Why we will not give you a ratio
Search "support agents per 1,000 patients telehealth" and you will find confident numbers with no source. We have not found a public, methodologically sound benchmark, and the variance between programs is large: a GLP-1 brand with weekly injections and dose changes generates a different queue from a hair-loss brand shipping a quarterly bottle. So this post does the more useful thing. It lists where tickets come from, then builds a staffing model from assumptions you can see and replace.
The one public signal worth knowing: on its August 10, 2026 earnings call, Hims & Hers said AI now handles about 80% of its support questions. That is a company with years of tooling and millions of subscribers. Plan for less.
Where the tickets come from
| Source | Typical questions | Owner | Can it be deflected? |
|---|---|---|---|
| Shipping and pharmacy status | "Where is my order?", "Did the pharmacy get it?" | Support | Yes, mostly, with live status in the portal |
| Billing and renewals | Surprise renewal, failed card, receipt for HSA or FSA | Support | Partly: renewal notices, card updater, self-serve receipts |
| Dose and how-to-use | Titration schedule, missed dose, injection technique | Clinical (protocol answers) | Partly: plain-language content tied to the patient's actual plan |
| Side effects | Nausea, injection-site reactions, anything alarming | Clinical, with prescriber escalation | No. Route faster; never deflect |
| Cancellation and pause | Cancel, skip a month, change plan | Support | Yes: in-portal cancel and pause |
| Account and other | Login, address change, eligibility questions | Support | Partly |
Two notes on the table. Shipping status is usually the biggest category, and it is almost entirely a visibility failure; the pharmacy-status post lays out the five states a patient should see. And cancellation is a category you should make easier to complete, not harder. The FTC's July 2026 complaint against Hims & Hers (which Hims denies) named hidden cancellation as a practice, and a patient who cannot find the cancel button files a chargeback instead.
The staffing model
Every number in this table is an assumption. Replace them with your own as soon as you have a month of tagged tickets.
| Assumption | Value | Why we picked it |
|---|---|---|
| Tickets per active patient per month, no self-serve | 0.5 | Placeholder for a GLP-1 style program with monthly shipments and dose changes |
| Category mix | Shipping 35%, billing 20%, dose 15%, side effects 10%, cancel 10%, other 10% | Our assumption; tag your own |
| Handle time, support tier | 8 minutes | Placeholder |
| Handle time, clinical tier | 15 minutes | Placeholder; includes chart review |
| Touches per ticket | 1.5 | Back-and-forth before resolution |
| Productive hours per FTE per month | 104 | 160 hours at 65% utilization |
| Deflection with self-serve | Shipping -60%, billing -40%, dose -30%, cancel -50%, other -20%, side effects 0% | Our assumption of what good portal status, billing self-serve and plan-specific content remove |
Run it at 1,000 active patients:
| Without self-serve | With self-serve | |
|---|---|---|
| Support-tier tickets per month | 375 | 195 |
| Clinical-tier tickets per month | 125 | 103 |
| Support hours | 75 | 39 |
| Clinical hours | 47 | 39 |
| Support FTE | 0.72 | 0.38 |
| Clinical FTE | 0.45 | 0.37 |
| Total FTE | 1.17 | 0.75 |
And at 5,000:
| Without self-serve | With self-serve | |
|---|---|---|
| Support FTE | 3.6 | 1.9 |
| Clinical FTE | 2.3 | 1.9 |
| Total FTE | 5.9 | 3.8 |
Read the clinical column. Self-serve barely moves it, because side effects should not be deflected and dose questions only partly can be. That is the honest shape of the work: automation shrinks the support tier and leaves the clinical tier roughly where it was. If a vendor tells you their tool cuts clinical messages by 80%, ask which messages, and who decided they did not need a clinician.
One more thing the FTE math hides. Coverage has a floor. If you promise a response within four business hours, seven days a week, a 0.4 FTE workload still needs a rota covering every day, which is why small brands usually start with the founder plus a part-time contractor and a clinical partner on a per-message rate rather than a hire.
Two tiers, and the line between them
The support tier answers anything that is not medical: order status, billing, receipts, account changes, plan changes, cancellation. Agents work from macros and the patient's record. They never interpret symptoms and never advise on dosing, even when the answer seems obvious.
The clinical tier (a nurse or a clinical support specialist working under the medical director's protocols, with escalation to the prescriber) answers dose questions, side effects and anything a patient describes in their own body. Protocol answers ("your plan says you move to the next dose after four weeks") can be templated. Anything outside the protocol goes to a prescriber.
The escalation rule we would write on the wall: if the patient describes a symptom, the ticket is clinical, whatever category they picked. Patients file side-effect reports under "billing" all the time, usually because they want to cancel because of the side effect. Your routing has to read the message, not the dropdown. The side-effect triage post has the severity tiers and the adverse-event documentation that goes with them.
Cutting tickets that never had to exist
In order of how much they removed in our model:
Live order status in the portal, pushed as notifications. Approved, sent to pharmacy, processing, action needed, shipped with tracking. The patient checks the app instead of writing to you. This is the 60% in the shipping row.
Billing that explains itself. A renewal notice a few days before the charge with the amount and the date, an automatic card updater, smart retries with a clear "update your card" message, and receipts downloadable from the portal (including HSA and FSA itemized receipts). Most billing tickets are surprises.
Cancel, pause and skip in the portal. Fewer tickets, fewer chargebacks, and a pause option that keeps some of the patients who were going to cancel.
Plan-specific dose content. Generic FAQ pages do not deflect dose questions. A screen that says "you are on week three of your 0.5 mg dose; here is what to expect and when your next dose change is" does.
AI for the support tier, with a hard boundary. An agent that can read order and billing status and answer in your brand voice handles routine support well. It should hand anything clinical to a human immediately and say it is doing so. We wrote up what works in production in agentic patient support.
Measure your own number in a month
Tag every ticket with one category from the table above for 30 days. Divide by average active patients and multiply by 1,000. Record handle time per category. You now have your own tickets per 1,000, your own mix and your own handle times, and the model above becomes a forecast instead of an illustration. Re-run it before January, when new-patient volume (and "where is my first order" volume) spikes.
Where Turbopills fits
Most of the deflection in our model is a product decision, and it is where we put the work. Turbopills ships live pharmacy status in the branded portal and mobile app, billing that starts after provider approval with renewal notices, retries and self-serve receipts, cancellation and pause in the portal, patient-to-care-team messaging, and a CRM where tickets sit next to the order and the chart, so a support agent sees what the patient sees. What it does not replace: your clinical tier, your medical director's protocols, and your judgment about response times. For the broader question of which roles a small team keeps, see the two-person telehealth company. We are in private beta and quote per program; put your own ticket mix into the model above and test our deflection claims against it in a demo.
FAQ
How many support staff does a telehealth company need per 1,000 patients? There is no reliable public benchmark. In our labeled model, a 1,000-patient GLP-1 style program needs about 1.2 support and clinical FTEs without self-serve tools and about 0.75 with them. Tag a month of your own tickets to get your real number.
What are the most common telehealth support tickets? Order and pharmacy status, billing and renewals, dose questions, side effects, cancellation, and account issues. Status and billing are usually the largest and the most preventable.
Should telehealth support agents answer medical questions? No. Non-clinical agents handle orders, billing and accounts. Dose questions and side effects go to a clinical tier working under the medical director's protocols, with escalation to the prescriber.
Can AI handle telehealth patient support? For non-clinical questions with access to order and billing data, yes, and Hims says AI handles about 80% of its support questions. Anything describing a symptom should route to a clinician, and the patient should be told when that happens.