Lighthouse's AI is built to be hired, not installed

We asked Brett Kohn to explain Lighthouse to someone with no connection to the industry. He calls it a commercial operating platform, and he defines commercial as everything that brings travellers to a hotel: sales, marketing, distribution, revenue management. Everything Lighthouse builds is aimed at helping those commercial teams price, market, and operate more effectively, toward profitable top-line growth. Underneath, it all comes back to data. Hospitality has always had plenty of data that's hard to access, interpret, and use, and Lighthouse has spent 14 years turning that data into insights, recommendations, and actions a commercial team can actually use.

Lighthouse isn't an RMS and it definitely isn't a PMS. He describes a lot of co-opetition in the space: Lighthouse works alongside RMS, PMS, and CRS partners and the brands, cleansing the data and turning it into tools that run inside any of those systems. For the team driving top-line growth, it wants to be the system of action.

Where the opportunity actually is

The illustration came from Jill's own experience, working at a property where the hotel and the spa ran separate PMS systems that didn't talk to each other. Brett used that to get at where he sees AI mattering most. Hotels have lived with fragmented systems and siloed data forever, and everyone has always said they want it all in one place. It hasn't happened, because the integration and data-normalisation work to join those systems has been conceptually possible but realistically too expensive and too slow.

So what happens instead is that people become the integration layer. Someone on the spa system, someone on the hotel system, someone piecing the data together by hand. That's the exact case Brett points to, because it's what AI is good at. Instead of a junior analyst pulling data into one spreadsheet, then another, then trying to combine them, work that's slow and full of mistakes and that nobody enjoys, you train the AI once on what the spa data looks like and what the hotel data looks like, give it the rules, and ask it what you want to know. He adds that this isn't about eliminating the analyst. It's about freeing that analyst from the grunt work so they can do the interesting part, building the strategy to grow revenue.

Ernest as a teammate, not a tool

Lighthouse's launch is Ernest, and the word Brett keeps using is teammate: Ernest is your new colleague, a commercial teammate, not yet another tool or software platform. The look and feel is a chat interface, you're having a conversation with Ernest, but the teammate idea isn't just branding. It changes how you're meant to work with it, you train it and correct it the way you would a person, rather than configuring it like software.

Ernest fills a gap. The general models, Claude, Gemini, ChatGPT, are powerful and let Lighthouse build things it couldn't have six months ago. But there's general AI, and then there's hospitality-specific outcomes, and Ernest is meant to be the last mile between the two. Lighthouse uses existing frontier models, then draws on its network of 80,000 hotels, not to train on those hotels' data or share it between them, but to teach Ernest how a revenue manager or a marketing manager actually thinks when demand spikes or prices shift. Ask Ernest on a Monday morning where to focus, and it knows to look at your pace and pickup, your comp set on pricing, what's happening to demand on your channels.

Intelligence, recommendation, action

Brett breaks down what Ernest does into three things. First, it delivers intelligence: ask how your KPIs are trending or what your channel mix looks like, and because it can connect to all those systems the way a human would, through a login, you don't need deep integrations to get the answer. Second, it gives recommendations, but hospitality-specific ones: you've got three days in the next 90 where your bar rate looks off against the comp set, go look there. Third, it takes action. Within the guidelines you set, it can reset a price in your RMS, take inventory on or offline in your PMS, or build a forecast model and produce a spreadsheet.

The trust question came up, and it was specifically about revenue management: do people using a tool like Ernest just accept its suggestions, or do they still want a human in the loop, where the system recommends and the person in charge checks and approves rather than letting it auto-apply. Brett thinks the human-check instinct softens with AI, and the reason is the teammate idea again. A traditional RMS is often a black box: you don't know why it made a recommendation, so you override it. Ernest can be asked why. Here's the dashboard, here's what happened, here's why I said raise the price 10 percent and not 20. That conversational layer didn't exist before. And you can correct it in plain language, "never make an adjustment of more than 10 percent without asking me," and it remembers, instead of digging through settings menus to change a rule.

Your own instance, not the hotel's

One detail stuck with us. Ernest isn't a single shared instance for the whole hotel, it's a separate one for each user. It works at property, portfolio, or brand level, with enterprise-style access rights, so a property-level user sees and trains a different Ernest than someone at portfolio level, and the two don't have the same access. You train it the way you'd train a new employee, sit down and explain how you want it to operate, and it does that, doesn't forget, doesn't sleep, and speaks every language. That accessibility, Brett thinks, is what will drive the trust that drives adoption.

Ernest launched publicly on June 9, with a growing wait list feeding a beta programme. Alongside the software, Lighthouse has four deployed builder teams it calls Ernest Crews, who spend time in the field getting hotels up to speed and making sure the right connections are in place. The point is to accelerate time to value, because one management company's problems look a lot like the next one's, and Brett thinks hotels that don't move now risk being left behind.

How it connects, and the data underneath

On the technical side, Ernest is a standalone product that connects to existing systems rather than replacing them, and there are three ways it can do that. The traditional one is an API. The newer one is MCP, which more systems are starting to add. But the third works differently: if neither is available, Ernest can act as an actual user, going to the website, app, or tool, logging in with its own credentials, and finding and grabbing the data the way a person would. Brett kept returning to that image: you give Ernest a seat and its own login, because it's an AI agent, a coworker. The systems all have a job to do and do it better working in unison, and that's the role Ernest plays.

He also says AI is only as good as the data underneath it, which is where he thinks Lighthouse is well positioned after 14 years of doing exactly that, giving hotels data they can trust and showing where a recommendation came from. On ownership, the data belongs to the hotel and isn't used to train other hotels, and large customers get enterprise security and the property/portfolio/brand access tiers. A five-room B&B won't learn as fast as the 1,000-room Grand Hyatt running constant pricing experiments, but it still benefits from Ernest's read on how the market, demand, seasonality, and comp-set pricing work.

Where the name came from

The name has a story. Ernest comes from three places. Ernest Shackleton, who led his team through every trial of the Antarctic expedition and got them through. Ernest Hemingway, who spent much of his life in hotels and wrote with such economy, which is why Ernest's answers are meant to be only what you need, not pages of data. And the plain word earnest, meaning trustworthy. Brett said the first choice had actually been Holly, after one of the first female concierges at the Waldorf Astoria in New York, but the SEO was hopeless, too many agentic "Hollys" already out there. At the company's Barcelona-influenced customer conference, Ernest briefly became Ernesto for half a day, which they had to keep an eye on.

One concrete use case

We asked for one concrete before-and-after. Brett's example: a group wants a block of rooms at a $150 rate. You ask Ernest to run the displacement analysis. It knows your cost per available room is $80, which is your floor, the lowest rate you'd accept, it knows what demand looks like and the odds of filling those rooms with transient business, and it comes back with a recommendation to counter at $160 to $170, and offers to write up the statement of work. A year ago a sales manager would be doing all of that by hand, partly on a spreadsheet, pulling numbers from a couple of different systems and calling the revenue manager to guess at demand. It was slow, and because it was manual there was always room for mistakes. Ernest does the assembly so the sales manager can spend their time on the judgment, which is the part that actually has value. Not replacing anyone, he said, empowering everyone.