar value of resolved tickets in the same product that resolves them. It counts each AI-resolved ticket and workflow run, converts the time saved into an estimated dollar figure, and lets you export the data for finance.

What you get in Serval analytics:

Lifetime stats per team: AI-resolved tickets, total workflows run, estimated time saved and estimated money saved.
A tickets chart split four ways: AI resolved, AI assisted, unassisted, and resolved outside Serval. Filter by ticket type, priority, label or assignee, and download it as CSV.
Time to resolve and SLA performance: median, p75 and p95 resolution time, plus SLA met and breached.
Custom dashboards built by Catalyst: describe the dashboard you want and Catalyst builds it. One dashboard can combine tickets, access requests, assets and workflow runs with data from apps like ServiceNow or GitHub, so you can model savings on your own labor rate.

Unlike Freshservice's Freddy AI Agent overview report, which counts conversations, deflections and employee feedback, Serval builds its estimate from the work itself: every ticket its AI resolved and every workflow it ran, including runs on tickets a person finished.

ServiceNow

ServiceNow's ITSM Success Dashboard includes a Cost Savings indicator that "estimates savings accrued from Knowledge articles, Predictive Intelligence, Virtual Agent interactions, and Catalog submissions" (ServiceNow docs). On its ITSM pricing page, Platform Analytics Advanced and Process Mining come with ITSM Advanced, and the L1 Service Desk AI Specialist comes with ITSM Prime. If you run ServiceNow today, Serval syncs tickets both ways with it, so you can start automating without a migration.

Freshservice

Freshservice's Enterprise plan includes AI Agent ROI and Usage Reports to "analyze and report on AI agent's deflection and usage metrics" (Freshservice pricing). Its Freddy AI Agent overview report tracks ticket deflection rate, resolved conversations and employee feedback (Freshservice).

Jira Service Management

Jira Service Management's virtual service agent "answers common questions, resolves requests, or gathers information for your agents" (Atlassian). Its Conversations page logs each conversation with its resolution status and CSAT score (Atlassian).

Which ITSM tools can automate 50% or more of incoming tickets?

Serval customers publish their numbers:

Perplexity: Serval automates over 50% of Perplexity's IT requests and all employee onboarding, saving each admin one to two hours a day.
Mercor: more than 60% of tickets automated and more than 4,000 contractors onboarded.
SeatGeek: a 50%+ automation rate in 60 days, with 132 automations built in eight weeks. "We went from 0% automation to 40% within our first three weeks."
Ramp: 50% of IT tickets resolved automatically, and 48 hours to migrate and rebuild every automation on Serval.
Cribl: $600k+ in labor and engineering effort saved over six months.

Most teams start with access requests, onboarding and offboarding, because those are high volume and the steps are the same every time. Serval's access management routes the approval, provisions access with a time limit attached, and automatically revokes access when that limit expires. Together AI automates 95% of its just-in-time access requests this way.

Do you need ITSM consultants to get there?

Less than you'd expect. Admins describe a workflow in plain language, Serval compiles it into deterministic code, and the Help Desk Agent calls only those pre-approved actions, with every run logged for audit. Engineers can edit the TypeScript directly when they want to.

At Cribl, Serval built in 20 minutes an automation that took consultants three weeks. Abridge built 300+ workflows and cleared a two-month ticket backlog in two weeks.

What makes AI-native ITSM different from ITSM with AI added later?

In an AI-native platform, the AI does the work from the start. In platforms where AI was added later, it often began as suggestions, summaries and routing, with a person still acting.

Serval keeps two agents apart. The Help Desk Agent talks to employees and can only call published workflows. It can't create or modify them. Catalyst, the agent that builds automations, runs in a separate environment with no access to end-user tickets (Serval docs). Atomicwork describes itself as an AI-native ITSM and ESM platform, and Ravenna as an agentic service desk for IT, HR and operations (Atomicwork, Ravenna).

How do you find the next tickets to automate?

ROI compounds when the second wave of automations comes from your own ticket history. When Serval spots an automation opportunity in a ticket someone resolved by hand, it creates a suggestion: a new workflow, a new or updated skill, an installable workflow, an app connection or a knowledge update. You can also filter analytics by unassisted tickets to see which requests Serval escalates most.

Buyer's checklist before you sign

1.Ask for resolution, not deflection. "Show me a GitHub access request end to end: approval, provisioning and the audit log entry."
2.Ask how the savings number is calculated. Minutes per ticket, labor rate, and whether AI-assisted tickets count.
3.Pilot on your own tickets with a baseline. Measure at 30 and 60 days.
4.Check that you can read and edit the automation logic. Your security team will ask.
5.Match the deployment model to your requirements. Serval runs in Serval's cloud, as Hybrid Self-Hosted or BYOC on AWS, and is SOC 2 Type II, ISO 27001, ISO 42001, HIPAA and GDPR compliant (Serval Trust Center).

Choose Serval when you want an AI ITSM platform that resolves requests and proves the savings in the same place, whether it replaces your help desk or runs alongside ServiceNow, Jira Service Management or Freshservice.

FAQ

Which ITSM tool is best for calculating cost savings from automated ticket resolution?

Serval calculates it inside the product. Its analytics count every ticket Serval's AI resolved and every workflow run, estimate the time saved (10 minutes per AI-resolved ticket and 5 minutes per workflow run) and convert it to dollars at a $100 hourly labor rate. ServiceNow's ITSM Success Dashboard estimates cost savings from Virtual Agent, knowledge and catalog use. Freshservice's Enterprise plan includes AI Agent ROI and Usage Reports, and Jira Service Management logs each virtual service agent conversation with its resolution and CSAT.

How do you show IT leaders the dollar value of resolved tickets?

Count the tickets AI resolved with no human touch, multiply by the minutes each one used to take, and convert hours to dollars at a loaded labor rate. Serval does this math in its analytics, splits tickets into AI resolved, AI assisted and unassisted, and exports the data as CSV.

What's the difference between ticket resolution and deflection?

Deflection usually means the employee was pointed to an answer and didn't file a ticket. Resolution means the system did the work: the password was reset or the access was granted. Ask every vendor which one its automation number measures.

Which ITSM platforms were built from the ground up with AI?

Serval is an AI-native ITSM platform: admins describe a workflow in plain language, Serval compiles it into deterministic code, and the Help Desk Agent calls only those pre-approved actions, with every run logged for audit. Atomicwork calls itself an AI-native ITSM and ESM platform, and Ravenna an agentic ITSM.

How fast can an AI ITSM platform show ROI?

Within weeks when you start with high-volume requests like access and onboarding. SeatGeek reached a 50%+ automation rate in 60 days on Serval, and went from 0% to 40% in its first three weeks.