His use of the technology does not stop there. The company has released a number of AI initiatives for customers in the past few months—most recently Cloudbeds Labs—but Harris has been using OpenAI's models internally for over a year and Anthropic’s Claude for about eight months.
That said, the company is “just getting started” with the technology, but Harris has himself gone from using AI a few times a day to hundreds of times per day across both his professional and personal lives.
“Some of that is because I don’t think the tools are far enough along to do what we really want to get to. There are a lot of individuals inside this organization that have been playing with the various agentic search tools for many months now,” he said.
Harris does not see AI taking over the world; he views it as a thought partner to what he calls the “thought leaders” at Cloudbeds, and the autonomous state is still far off in his eyes. But as the technology evolves, the company has been evolving how its teams use it.
“How we can use it not only in our applications but also for our employees, to ultimately get them to start challenging the problem and almost flip it on its back and approach it from where we are trying to get to? And, how do we get there in the most effective way?” he said.
Knowledge is power
One of the biggest challenges Harris wants to address in the company through AI is knowledge transfer, which he said is the biggest bottleneck and bane in any organization.
“There is so much knowledge across my 700+ people in 41 countries, it gets trapped in silos,” Harris said.
“We needed to break that down, and to break that down, we needed to put all of our knowledge in one place, make it accessible, let AI look at it and look at it with the degree of accuracy that might also make some organizations worry and [feel] uncomfortable.”
The discomfort comes from potentially opening up the company’s internal systems when there is sensitive information—such as financial and shareholder information or employee agreements—that is not for all eyes. The goal is to create a “department-level capacity without worrying about cross-organizational complexity and fear of going off the cliff with data leaks or data integrity,” he said.
The solution for Cloudbeds has been to decide which systems—from Slack, email and Zendesk to Salesforce, Confluence and GitHub—to give the AI access to, while also considering who would have access to them.
“We've been able to construct a system, and we're actually going to evolve that system again to something that is very searchable [with] actionable ways of approaching that information,” Harris said.
The evolution of that system, dubbed internally the “RAG (retrieval-augmented generation) knowledge base” will involve centralizing data so that everyone can access it and reach another level of productivity. From there, the company wants to build custom agents for each department that can carry out more tasks and eventually broaden them to a company level so that a piece of information can be accessed or a task set in motion for several different departments.
Harris acknowledges it is not just about the technology—there is also a big change management process to implement simultaneously, because it is not about replacing humans.
“I really want to instill a change management style. I want them to be the thought leaders. I don't want them to stop doing things. I need the humans. That's why I hired them,” he said.
Low-hanging fruit
Beyond addressing the knowledge transfer challenge, Cloudbeds has been able to find some quick wins using AI.
Early productivity gains include a member of the human resources team using the technology to reduce a task related to meetings and follow-ups from half a day down to minutes. And this is just the tip of the iceberg, according to Harris.
The company has also embarked on a more significant project to scan the resumes of the hundreds of thousands of people who have applied for a role at Cloudbeds in the past three years.
“That's a lot of resumes, and we have the systems trying to scan those and look for candidates that weren't a good fit then but could be a great fit now [so] we would not have to have a recruiter go outbound to try to find candidates,” he said.
A further interesting use case has been in exploring ways to make board meetings more efficient. Cloudbeds has trawled through all the questions being asked in those settings and to look for gaps in that information it has to answer those queries.
But it’s not just about efficiency and productivity gains. Harris said the company is also using AI to gauge staff sentiment, something he described as “beautifully powerful.”
“We’re trying to be way more real-time in understanding of sentiment. How happy are they? What are the things we’re missing? We’re getting all sorts of pulses coming back from the organization.”
Harris said he believes the company is in the 1% of organizations using AI this extensively, basing this on the fact they have access to large language models, specific teams building agents for targeted use cases and a significant portion of their engineers using it for coding.
“There is the 0.01% of companies that I strive to get to. That’s the Klarnas of the world who have removed major systems like Salesforce and Workday and gone completely autonomous with internal tools. That is another level, that is obviously a $32 billion company that's about to go public and has much deeper resources, but that's the next chapter,” Harris said.
Staying focused
Not getting swept away by the shiny stuff can be tough, Harris said, admitting that the team has to slow him down sometimes.
“I'm like the AI czar—I don't like that term—because I'm pushing it so frequently because I'm so excited by it,” he said.
“My team is constantly telling me that we have to slow down, stay more focused and we can't do everything. We need to be mindful. So what are the lowest-hanging fruit? What are the things that we need to do best?”
Harris uses “wow” moments to describe things that go right and “oh shoot” for more challenging elements.
“‘Oh shoots’ are when we miss things that would have been really good to think about or where we made decisions and something didn’t smell right after we made the decisions and went back retrospectively. Those moments are happening,” he said.
Many experts talk about AI in travel and beyond as being very much in its infancy, but Harris feels Cloudbeds is taking it’s first steps.
“This has been about seven months of standing up and falling down. My team that's actually building product for our customers—they are seasoned professionals, they are running because they have built, for the last two years, all the market infrastructure to enable them to run. The organization has not prioritized that up until about a year ago, and now we're starting to do all the things where we're now standing up and we're taking steps forward. I think in the next six months we are going to be running faster than we ever.”
There are other challenges besides the speed with which Harris would like to proceed and the team bringing him back to base.
There are employees with environmental concerns, for example, with AI seen as “one of the most polluting technologies on the planet,” Harris said.
Then there are trust issues with the technology, and there are some people “who use it every single moment of their day,” he said.
“They rely on it too heavily. They're not checking their work all the time.”
This is a further reason why Harris stresses the need for a relationship with the technology where his teams act as the thought leaders above the AI.
His hope is that by bringing people with him, encouraging them to use AI through everyday examples that make work life easier, having ambassadors sharing use cases and training, everyone else will be as excited about the potential as he is.
Tags: AI initiatives, Adam Harris, Cloudbeds, Technology
