The deflection win: what 40% fewer work orders looks like on paper
The loudest metric in our dashboard is one of the quietest in the industry: deflection.
Deflection is the count of tenant maintenance conversations where Clara walked the tenant through a fix in the first exchange and no work order had to be created. No vendor dispatched, no PO written, no PM tapped on the shoulder. The tenant got their problem solved; the operator got a non-event.
Why deflection matters more than resolution time
Most AI maintenance products are measured on mean time to resolution — how fast can the agent assign a vendor. It is a legacy metric. It presumes the work order was always going to exist. It presumes the only question is how fast you can route it.
A meaningful percentage of tenant maintenance requests are not work orders. They are questions. The garbage disposal is jammed and the reset button is on the bottom of the unit. The smoke detector beeps because the battery is dying. The thermostat is set to heat instead of cool. These are fifteen-second fixes — if the tenant knows where to look.
Historically, your PM became a help desk for these. The inbox filled up. The after-hours service dispatched a handyman for a jammed disposal at double time on a Saturday. The operator paid $110 to have someone press a reset button.
What deflection looks like in the Clara loop
When a tenant texts that their disposal is jammed, Clara does three things in sequence. First, she pulls the unit's appliance inventory from the PMS and identifies the exact model. Second, she searches the manufacturer's manual for the troubleshooting steps. Third, she walks the tenant through the fix — in plain language, one step at a time, asking for confirmation before moving on.
If the steps work, the conversation ends there. Clara logs the interaction, tags it as a deflection, and the work order never gets created. The tenant is unblocked in under five minutes with no truck roll.
If the steps fail, Clara pivots. She asks for a photo, creates the work order, pulls the available vendor roster for the trade, proposes scheduling windows, and writes the outcome to the PMS. That path is the resolution path. Deflection is the path before it.
How we measure it
Two numbers, both on the dashboard:
(1) Deflection rate — conversations where Clara resolved the issue without creating a work order, divided by total maintenance conversations. Measured monthly.
(2) Avoided work order value — deflection count × average cost of a work order in that category. For a jammed disposal on an after-hours dispatch, that is often the full service-call minimum. For routine daytime issues, it is the labor hour.
We do not compare against a synthetic baseline. We compare against the operator's own pre-Clara run-rate, which we pull at onboarding.
What we see in production
Across the first-access cohort, deflection is tracking at 34 to 42 percent of maintenance conversations depending on property mix and age of stock. Newer stock with better appliance documentation deflects higher. Older stock deflects lower but triages faster — Clara still shaves hours off the PM's triage queue even when the fix requires a vendor.
The second-order effect is bigger. PMs who stop drowning in work-order intake have time for the high-leverage work — retention, vendor management, turns. One regional we work with reassigned their after-hours pager entirely after two months of Clara deflection, because the after-hours volume had dropped below the threshold that justified a dedicated on-call.
What this does not mean
Deflection is not a goal-in-itself metric. The goal is tenant experience with lower operating cost. If Clara deflects something she shouldn't — a safety issue, a hidden leak, a compliance matter — the downside is much larger than the savings. That is why our escalation rules are explicit and conservative: gas smell, active flooding, no heat below freezing, anything where the tenant sounds distressed. Safety beats deflection every time.
The takeaway for operators
If you are evaluating an AI maintenance vendor, do not ask how fast they dispatch. Ask what percentage of conversations they resolve without a dispatch at all. That is where the operating leverage hides.
When the work order never gets created, that is the best outcome for the tenant, the PM, and the operating budget. Measure it.
Keep reading
New posts, in your inbox, when they publish.
Join the waitlist to get the weekly and biweekly PropFlow blog — originals only, bylined, operator-centric.