AI receptionist: set it up yourself or have it done for you?

Nobody types "DIY AI receptionist vs agency" into a search box because they care about the label. The real question underneath is more useful: what is your time worth, and how complicated is the business this thing has to understand? Answer those two and the choice mostly makes itself.

Both routes end in the same place: an AI receptionist that answers enquiries in seconds, around the clock, takes details when it cannot help, and books people into your diary. The case for having one at all is mostly settled by speed: enquiries answered within the first minute convert up to 391% better (Velocify, from an analysis of around 3.5 million leads). The open question is who builds it and, more importantly, who keeps it sharp. You, or me.

Until recently there was only one answer, because done-for-you was the only route we offered. This September that changes: Pikul Reception, the self-serve version, opens, which means I now sell both and have no reason to steer you to the wrong one. Here is the comparison I would want if I were on your side of the table.

What setting it up yourself actually involves

The platform is designed so setup is mostly reading and correcting, not writing from a blank page. Step by step:

• A guided wizard reads your website and drafts your AI's knowledge from it: your services, your published answers, the basics of who you are.
• You correct anything stale. The old price, the treatment you dropped, the practitioner who left. This is the step that matters most, because the AI only ever answers from what you approve.
• You connect your channels. Website chat connects with one line of code; socials take a couple of clicks.
• You test it like a difficult customer and fix whatever wobbles.

Most people finish that first pass in about half an hour, with a help library and support behind them for anything unclear. Half an hour gets you live. It does not get you finished, and the difference is worth being straight about.

One design decision does a lot of quiet work here: the AI never guesses. Ask it something outside its approved knowledge and it says it does not know, then takes the person's details so you can follow up. A half-finished setup is therefore safe, just quieter than it could be: no invented prices, no confident nonsense, only gaps that behave like a polite "let me check with the team".

Where DIY setups go quiet

I have read enough transcripts from self-built bots, ours and other people's, to know the three ways they fade. None of them look like failure on day one:

• Thin knowledge. The wizard's draft gets approved as-is and nothing is added from the owner's head, so the AI knows what the website knows: which is never the twenty questions that actually decide bookings.
• Untested edge cases. "What are your hours?" works beautifully because it got tested. "Can you fit me in Friday if I pay today?" was never tried, so nobody knows what happens next.
• Nobody reviewing conversations. Prices move, staff change, and the transcripts that would show you what customers keep asking go unread. The bot does not break; it drifts from useful to merely present.

None of this is hard to fix. It needs a habit: an hour or so a month reading conversations and topping up the knowledge. If you know you will keep that habit, self-serve can suit you very well. If you already suspect you won't, better to admit it now than discover it in November.

Pikul Reception opens this September. Details are on the platform page.

What done-for-you actually involves

The done-for-you route starts with the part no wizard can do: I learn your business personally, in a discovery session of two to four hours. What you always ask first, which enquiries you politely decline, how you handle a price shopper, what must never be said. I unpacked all of it in what "trained on your business" actually means; the short version is that the answers live in your head, not on your website, which is exactly why a scrape cannot draft them.

Everything is then written, trained and tested, and you sign it off before go-live. Not a demo you glance at politely: the actual answers, read and approved by you, so the first customer meets a receptionist you have already vetted.

The scope is wider too. The full AI Receptionist service covers WhatsApp, web chat, Instagram and Facebook DMs, SMS and email. It understands how people actually message: voice notes from the car, photos of the problem, the occasional video. It books straight into your diary with the email captured first, so every appointment has a proper confirmation trail. It never gives medical, legal or financial advice; sensitive cases hand to your human team with the conversation attached. And on the bigger plan there is optional live phone answering with Voice AI, because in plenty of businesses the phone is still where the money calls from.

After launch, the maintenance habit becomes my job rather than yours: monthly conversation reviews, weak answers rewritten, new services and prices added, so the system keeps sharpening instead of drifting.

And the honest difference in one line: the platform never includes me training your AI personally. That is what done-for-you buys. Same engine underneath, different hands on it.

A plain decision framework

Choose self-serve if most of these are true:

• You are comfortable with tools. You have set up a booking system or an email tool before and it did not ruin your week.
• Your service list is simple: you could explain it to a new hire in an afternoon.
• An imperfect answer is low-risk. Mildly annoying, not expensive or dangerous.
• You want to start today-ish, not after a project.

Choose done-for-you if any of these are true:

• Every enquiry carries real money, so one fumbled conversation costs more than a month of the service.
• You work in a regulated or sensitive field: clinics, dental, aesthetics, anywhere a wrong sentence is a real problem. Expectations run high there too: 80% of patients say they want to book any time, from home or their phone (Experian Health, 2025).
• You want six channels live at once, tested, from day one.
• You have no appetite to maintain it, and you know it.

Notice the asymmetry: "most" on one side, "any" on the other. A single strong reason, regulation or high-value enquiries especially, is usually enough to justify done-for-you. Self-serve earns its place when the whole picture is simple.

The hybrid path

This is not a fork you choose once. A sensible pattern is to start on the platform, run it yourself for a few months, and upgrade to the done-for-you service when the enquiries justify it. Everything you built carries over: the knowledge you approved, the channels you connected, the conversations already had. Upgrading adds my discovery, testing and monthly reviews on top of your foundation instead of replacing it.

The platform is also a low-stakes way to find out which kind of owner you are. If the monthly review turns out to be a habit you keep, carry on. If the transcripts sit unread by month two, you have learned something useful about yourself, cheaply.

The real cost is not the subscription

Current prices for both routes are on the pricing page, so I will skip figures and talk about the cost that never appears on an invoice.

The platform costs less cash and more attention. You are the one approving knowledge, testing awkward cases and reading conversations, and that attention is real spending even though nothing gets billed. Done-for-you is the reverse: more cash, and almost none of your attention beyond the discovery session and a monthly check-in.

And the most expensive option of all is the one nobody prices: the unmaintained bot. It sits on your website looking like coverage while its knowledge quietly dates, declining more and more questions it should be answering, taking details for enquiries it could have booked on the spot. The losses never show up anywhere, which is exactly why they compound. Whichever route you choose, the real decision is not who does the setup. It is whose job the maintenance is, agreed in advance.

Who should choose neither yet

One more honest answer. If you get a handful of enquiries a month, neither route is your first move. An AI receptionist multiplies what arrives; it cannot conjure arrivals. Fix attention first: content, visibility, a reason for people to message you at all. I laid out how the pieces fit in attention in, booked clients out: the receptionist is the layer that catches demand, not the layer that creates it. Come back to this page when missed enquiries are a weekly cost rather than a rare event.

I sell both routes, so I have nothing to gain from talking you out of either. What I care about is where you are in six months: a receptionist that is still accurate, still tested and still read by someone is what books clients, far more than who set it up. Pick the route where that is most likely to stay true for you, and if you are torn, start on the platform and let the upgrade path do the rest.