What is a voice AI handoff rate?

The handoff rate is the percentage of calls your AI voice agent cannot finish on its own and escalates to a human, either as a live transfer or a queued callback. It is the single most honest health metric for a production voice agent. A rate that is too high means the prompt, knowledge base, or tool integrations are not carrying the conversation. A rate of zero is usually worse: it means the agent is bluffing through calls it should be escalating, which is how you lose a client's trust in week two. Healthy deployments land somewhere in the middle, and the right target depends heavily on the vertical and how complex the agent's job is.

What handoff rate is normal for an AI voice agent?

For tuned inbound agents doing booking and FAQ work, 8 to 15 percent is a common range across local service verticals like dental, home services, and auto repair. Verticals with regulated or high-stakes conversations run structurally higher: insurance often lands near 15 to 25 percent, and law firm intake can sit at 20 to 30 percent because callers frequently need judgment calls the agent should not make. Outbound qualification tends to run lower because the agent controls the script. Treat the estimator's output as a planning baseline, then measure your real rate from call logs after the first two weeks and tune from there.

How does prompt complexity change the handoff rate?

Scope is the biggest lever you control. A simple agent that greets, answers a tight FAQ set, and books appointments hands off far less because there are fewer places to fall outside its lane. A moderate agent that qualifies leads, checks availability, and writes to a CRM adds tool-failure and edge-case surface. A complex agent handling multi-step workflows, policy questions, or compliance-sensitive language escalates the most, and it should. In this estimator, simple scope multiplies the vertical baseline by 0.7, moderate leaves it unchanged, and complex multiplies it by 1.5. The practical lesson: ship the narrow agent first, watch the logs, then widen scope deliberately.

How do I reduce my handoff rate without hurting quality?

Work the causes, not the number. First, read your handoff transcripts weekly and sort them into three buckets: missing knowledge, missing tool access, and genuinely human judgment. The first bucket is fixed by expanding the knowledge base with the exact questions callers asked. The second is fixed by wiring the action the agent needed, like rescheduling or taking a payment commitment. The third bucket should stay human, permanently. Second, tighten the escalation triggers so the agent hands off early and cleanly instead of struggling first. A fast, graceful handoff at 15 percent beats a messy 8 percent where callers repeat themselves to a confused human.

How are the human hours and staffing model calculated?

Handoff calls per month equal your call volume times the estimated handoff rate. Each handoff consumes human minutes: the default of six covers reading the AI's call summary, picking up the context, and resolving the request. Multiply and divide by sixty and you get monthly human hours, which the tool also expresses as a fraction of a full-time load at 160 hours per month. The staffing recommendation maps those hours to a model: an owner-handled callback queue under 20 hours, a part-time coordinator up to 80, a dedicated human-in-the-loop hire up to 160, and a routed escalation team beyond that. Adjust the minutes-per-handoff input to match your reality.

How does Hermes help agencies manage handoffs?

Hermes is the operating platform agencies use to run voice agents for clients under their own brand, and handoffs are treated as a first-class workflow rather than an afterthought. Every call lands in the built-in CRM with a transcript and summary, so the human picking up a handoff has full context instead of starting cold. Escalations can trigger campaign follow-ups, and you see handoff patterns across every client workspace from one dashboard instead of stitching together Retell, GoHighLevel, Zapier, and spreadsheets. Plans run $149 per month for Starter with 3 workspaces, $399 for Business with 7, and $699 for Agency with 20. Book a walkthrough and we will map your handoff flow together.

free tool

Voice agent handoff rate estimator.

Every voice agent hands some calls to a human. Estimate your handoff rate by vertical and scope, see the human hours behind it, and get the staffing model that actually fits.

/ vertical/ agent scopeMore tools, more edge cases./ calls handled per month/ human minutes per handoffIncludes reading the AI call summary and resolving the request.

/ expected handoff rate

12%

120 calls per month escalate to a human.

human hours

12/mo

fte load

0.08

/ recommended staffing model

Owner-handled callback queue

Route handoffs to a callback queue with an SMS confirmation to the caller. No dedicated hire needed at this volume.

Book a 20-min walkthrough to cut this number

Want the full breakdown for your agency? Book 20 min.

Your inputs are encoded in the URL. Send it to a partner and they see the same numbers.

how it works

Plan the human side before launch.

The most common failure in a voice agent deployment is not the AI. It is the missing plan for the calls the AI correctly refuses to handle. Agencies pitch the automation, land the client, and then discover in week one that 10 to 25 percent of calls escalate to a human who does not exist yet. The client experiences that as broken, even when the agent did exactly what it should. Sizing the handoff load before launch is what separates operators who keep clients from operators who churn them.

This estimator starts from a vertical baseline, because the work differs: booking a dental cleaning is not triaging a legal intake call. It then applies a scope multiplier, since a narrow FAQ-and-booking agent hands off far less than one running multi-step, compliance-sensitive workflows. Multiply the resulting rate by your call volume and by the minutes a human spends per escalation, and you get the real monthly hours, plus the staffing model those hours justify, from an owner-handled callback queue up to a routed escalation team.

Use the output two ways. Internally, it tells you what to hire or automate next, and whether tightening the prompt or expanding the knowledge base pays for itself in saved hours. Externally, it is a pricing tool: human-in-the-loop coverage is a line item your agency should charge for, not absorb. Hermes gives every handoff a transcript, summary, and CRM record under your brand, so the human side runs as cleanly as the AI side, on plans from $149 to $699 per month.

frequently asked

Common questions.

next step

Run the handoff loop on one platform.

One platform. Your brand. From $149/month. Book a 20-min private demo and we will map this to your agency.

Book a 20-min private demo Apply for the Founders' Beta

Voice agent handoff rate estimator.

Plan the human side before launch.

Common questions.

Run the handoff loop on one platform.

/ Keep reading