Actabl wants AI that won't make things up

We started the way we start all of these, with the birthday-party test: if you were at a party and someone asked what Actabl does, how would you explain it to them? Jerimi Ford, Actabl's chief innovation officer, was happy to take it. Actabl is an operations platform for hotels, built by combining four established point solutions into something that didn't exist before. He went through them. Transcendent, the company he founded, does asset lifecycle management, the building and its equipment from birth to death, capital budgeting, life safety, water safety, everything in between. Alice runs front of house, concierge, guest services, housekeeping, and messaging. ProfitSword is the BI and operational-budgeting piece, with a patent on normalising data so it can pull from hundreds of different system types and land profitability in one place. Hotel Effectiveness handles labour, benchmarking a hotel's hours and pay against everyone else.

What connects them is the data, not the apps. Take a housekeeper who finds a broken fixture. In most hotels that means a radio call to engineering and a hope that it gets logged. In Actabl it goes straight into the system, because Alice and Transcendent share the same data underneath even though different people use them. That's the whole design: the data funnels into one place, but a finance person never has to touch the maintenance app, because each role keeps its own. Actabl began taking shape about six years ago, when Jerimi's Transcendent was acquired by Alpine Software Group, and the plan became to bundle the best point solutions into a single platform.

The word that didn't come up

Normally the word AI turns up early in our conversations. This time it took a while. A few minutes in we pointed that out, and Jerimi liked it. They don't lead with AI, not because they aren't doing it, they build their own code with it every day, but because he's wary of stapling it onto everything.

Altitude, and why it doesn't guess

At HITEC, Actabl announced Altitude: a chat layer over the data, where an operator asks a question in plain language and gets an answer back. The difference is that it doesn't make things up. With most AI tools, you ask a question, and if the model doesn't know, it fills the gap with a guess. Altitude can only draw on the data you're allowed to see, and it answers from that alone, so it has nothing to invent from. It runs inside Actabl's own system, a chat interface for now, with no MCP or outside model. And because people tend to ask the same question again and again, you can turn an answer into a dashboard: ask for your ADR over the last quarter, say you want it as a trend line, and it builds the visual so the question stays answered. If you run a whole portfolio, you can ask it to compare and benchmark across properties.

It goes beyond handing a model a spreadsheet. A model reads structured data well enough, but Actabl went further: they built in algorithms that show it how the different metrics relate to each other, and they taught it to write its own SQL, the query language used to pull exact figures from a database, against Actabl's data model. That's the part that produces accurate numbers, because the model has been trained on how the data actually works rather than left to guess. Jerimi knows the failure mode well: you hand AI an office spreadsheet, ask it something, it answers with full confidence, and then you do the math yourself and it doesn't add up, because it quietly misread the table. "You can argue with AI for a while" before you get what you want, he said, and the algorithms and guardrails are what stop that. Rob added that the need only grows: the flood of data is going to keep rising, especially once robots start moving around hotels feeding everything back, and without a tool to sort it, it's just noise.

Take a picture of the boiler

When we asked for a real example, Rob Bahl, who had recently joined Actabl as senior vice president of AI asset management after a long career at Marriott, gave a concrete one. Getting a hotel's equipment into Transcendent used to be slow and error-prone, because someone had to enter every asset by hand. Now you photograph the nameplate on a boiler or a chiller, the AI reads it, and it works out what the equipment is. It goes further than logging a label: from that nameplate it builds a profile of the asset, the brand, a Hoshizaki or a Scotsman, the model, the serial number, and pulls in the wider product information, so the hotel ends up with a proper knowledge base for that machine rather than a line in a list. Jerimi built the original product before AI existed, so back then a photo of an asset was just a photo, nothing could look at it and recognise an ice machine. Since this launched earlier this year, the data comes in accurate and enriched, and it sets up everything downstream: capex planning, and procurement integrations where the system can go find the right parts when something breaks.

We asked about the next step: photographing a piece of equipment and having the system diagnose it, tell you what's wrong and how to fix it. They don't have that yet, and Jerimi explained why. Every recommendation the system makes has to be correct and trustworthy, because the stakes aren't abstract. At home he uses AI like that all the time. Commercially, if it makes a wrong call, you've broken a $2,000 compressor, or turned it into a $10,000 problem, "or worse, we hurt someone." So they'd rather work out how to do it responsibly first. Even flagging that something looks wrong and should be checked is something he wants to get right before it ships.

Fewer dashboards, not more

The idea underneath all of it is one Jerimi keeps coming back to: people should know what was done, and why, when, how, and who did it, and the difference it made. It sounds simple, but the point is that they don't want to build more dashboards. Every product in the industry ships a dashboard, sometimes two, and a manager ends up trained on each one, then left to work out what matters across all of them.

His example is IoT sensors. A hotel runs lots of vendors, and they all raise alarms their own way, some in an app, some by email, some by text, so the engineering team is watching three apps and an inbox and nobody's sure who owns what. Actabl pulls those feeds into Transcendent through APIs and, instead of drawing another dashboard, tells a specific person the single highest and best use of their time right now. That insight layer runs across the whole platform, and Actabl switched it on first in Hotel Effectiveness, for labour: the system spots looming overtime, suggests a fix, slot this person in and save $180, you approve it, and it shows you the difference it made. The longer goal is for an agent to handle the repetitive clicks behind the scenes, which Rob said gets people out from behind the desk and back with the guest.

What won't change

In one of the HITEC sessions earlier that day, someone had flipped the usual question. Don't ask what will be different in five years, ask what won't have changed in five years. We put that to Rob and Jerimi. Rob went straight to hospitality, the personal touch people travel for. He loves the technology and the gadgets, but he doesn't look forward to a day when your drink by the pool arrives on a robot that might roll into the water. You want a person to come along, to have that bit of interaction. The basic human condition, he says, is wanting to be connected to other people, and that's what hospitality and travel are for.

One thing he kept coming back to is who builds Actabl. Most of the people there came out of hospitality, so they know what it means to work in a hotel, and how hard it is behind the scenes. He used the duck on the water: gliding on top, paddling like mad underneath. The guest should never see the engineers in the boiler room or the housekeepers working flat out. Getting all of that out of sight is the whole job.