Runnr.ai vs HiJiffy: two WhatsApp-first tools, one boutique hotel

Both of these tools open their pitch with the same three words, “WhatsApp-first AI”, and both mean it. Runnr.ai vs HiJiffy is not the usual case of one tool doing a thing well and the other faking it. They are direct rivals, built around the same channel, sold to the same nervous independent hoteliers. Which is exactly why I ended up running both. I chose HiJiffy deliberately and rolled it out across all three of my properties. Runnr.ai I trialled later, at the Cotswolds, because enough people kept describing it as the WhatsApp-first tool to beat that I wanted to see whether it actually beat the one I already had.

This isn’t a spec sheet. We have two full write-ups for that: my HiJiffy review and Thomas‘s Runnr.ai review, written from the perspective of someone who runs technology for fifteen French properties and reads the API docs before the marketing page. This is the narrower question a small operator actually faces. If two tools promise the same thing, where do they part company, and which one belongs at a boutique hotel? For the rest of the field, the other guest communication tools we’ve tested sit alongside these two.

Where they agree

More than either sales deck would like you to know.

Both were built around WhatsApp from day one rather than bolting it on afterwards, and in both cases you can feel the difference. Pre-arrival messages get opened. My HiJiffy open rate at the Cotswolds sits around 85%, against roughly 40% for the emails it replaced, and Runnr’s numbers in testing were in the same territory. Both are European companies that will tell you, accurately, that guest data stays in EU jurisdiction, and both then route the actual AI generation through an American processor underneath. Both put you through Meta’s WhatsApp verification, which is a bureaucratic ordeal regardless of which vendor is holding your hand through it. Budget a fortnight before you send a single message with either.

And both AIs are competent automation rather than intelligence. Ask either one a tidy single question (“is there parking?”, “what time is breakfast?”) and you get a fast, accurate, correctly-translated reply. Ask either the sort of compound message real guests actually send (“can you sort a cake for my wife, move dinner to 8, and is the pool heated?”) and it loses the thread. Both are also thinner inside WhatsApp than the marketing implies. Rich media, interactive buttons, product catalogues, the components you’d want for a proper pre-arrival flow, are limited on both. Runnr’s are constrained because it doesn’t connect to the WhatsApp Business API directly but routes through a third-party messaging provider that sits between it and Meta. HiJiffy connects more directly and is still thin. Either way, a slick multi-step concierge flow turns into a string of plain text messages. Lower your sights a notch for both.

So the overlap is real. The decision lives in three places where they don’t overlap at all.

Where the AI actually differs

Both AIs get things wrong. What matters is how they get things wrong, and this is the sharpest divide between them.

HiJiffy’s failures are the embarrassing-but-visible kind. It occasionally muffs a simple lookup: a guest asked what time our restaurant opened on a Sunday, the information was correctly in the system, and the bot handed back the weekday hours. It falls apart on compound messages. But it flags when its own confidence is low, and that flag is the whole point. My staff see the wobble and step in before the guest is misled. HiJiffy shipped Aplysia 3 in May 2025, a retrieval-augmented setup that grounds answers in your own knowledge base before generating, and colleagues who’ve run it since say the compound-message failures are meaningfully reduced. The bot is imperfect, but it fails where you can see it.

Runnr’s failure mode is the one that worried me, and it’s the one Thomas flags in his review too. It doesn’t just get answers wrong. It makes promises. In testing and in the accounts I’ve heard from other hoteliers, the AI offers guests things the front desk never approved: a late checkout nobody signed off, an upgrade to a room that isn’t free, a recommendation for a restaurant that closed last year. The guest arrives expecting something real, and the desk has no idea it was ever offered. Intercepting a bad reply mid-conversation is harder than it should be, and, for a company that built its whole product around messaging, the alert that a conversation needs attention arrives by email. Email. Nobody at a busy front desk is refreshing an inbox.

For a boutique hotel, where the entire proposition is that the guest feels personally looked after, an AI that quietly writes cheques the hotel can’t cash is a different order of problem from one that gives out the wrong opening hours. A wrong answer is a small dent. A broken promise is a guest standing at your desk feeling let down before they’ve put their bags down.

Where else they pull apart

Integration depth is Runnr’s strongest card, and it’s a real one. This is Thomas’s territory more than mine, so read his review if it’s yours too, but the short version is that Runnr does something HiJiffy doesn’t. It writes conversation summaries back into the guest profile in the PMS. When the front desk opens a reservation, they can see the guest asked about late checkout over WhatsApp and was quoted €30 for it. The webhooks carry enough data to build on: Thomas wired up a CRM automation that tagged guest profiles by conversation topic and it worked on the first attempt. HiJiffy leans the other way. Its website chatbot and campaign tools are good, but changing a campaign or adjusting an integration often means asking their support team rather than doing it yourself, which slows an agile team down. If your priority is data flowing back into your systems, Runnr is ahead and it isn’t close.

On cost, both publish their pricing, which after my Quicktext experience I could kiss them for. But published and cheap aren’t the same thing. HiJiffy runs €99 Basic, around €159 Pro, around €319 Premium, only Enterprise hides behind a call, and it works out to roughly €3 to €5 per room per month. The Cotswolds lands at about €180. Runnr charges a platform fee (€100 Pro, €200 Plus, €400 Enterprise), then WhatsApp messaging on top per conversation, then locks you into a twelve-month contract. Its “from €3/room/month” headline only holds once you have enough rooms to dilute that base fee. A single boutique property doesn’t. At 45 or 60 rooms you’re paying north of €5 and can drift towards €10 a room once messaging is counted. For a small hotel, HiJiffy is plainly the cheaper of the two, and Runnr’s base fee punishes exactly the properties most likely to be reading this.

On provenance, HiJiffy is the steadier bet. It’s a Lisbon company that’s been at this since 2016, serving over 2,500 hotels across 60-plus countries, and its Aplysia AI runs across several models including Mixtral, the European option from Mistral, which at least shows someone thought about the question. Runnr is a 2022 Utrecht startup, a lean team of ten to fourteen, around 500 properties, founders out of Booking.com, and I couldn’t find a single published security certification for it, no SOC 2, no ISO 27001. Marc would have opinions about the Booking.com pedigree, and Anna would want the missing certifications explained before signing anything. For a tool you’re committing to for years, the longer track record and the shorter list of open questions both point the same way.

The morning a guest arrived expecting a room we hadn’t promised

The trial moment that decided it wasn’t a feature. It was a guest at the desk.

During the Runnr trial at the Cotswolds, a guest had messaged before arrival asking whether she could have a room with a view over the garden. The assistant, trying to be helpful, told her it would “arrange” it. Nobody arranged anything. The garden rooms were full that night, the desk knew nothing about the exchange, and the first anyone heard of the promise was the guest mentioning it at check-in, mildly deflated, holding a booking for a room that faced the car park. It was recoverable. My front desk manager is good and she smoothed it over. But I only learned a promise had been made at all because I scrolled back through the conversation that evening, prompted by an email notification I hadn’t seen until then.

Put the same guest in front of HiJiffy and the machine behaves differently. It doesn’t invent an upgrade it can’t guarantee. When it isn’t sure, it flags low confidence and hands the thread to the desk with the full conversation attached, and a human decides whether the garden room is actually free. That’s less clever on paper. In a hotel it’s the difference between a guest who feels looked after and a guest who feels let down at the door. For me that outweighed everything Runnr does better on the integration side, because the integration side isn’t what my guests experience. The front desk is.

What I’d tell a colleague

If you run a small or independent hotel and your priority is WhatsApp done well, a price you can see without a sales call, and an AI that fails where you can catch it, HiJiffy. That’s the one I kept and rolled out across all three properties, and my front desk manager at the Cotswolds volunteered, without being asked, that she spends less time on repetitive messages now. Staff praising software unprompted is the most reliable review you’ll get. HiJiffy has real faults, the dashboard is awkward to move around and the AI still muffs the odd simple question, but the faults are the visible kind.

If you’re an integration-minded operator, you run a modern PMS, and you want WhatsApp conversations feeding back into your guest data and your CRM, Runnr.ai is the stronger tool for that specific job, and I wouldn’t argue you out of it. Thomas rates its integration work above anything HiJiffy offers, and he’s right. But go in clear-eyed about two things: the base fee makes it dear for a small property, and its habit of promising guests things nobody approved is a real risk in a hotel where the guest relationship is the product. Fix the second by keeping a closer eye on the handovers than you’d like to.

What I wouldn’t do is pick on the WhatsApp pitch alone, because on that headline they’re matched. Two tools, same three words. Let the cost, the integration needs, and the way each one fails when it fails decide it. If it’s the money you’re weighing, what these tools really cost lays the numbers out properly.

• James