The fear around automating onboarding is always the same: it becomes impersonal. That fear is justified, but it's pointed at the wrong thing. What a client experiences as impersonal is rarely that an email was sent automatically. It's that they had to tell the same story three times.

This piece walks through the chain as it looks in practice: form, system, confirmation, scheduling. Per step, what should happen and where it goes wrong.

01 / the diagnosisWhy does onboarding feel messy?

Because the information sits in four places at once and is complete nowhere.

Part of it sits in the email exchange in which the appointment was actually made. Part in the head of whoever had the conversation. Part in a form the client half filled in. And part nowhere, because nobody asked.

The result is that the first real appointment starts with a quarter of an hour of reconstruction. To the client that feels like a company that doesn't know who they are, and that's exactly the feeling you were trying to avoid with warm manual emails.

Impersonal isn't an email arriving automatically. Impersonal is that you'd already been given the answer.

02 / the chainWhat happens between yes and the first appointment?

Four steps. They look trivial, and that's exactly why they're rarely designed.

At most companies steps 1 and 4 exist, but separately, and step 2 is missing entirely. Then you have a form that sends an email and a calendar that knows nothing about that form. That's not a chain but two loose pieces with a human tying them together in between, every single time.

03 / the formWhat do you ask, and what don't you ask?

Ask only what you need to make the first appointment useful. Not what might come in handy someday.

This is where most onboardings run aground. A form with eighteen fields gets half filled in or abandoned, and then the email exchange you were trying to avoid starts anyway. A form with six fields gets finished. In some sectors it's not a form but a stack of documents: payslips, policies, an annual statement that doesn't exist yet. Then the work shifts from filling in to chasing, and that pattern is set out for financial planners.

Two rules of thumb that make the most difference in practice:

• Don't ask what you already know. You usually already have name, company and email address from the first contact. A form asking for it again tells the client nothing has been kept track of.
• Don't ask what you'll only use in three months' time. That can wait until later, in a context where the client understands why it's being asked. Now it just costs you your completion rate.

What does belong in every form is one open field with a good question in it. Not "comments", but something like: what should be different in six months compared to now. That one answer is often worth more than the ten structured fields above it, because it's in the client's own words.

04 / the confirmationHow does an automatic email stay personal?

By being specific. Not by sounding warm.

A confirmation email saying "we've received your registration and will be in touch as soon as possible" is empty, and everyone notices. An email saying what's happening now, who's doing it and when, reads like a company that has it under control, even if it was sent automatically.

Three things that make a confirmation concrete:

And the boundary we hold ourselves to: a system may confirm, remind and prepare. It doesn't write a substantive reply on your behalf without you having seen it. Why that's not a technical limitation but a choice is set out in why we don't publish automatically on your behalf.

05 / the schedulingWhy the calendar is the real bottleneck

Because it's the only step where two calendars have to align, and by email that nearly always fails in three rounds.

Let the client choose from what's genuinely free. That sounds obvious and in practice isn't: many companies send three proposals that are already full by the time they arrive. A calendar that's live and where the client books themselves takes on average a few days off the lead time and a handful of messages out of everyone's inbox.

When setting this up, pay attention to two things. Build in a minimum preparation time, otherwise there's an appointment at eight tomorrow morning that nobody knew about. And make the appointment confirmation carry the same information as the form confirmation: what, when, with whom, and where the client needs to go.

The client doesn't notice that it's automated. They notice that it's right.

06 / the boundaryWhat you're better off not automating

Three things, and they have one thing in common: there's a judgement call in them.

The decision whether you say yes. Whoever takes on a client based on a form takes on clients who don't fit. That costs more later than it earns now.

The first substantive reply. A system may prepare, it may not send. The difference between a draft you check and a message that goes out the door on your behalf is exactly the difference between time saved and reputational risk.

The bad news. A rejection, a delay, a price that ends up higher than discussed. That should come from a human, even if it costs more time.

How we build up such a chain in steps is set out on the page your software. Automate what repeats and where no judgement is involved. Leave alone what depends on who's sitting across from you. That boundary isn't technically determined and isn't hard to draw either, but it does need to be drawn before you start building. How to take an existing process apart along those lines is set out in your tools don't talk to each other.

This text was produced with AI assistance and checked and approved for publication by a human.