In a recent presentation, Ben Dixon, CTO of Sona, tackled one of the most talked-about but least understood topics in tech today: Agentic AI. His talk cut through the noise to provide not just a technical explanation, but a deeply human one, framing AI not as a threat to Hospitality, but as a tool to strengthen it.
Ben's definition of AI is simple: it’s the tech that feels like magic, until it becomes commonplace. Where traditional machine learning presents data for us to interpret, generative or "agentic" AI takes it a step further. It can take objectives, navigate data, suggest actions, and even monitor results. In Ben’s words, it’s like a “really, really good chief of staff” working quietly behind the scenes.
Until recently, Hospitality businesses relied heavily on machine learning to do things like forecast demand or set dynamic pricing. These systems could take historical data and make useful predictions. But their effectiveness had limits.
Machine learning is simplistically using complex mathematics to take a bunch of data on one side and make predictions on the other.
Ben explained that while machine learning excels within clearly defined parameters, it struggles to adapt when those parameters shift—something many businesses painfully experienced during the COVID-19 pandemic. Models trained on years of stable patterns broke down almost overnight.
That’s where agentic AI comes in. Unlike traditional systems that require human interpretation, agentic AI systems are designed to explore, adapt, and respond to real-world complexity on their own.
You can create these agents which have access to a large amount of information and make a plan for how to reach an objective.
This means businesses aren’t just getting better predictions. They’re getting intelligent systems that can decide what to do next, even when things change. For Hospitality, this means helping managers and floor colleagues make better decisions, faster - without being buried in dashboards or spreadsheets.
To help the audience picture this in action, Ben shared a practical scenario. Say an area manager notices one of their restaurants is underperforming compared to a similar location. Instead of manually digging into rosters and weather reports, they could ask an AI agent to investigate.
It will have a look at historical rosters, weather, item mixes, and might identify that peak shifts in the high-performing store have managers and more experienced team members working - while the underperforming one doesn’t.
But the power of agentic AI goes further. Once it uncovers the pattern, the agent doesn’t just report the insight. It can suggest solutions, offer to communicate them to the relevant GMs, and continue monitoring the issue long term. The system doesn’t forget, and it doesn’t sleep.
What started with a reactive problem has now been encoded into the behaviour of that organisation. This shift - from one-off fixes to long-term organisational intelligence - is where agentic AI really shines.
Understandably, this kind of technology can raise concerns. If AI agents are taking on analytical and operational tasks, what happens to the people who used to do them?
Ben addressed that head-on, pointing out that the real opportunity lies in freeing people up to focus on the work that only humans can do, especially in an industry so dependent on personal connection.
80% of what agents are going to do are not activities that people are currently doing. They’re things you wish people were doing, but they don’t have time to. A GM probably adds the most value when they are out in front building capability in their team member, not sitting at a computer.
For example, GMs don’t enter the Hospitality world hoping to spend their days managing spreadsheets. Their value is in the relationships they build, the team culture they shape, and the experiences they help create.
In this light, agentic AI isn’t about reduction—it’s about amplification. It’s about letting people do more of the work that matters.
Despite the hype on social media, Ben was quick to dispel the idea that everyone else is already ahead.
There are almost no genuine agent deployments out there… it was only March or April this year that the models got good enough to handle the messiness of real operations.
In other words, we’re still in the early innings. The next year will be pivotal, and there’s still plenty of time for businesses to prepare, experiment, and implement.
And when those first deployments go live, Ben expects real, measurable impact—starting with operational performance, then flowing into employee retention and experience delivery.
For companies wondering how to take the first step, Ben had simple advice: focus on your data and your people.
Agents can only be as good as the data that goes into them. Ask: where could someone in your organisation be 5x more effective if you gave them leverage?
Ensure you know where your data lives, how clean it is, and that you have access to it. But just as importantly, think about your teams. Where could a little bit of leverage make a big difference?
At Sona, that focus was engineering. Deploying agentic coding tools led to a 2x, 3x, even 4x increase in productivity—results that are now inspiring the rest of the organisation to follow suit. When people see what’s possible, excitement spreads naturally.
The numbers keep going up… and all the other teams have now moved from trepidation to trying to pull it in.
Ben’s closing message hit a note of optimism that felt true to the industry he’s spent the last decade serving.
Technology, especially agentic AI, should be in service of that mission—not a distraction from it. By giving people the tools to spend less time on admin and more time on human connection, we’re not just making businesses better—we’re making Hospitality better.
No one really got into Hospitality to use Hospitality tech products. They got into Hospitality to provide great experiences. I think what Agentic AI has the capability to do is to unlock that, and get people back out there focused on providing experiences.