The last post was about the hours I'm not there. This one is about the hours I am, and the job that in any normal medical office belongs to a person at the front: the one who answers the phone, takes the message, knows who's calling, and makes sure it gets to the right place. A telehealth practice of one has no such person, and patients still show up the same way they always have, through every door at once. This post is how software does the front-desk job so that nothing coming in gets dropped.

I want to be precise about "no front desk," because I have a little administrative help and I'm not going to pretend otherwise. What I don't have, and what a practice my size can't justify, is a person whose whole job is to sit at the front and field the constant stream of inbound patient communication. That role, the receptionist, is the one software replaces here. The small amount of genuinely human follow-up, the callback, the difficult conversation, still comes from a person. Everything mechanical in between comes from software.

Patients arrive through every door

Here's the thing about patient communication that you don't appreciate until you're the one catching all of it: patients contact you however they want, and you don't get a vote.

Some email. Some text. Some call and leave a voicemail. The occasional one faxes, because a pharmacy or another office does. Some use the portal the way I designed them to, and many don't. In a traditional office this didn't matter, because every one of those channels terminated at the same front desk, and the same person handled the walk-up, the phone, and the fax with the same brain and the same notepad. The channel was just how the message arrived; the handling was unified by a human sitting in one chair.

Take away that chair and every channel becomes its own separate inbox, and separate inboxes are exactly how things get dropped. A text sits unseen because I was in my email. A voicemail waits because I was in the portal. Five places to check is the same problem the dashboard post was about, and it has the same answer: the first job is to funnel every channel into one.

One inbox for every channel

So every inbound message, no matter which door it came through, lands in one place: a single message view where email, text, fax, and portal messages all show up together, in order, as one stream.

The design principle is that I should never have to remember which channel a patient prefers in order to not miss them. I check one place, and if a patient emailed and then texted a follow-up, both are right there together, not scattered across two apps that don't know about each other. This is the live, incoming side of the patient history from the dashboard post: that post was about looking up what we've sent someone, and this is the same idea pointed at what's arriving right now. Unify the channels, and "did I miss anything from anyone" becomes a question with one place to look instead of five.

Knowing who's reaching out

A front desk knows who's standing in front of it. Software gets a phone number, or an email address, and has to work backward to a person, and that turns out to be harder than it sounds.

Most of the time it's easy: the number or the email matches a patient on file, and the message attaches to their record automatically. The trouble is the edges. Someone who filled out the intake but hasn't finished becoming a patient yet doesn't have a full record to match against. A patient texts from a different number than the one on file. A message arrives from someone I have no record of at all. Each of those is a message that, handled naively, attaches to nobody and effectively disappears, which is the software equivalent of a phone ringing in an empty office.

So the matching tries every identifier it has, the phone number and the email both, and when there's no patient record yet it falls back to the intake submission, so someone who's mid-signup still gets recognized instead of showing up as a stranger. And when it genuinely can't place someone, it doesn't drop the message, it surfaces it as unknown and waiting, because an unidentified patient is still a patient, and the one thing a front desk must never do is let someone reach out and get ignored.

Acknowledge, route, and remember

Strip a front desk down to its essence and it does three things with every message: it acknowledges, it routes, and it remembers. Software can do all three, and it does the third one better than any human ever could.

Acknowledge means the patient knows the message landed. Nothing feels worse than sending a message into silence, so every inbound gets a response, whether that's an instant automatic acknowledgement or a real reply, but never nothing.

Route means the message gets to where it needs to go, which as the coverage post laid out means sorting it: this one is auto-handled, this one needs an admin callback, this one needs me, this one is clinical and can't wait. The routing is the part that used to live in a receptionist's judgment, and a lot of it can be encoded, with the genuinely ambiguous cases pushed to a human rather than guessed at.

Remember means every message, in and out, gets logged to the patient's history, so the next interaction starts with context instead of amnesia. This is where software laps a human front desk, because it never forgets a conversation, never loses a sticky note, and never has to ask "remind me what we discussed." The memory is perfect and permanent, and for continuity of care that is not a small thing.

The machinery patients never see

A good front desk also does the outreach nobody notices when it works: the reminder the day before, the nudge to finish the paperwork, the "just checking in." A practice of one automates that, and it's a large share of the total message volume.

Appointment reminders go out on a schedule. Someone who started the intake and stalled gets a gentle sequence of follow-ups. Logins arrive as magic links, verification codes arrive when they're needed. None of it requires me to remember to send it, which is the only reason it happens reliably, because I would absolutely forget.

There's one unglamorous lesson buried in here that cost me to learn. If the software sends something, the software has to record that it sent it. Early on, the automated messages went out but weren't all being logged to the patient's history, so when a patient replied to a reminder, the conversation looked like it was starting from nothing, with no sign of the message they were replying to. The left hand didn't know what the right hand had done. The fix was simple once I saw it, log every outbound message the same way inbound ones are logged, but the principle is general and easy to miss: a system that acts on a patient's behalf has to remember its own actions, or the humans reading the record later are missing half the conversation.

What still needs a human

None of this is an argument for automating patient communication end to end, and I want to be clear about that, because it would be the wrong lesson to take. Some messages are emotional. Some are from a patient who is frustrated, or scared, or delivering news that deserves a person on the other end. The software's job with those is not to handle them. It's to recognize that they need a human and get them to one quickly.

That's the whole philosophy, and it's the same one running through this entire series. Put the robot on the mechanical work, the acknowledging and routing and logging and reminding, the enormous volume of necessary, unglamorous handling that would otherwise bury a solo doctor. Reserve the human, me or my admin help, for the messages that actually need a human, which are fewer than you'd think but matter more than all the rest combined. A front desk staffed by software isn't about removing people from patient communication. It's about spending the scarce human attention only where it changes the outcome.

Why this matters for a solo practice

A receptionist is the first hire most practices make, and the one a solo telehealth practice can't justify, and yet the receptionist's job, catching every inbound message, knowing who it's from, acknowledging it, routing it, and remembering it, still has to get done. A patient who reaches out and hears nothing back doesn't assume you're busy, they assume you don't care, and they're not entirely wrong to leave.

So software does the job. It funnels every channel into one inbox, identifies who's reaching out, acknowledges and routes and remembers each message, and runs the outbound machinery that a front desk would otherwise run by hand. What's left for a human is small and deliberate: the callbacks and the conversations that need a voice. One doctor and a little help end up covering what would traditionally take a front office, not by working harder, but by making software do the part of the front desk that was never really about being human in the first place.

Next time

The front desk catches everyone who reaches out. The next post is about a harder and more uncomfortable question: who you actually take on. Not everyone who contacts a practice is a good fit for it, and for a solo practice especially, saying yes to the wrong patient is a mistake that costs everyone involved. The next one is about eligibility, screening, and the underrated skill of saying no well.