Why Inductio runs on WhatsApp, not an app

When I describe Inductio to people, the reaction is usually the same: “so it’s an app?” It isn’t, and that’s the most deliberate decision in the product. We deliver guided site inductions through WhatsApp, and the longer we build, the more convinced I am that for this job, an app would have been the wrong call.

The short version: induction software is used by people on their first day, often for the only time, frequently on their own phone. Every one of those conditions argues against an app and in favour of a channel they already have.

Day one is the worst possible day to ask for an install

Picture who actually turns up to be inducted. A permanent hire, sometimes. Just as often: an agency worker who got the booking last night, a contractor in for a maintenance shutdown, a seasonal starter on their first job. They have a personal phone, no corporate email, no company device, and somewhere between five and zero minutes of patience for “first, download our app”.

An app means an app store search, a download over whatever signal the site gate has, an account, a password, a verification email they can’t receive because they don’t have a work address yet. Every step sheds people, and the people it sheds first are exactly the high-turnover, here-for-a-week workers whose inductions are most often the rushed ones. The evidence problem starts right here.

A WhatsApp induction starts with a link or a QR code at the gate. The “install base” is the phone already in their pocket. The onboarding flow for our onboarding tool is: send one message.

What WhatsApp gives us out of the box

Choosing a channel everyone already uses isn’t just about removing friction. A surprising amount of product comes built in:

• Zero training. Nobody needs to learn the interface. If you can message your family, you can complete an induction. That matters when your users span every age, language background and level of tech comfort a workforce contains.
• Photos in the flow. Location challenges (“photograph the lockout point”) use the camera button people already know.
• A natural transcript. The conversation is the audit trail: every question, answer and timestamp in order.
• Groups. A cohort being inducted together shares one conversation, the way they’d naturally walk the site together.

We’d have spent months building worse versions of all of this in an app, and then more months getting anyone to install it.

The conversation with IT

Here’s the part I underestimated: not building an app changes the sales conversation, because it changes the procurement one.

An app on workers’ phones (or worse, on company devices) means MDM policies, account provisioning, password resets, a deprovisioning process for every leaver, and a security review of software running inside the estate. For a tool someone uses once, on their first day.

Inductio asks IT for nothing on the device side. There are no accounts to create or revoke, no credentials to manage, nothing to enrol. The service runs on our infrastructure; the site’s team gets a dashboard and the compliance record. The security review still happens, as it should, but it’s a review of our processing, not of new software inside their network.

And because the data questions come next: messages travel over WhatsApp’s encrypted transport; all processing and storage runs on Cloudflare infrastructure; the AI that interprets natural-language answers runs there too, so induction conversations are never sent to a third-party AI provider. Identifiable run data is automatically anonymised within hours of a run finishing. Only the compliance statistics survive. That’s not a settings page, it’s the architecture.

What we gave up

Honesty requires the other column. Building inside WhatsApp means:

• No custom interface. Our UI is text, buttons, images and whatever else a chat can carry. When an induction stage genuinely needs interactivity, like a spot-the-hazard exercise or a code puzzle, we link out to a small web page and bring the result back into the chat. It works well, but it’s a workaround we’d not need with full UI control.
• A platform dependency. We’re building on Meta’s rails, with their rules, their review processes and their pricing. We take that seriously: the engine is channel-agnostic (its first version shipped on Telegram), so the dependency is on a messaging platform rather than on WhatsApp specifically.
• No lock-screen real estate. An app gets push notifications and a home screen icon. We get a chat thread. For a use-once tool this costs us nothing; for a daily-use product it would be fatal.

That last point is the honest boundary of the argument. If your workforce will use a tool every shift (rotas, timesheets, permits), build the app; the install cost amortises. Induction is the opposite case: maximum friction-sensitivity, minimum repeat use. The economics point the other way.

The principle underneath

The design rule we keep coming back to: meet people where they already are. A site induction’s job is to transfer understanding to someone on the most chaotic day of their employment, and every screen between them and that understanding is a place to lose them.

If you want to see what it feels like, the browser demo runs a full induction in two minutes, no WhatsApp required, which is admittedly slightly ironic.