Elizabeth Walsh
Yet one metric showed up consistently across every AI strategy workshop INRY facilitated over the past several months. These workshops spanned healthcare, retail, manufacturing, higher education, engineering, and professional services. It is not a new metric. Most organizations already track a version of it. What is new is how directly AI investment can move it, and how few organizations are using it as the primary anchor for their AI business case.
That metric is deflection: preventing work from being created in the first place.
What ticket deflection looks like in practice
One large IT organization INRY worked with identified that more than 60 percent of Level 1 tickets were resolvable by following a standard operating procedure. Yet every one of those tickets required a human agent to look up the procedure and execute it manually. The knowledge existed and the resolution path was clear, but the work was landing in a queue anyway because there was no mechanism to intercept it before it arrived.
A higher education client was managing between 14,000 and 18,000 IT incidents per month. Analysis revealed that 25 to 35 percent of those incidents were resolvable by instruction alone, with no need for investigation or escalation. The deflection opportunity was massive. The barrier was not technology. It was outdated knowledge articles and low self-service adoption. The AI capability to intercept that work existed, but the content infrastructure to support it had not been maintained.
In an HR environment, a single person was manually triaging approximately 19,000 cases per year. There was no auto-routing or SLA tracking, and no AI governance layer existed. Every case was a manual decision, which meant every routing error added time and cost to a process that was already consuming more capacity than it should.
These are not edge cases. They are representative of what INRY found in every workshop across every industry. A significant percentage of incoming work, in many organizations the majority of Level 1 and routine-complexity cases, reaches a human agent not because it requires human judgment but because there is no structured path to resolution before it arrives.
Why deflection is the right anchor metric for AI ROI
Executives in the workshops consistently asked three questions about AI ROI: How many tickets can we avoid entirely? How many phone calls can we eliminate? How many repetitive questions can employees answer through self-service without ever contacting the help desk?
These questions are smart because deflection is measurable, attributable, and fast. It does not require a 12-month implementation to demonstrate value. Organizations that build a capable AI front door, a virtual agent backed by well-structured knowledge and embedded in the channels employees already use, can start measuring containment rates within weeks of deployment.
“The phone is the most expensive channel we have. Every call that becomes a self-service interaction is a win we can actually measure.”
— VP of HR Operations, INRY workshop participant
That statement captures why deflection anchors effective AI business cases. The cost of a phone call, an agent-handled ticket, or a manually routed case is knowable. The volume of those interactions is measurable. The percentage that AI can intercept is estimable with reasonable confidence before you deploy anything. That combination of known cost, measurable volume, and estimable interception rate produces a business case that holds up in a CFO conversation.
What the deflection metrics framework looks like
Successful AI programs measure deflection across several specific indicators. Self-service containment rate tracks the percentage of inbound requests resolved without agent involvement. Ticket deflection percentage measures how many would-be tickets never enter the queue. Call reduction tracks volume changes in phone channel demand over time. Knowledge utilization measures whether the content infrastructure is performing, meaning whether employees are finding and using the information that exists. Virtual agent adoption tracks whether the front door is being used.
INRY builds these metrics into every workshop from day one because they prevent a common failure mode: AI programs that generate real operational improvement but cannot demonstrate it clearly enough to secure continued investment.
Organizations that establish their measurement framework before deployment can point to specific, attributable results within months. Organizations that try to retrofit measurement after the fact spend months arguing about attribution and often settle for anecdotal evidence. That distinction matters enormously when AI program sponsors are defending budget in the second year.
The capacity conversation that deflection opens
There is a second-order benefit to deflection that does not always make it into the business case but should. When AI intercepts routine work before it reaches skilled employees, those employees do not disappear. They redirect.
This matters because the loudest objection to AI investment in most organizations is still job displacement anxiety. Deflection reframes that conversation in concrete terms. The HR specialist who no longer spends three hours per day triaging cases has three hours per day to do the work the role was designed for: supporting managers, resolving complex employee situations, and contributing to programs that require judgment and experience.
“I want to spend time with my teams, developing people. But instead I'm clicking buttons all day.”
— Frontline HR professional, INRY workshop participant
That describes how capacity is being consumed by work that does not require the skill level of the person doing it. AI does not eliminate that person's role. It returns their capacity to the work that justifies their salary and their institutional knowledge.
This framing, redirection rather than reduction, consistently built organizational support in the workshops in ways that efficiency-only narratives did not. When employees and managers understand that the goal is to free expert time rather than replace expert judgment, the adoption conversation becomes significantly easier.
The third post in this series addresses the dimension that determines whether AI programs scale across the organization or stall after the first pilot: what it actually takes to build the consistency and governance foundation that makes enterprise leaders comfortable moving beyond controlled deployment.
The full INRY eBook, “The AI Advantage: What Executive Workshops Reveal About Generating Real Value from ServiceNow AI,” is available now. It contains the complete findings from 12 workshops, field-tested use case frameworks, and the methodology INRY uses to help enterprise leadership teams build credible AI roadmaps grounded in real outcomes.
Frequently asked questions
What is ticket deflection in AI?
Ticket deflection is the practice of preventing work from being created in the first place, intercepting routine requests before they reach a human agent. Instead of a person looking up and executing a standard procedure, an AI-powered virtual agent backed by well-structured knowledge resolves the request through self-service. Deflection is measured through indicators such as self-service containment rate, ticket deflection percentage, and call reduction.
Why is deflection the fastest AI ROI to demonstrate?
Deflection is measurable, attributable, and fast. The cost of a phone call, an agent-handled ticket, or a manually routed case is knowable; the volume is measurable; and the percentage AI can intercept is estimable before deployment. That combination produces a business case that holds up in a CFO conversation, and organizations can start measuring containment rates within weeks rather than waiting for a 12-month implementation.
How do you measure ticket deflection?
Successful AI programs measure deflection across several indicators: self-service containment rate (requests resolved without an agent), ticket deflection percentage (would-be tickets that never enter the queue), call reduction (phone channel volume change over time), knowledge utilization (whether content is found and used), and virtual agent adoption (whether the AI front door is being used). Establishing this framework before deployment allows attributable results within months.
Does ticket deflection reduce headcount?
Deflection redirects capacity rather than reducing it. When AI intercepts routine work before it reaches skilled employees, those employees redirect their time to higher-value work that requires judgment and experience, such as supporting managers and resolving complex situations. This redirection framing, rather than reduction, consistently builds broader organizational support for AI adoption.