# What Is Ticket Deflection? _2026-09-01_ **Categories:** ITSM, Automation Learn what ticket deflection is, how to calculate your deflection rate, what a good rate looks like, and how AI can reduce IT support tickets. Ticket deflection is the share of support requests resolved before they ever become a ticket, through self-service, automation, or AI. You measure it with a simple formula: Deflection rate = (requests resolved without an agent / total requests) x 100. Higher deflection means a smaller queue, not a worse experience. The employee still gets an answer. A human just never has to touch the request. ## Key Takeaways - Ticket deflection is a percentage, and the formula is straightforward: requests resolved without a human agent, divided by total requests, times 100. - The number is only as honest as its definition. A strict definition counts autonomous resolution with no human and no re-open within 72 hours. Loose definitions that count auto-replies or article views inflate the figure by roughly 10 points. - A realistic good rate is lower than vendors imply. Industry average deflection is 20-30%; best-in-class is 40-60%; mature deployments plateau around a 55-65% ceiling (servicedeskagents.com, 2026). - Most deflection tools fail on content, not on people. Employees will self-serve. They stop when the underlying knowledge is stale, unsearchable, or wrong. - AI agents deflect differently from FAQs and chatbots. They resolve the request and execute the action, rather than pointing the employee at an article and hoping. ## What is ticket deflection? Ticket deflection is the percentage of incoming support requests that get resolved without a human agent handling a ticket. When an employee needs a password reset and a self-service flow completes it, that request was deflected. It never entered the queue, and no technician spent time on it. The standard formula is: Deflection rate = (requests resolved without an agent / total requests) x 100. If 1,000 employees reach out in a month and 300 get what they need without a human touching the request, the deflection rate is 30%. The metric measures how much of your inbound volume the front line absorbs on its own, so the human team can focus on the requests that genuinely need judgment. Deflection is not the same as ignoring a request or making it hard to reach IT. A properly deflected request ends in a resolution the employee is satisfied with. Deflection that comes from friction, a portal so painful that people give up, is not deflection. It is abandonment, and it shows up later as shadow IT and repeat contacts. ## How ticket deflection is measured, and why the number is slippery Ticket deflection is the one service desk metric most likely to be quoted dishonestly, because the definition does the heavy lifting. Two teams can both claim "50% deflection" and mean completely different things. Under a strict definition, a deflected ticket means autonomous resolution: the AI or self-service flow fully resolved the request, no human was involved, and the request did not re-open within 72 hours. That last clause matters. A "resolution" the employee reopens the next morning was never a resolution. Under loose definitions, teams count things that are not resolutions at all: an auto-reply that acknowledges the message, a knowledge base article that was merely viewed, a chatbot session that ended without a follow-up ticket even though the employee gave up. These inflate the headline number. Converting a loose vendor figure to the strict, Gartner-style definition typically compresses it by around 10 points. One published 65% deflection claim maps to roughly 55% under the strict definition (servicedeskagents.com, 2026). The practical takeaway: when you measure your own deflection, pick the strict definition and hold to it. When you read a vendor's number, ask what counts as "deflected" before you compare it to your own. ## How ticket deflection works Deflection happens at four points, each further from a human than the last. - Self-service knowledge. The employee searches a [self-service knowledge base](/platform/knowledge-base-generation), finds the right article, and solves the problem themselves. Cheapest to run, but entirely dependent on content quality and search. - Chatbots. A scripted or retrieval bot answers common questions and points to articles. It deflects informational requests but rarely completes an action. - [AI agents](/agents). A conversational AI agent understands the request in natural language, checks who the employee is and what they are allowed to do, and executes the action: resets the password, grants the access, installs the software. - Automated resolution workflows. Defined conditions trigger defined actions with no conversation at all, such as auto-remediating a device that drifts out of compliance. The further right you move, the more you deflect and the less the employee has to do. A knowledge base asks the employee to find and follow the answer. An AI agent just resolves it. ## The need for ticket deflection Three pressures are pushing IT teams toward deflection at the same time. 1. Ticket volume keeps rising. Growing companies generate more requests every quarter as they add employees, apps, and devices. The queue grows whether or not the team does. 2. Headcount stays flat. Few IT organizations get to hire in proportion to volume. So the same team absorbs more, resolution times slip, and the backlog compounds. 3. Employee experience is under scrutiny. IT issues cost real productive time. Employees lose up to two work weeks a year to IT problems (Nexthink, 2020), and much of that is spent waiting on simple requests, not hard ones. A password reset that takes under a minute of actual work can sit in a queue for hours. Deflection is the only lever that eases all three at once. It removes the high-volume, low-judgment requests from the human queue so the team can keep up without growing, and employees get instant answers instead of waiting behind a backlog. ## Main capabilities that drive deflection Not all deflection tooling is equal. The capabilities that actually move the number: - AI-suggested articles. Surfacing the right knowledge at the moment of need, phrased the way the employee asked, rather than making them search a portal. - Conversational bots and agents. Handling the request in natural language inside the tools employees already use, so there is no portal to learn and no form to fill. - Auto-resolution workflows. Executing the action end to end: verifying identity against the identity provider, checking permissions, and completing the task without escalation. - Feedback and content generation. Turning every resolved request into updated knowledge, so the content stays current and the next identical request deflects faster. The distinction that matters most is between answering and resolving. A tool that answers questions raises deflection modestly. A tool that executes actions raises it structurally, because it removes the whole class of requests that need a task done, not just a question answered. ## Common use cases for ticket deflection The best candidates for deflection are high-frequency and low-judgment: - Password resets. The single highest-volume ticket type on most desks, 15-20% of total volume even with single sign-on and multi-factor authentication in place (Gartner, via ScreenMeet/InvGate, 2024). By some estimates a manual reset costs around $70 in fully loaded labor (Gartner, widely cited); self-service drops that to roughly $1-$4 (ScreenMeet, 2024). Ideal for full automation. - Access requests. Granting an employee access to an app or a shared resource, once identity and entitlement are verified, is a rules-based action an agent can complete instantly. - Common troubleshooting. Repeatable, documented steps, VPN (virtual private network) connection issues, printer setup, software installs, deflect well because the resolution path is known and consistent. These are exactly the requests that consume the most L1 (level one) time and require the least expertise. Deflecting them frees the human team for the requests that genuinely need it. ## Deflection strategies compared The three common approaches deflect very different amounts of volume for very different levels of effort. ```json { "_key": "ticket-deflection-b49", "_type": "table", "rows": [ { "_key": "ticket-deflection-b49-r0", "_type": "tableRow", "cells": [ "Strategy", "What it deflects", "Typical deflection ceiling", "Effort and upkeep" ] }, { "_key": "ticket-deflection-b49-r1", "_type": "tableRow", "cells": [ "Static FAQ / knowledge base", "Informational \"how do I\" questions where the employee will read and follow steps", "Low. Held back by search quality and content freshness", "High ongoing content maintenance; low technical build" ] }, { "_key": "ticket-deflection-b49-r2", "_type": "tableRow", "cells": [ "Chatbot", "Common questions and article routing; simple guided flows", "Moderate. Deflects answers, rarely completes actions", "Moderate; needs intent training and scripting upkeep" ] }, { "_key": "ticket-deflection-b49-r3", "_type": "tableRow", "cells": [ "AI agent resolution", "The full request, including actions: resets, access grants, installs, executed and logged", "Highest. Approaches the 55-65% strict ceiling in mature deployments", "Higher integration setup; low manual upkeep once connected" ] } ] } ``` The pattern is consistent: static content and chatbots deflect questions, AI agents deflect the actual work. That is why the ceiling climbs as you move down the table. ## Benefits of ticket deflection Lower cost per resolution. A level-one human-handled ticket costs about $22 on average, and commonly $20-$40 (HDI/MetricNet, cited by ScreenMeet, 2021-2024). Automated resolution runs roughly $1-$4 per request (ScreenMeet, 2024). Every deflected ticket is a direct cost avoided, and costs climb sharply once a ticket escalates to L2 or L3. Faster time to resolution. A deflected request resolves in seconds or minutes, at any hour, with no queue. The employee is not waiting on a technician's availability. IT capacity freed for higher-value work. When repetitive L1 volume resolves itself, the team stops firefighting and moves to infrastructure, security, and strategic projects. The role changes without the headcount changing. ## Common mistakes that hurt deflection rate Teams tend to blame low deflection on employees who "just want to talk to a person." The data says otherwise. Given the option, employees submit around 70% of IT requests through Slack rather than a portal (Slack, 2024). People will self-serve. The tooling is usually what fails them. Most deflection tools fail because the underlying content is outdated, not because employees won't self-serve. Between 20-40% of knowledge base articles go stale without active maintenance (industry benchmarks, 2024-2026). An employee who follows a wrong article, hits a dead end, and files a ticket anyway has been trained not to trust self-service next time. The recurring failure modes reinforce each other: - Weak search. The article exists, but it does not match how the employee phrased the problem, so it never surfaces. Right answer, invisible. - Outdated content. Organizations over-invest in building initial content and under-invest in maintaining it, so accuracy decays exactly where volume is highest. - No feedback loop. Nothing closes the gap between what employees actually ask and what the knowledge covers, so the same misses repeat month after month. Fix the content and the search, and deflection rises without any change in employee behavior. ## The Harmony solution: resolution before a ticket exists [Harmony is an AI-native enterprise service management (ESM) platform](/). Its approach to deflection is different from a bolt-on: instead of pointing employees at content and measuring how many avoid a ticket, Harmony resolves the request autonomously before a ticket is ever created. Harmony's AI agents live inside Slack and Microsoft Teams. An employee sends a message. The agent already knows who they are, what access they have, and what device they are on, because Harmony connects to the identity provider, device management, HR system, and knowledge base before the first interaction. The request resolves in the same conversation. Most requests resolve without a ticket. The ones that need a human escalate cleanly, with full context, into the existing [IT help desk](/platform/ai-service-desk-agent) system of record. Because the agent executes the action rather than suggesting an article, deflection climbs to the top of the realistic range. At Cyera, a $12B cybersecurity company, Harmony reached 48% deflection by week two, from two 60-minute onboarding sessions. That figure grew to 75% sustained deflection at month three. "Harmony was a game-changer for us to support our rapid growth. The transition from our legacy ITSM to Harmony's AI Service Desk agent and asset management platform was quick and seamless." Shay Ankory, Director of Global IT, Cyera. Harmony does not typically require replacing ServiceNow, Freshservice, or Jira Service Management (JSM). Those stay as the system of record. Harmony is the AI resolution layer employees interact with directly, and every action it takes can log a ticket back to the platform the IT team already runs. For the broader shift this represents, see [what AI ITSM is](/insights/what-is-ai-itsm). ## Summary Ticket deflection is the share of requests resolved without a human, measured as requests resolved without an agent divided by total requests, times 100. The honest range is lower than vendor headlines suggest: 20-30% average, 40-60% best-in-class, a 55-65% ceiling under a strict definition. Most teams fall short not because employees refuse to self-serve, but because the underlying content is stale and unsearchable. The strategies that deflect the most are the ones that resolve the request and execute the action, rather than pointing at an article. Start by measuring your current deflection rate honestly, using the strict definition, so you know your real baseline before you try to move it. See how Harmony resolves tickets before they're ever created. [Book a demo.](/demo) ## Frequently asked questions on Ticket Deflection ### What is ticket deflection? Ticket deflection is the percentage of support requests resolved without a human agent handling a ticket, through self-service, automation, or AI. The employee still gets an answer, but the request never enters the queue and no technician spends time on it. ### How do you calculate a ticket deflection rate? Divide the number of requests resolved without an agent by the total number of requests, then multiply by 100. So 300 requests resolved on their own out of 1,000 total is a 30% deflection rate. Use a strict definition: count only requests fully resolved with no human involvement and no re-open within 72 hours. ### What's a good ticket deflection rate for IT support? Under a strict definition, industry average deflection is 20-30%, best-in-class is 40-60%, and mature deployments plateau around a 55-65% ceiling (servicedeskagents.com, 2026). Be skeptical of vendor claims of 50-75%, because converting those to the strict definition typically drops them by about 10 points. ### What tools help improve ticket deflection? A self-service knowledge base with strong search deflects informational requests. Chatbots deflect common questions. AI agents deflect the most, because they resolve the request and execute the action rather than pointing to an article. Auto-resolution workflows handle predictable maintenance with no conversation at all. ### Does ticket deflection hurt employee satisfaction? Done well, no. Deflection that resolves the request instantly, in the tools employees already use, tends to raise satisfaction because people get answers faster than a queue can deliver. Satisfaction only drops when "deflection" is really friction, a painful portal or a dead-end bot that pushes people to give up rather than resolving their request.