Guest complaints can tell you where to look. They don’t necessarily tell you what to fix.
Drew Bridges, principal at JMI Realty, saw this firsthand while evaluating the Hilton Garden Inn Austin Downtown. Analysis of guest feedback showed two recurring problems: the elevators and the air conditioning.
Both sounded like significant problems. Both could have cost a lot of money to fix.
Only one did.
The elevators were a real issue. JMI identified the problem before acquiring the hotel and factored in the cost of fixing it.
The air conditioning was different.
“The more we dug into it, the more we realized, hey, this actually isn’t a problem,” Bridges says. “We came to realize that this was just a matter of not doing preventative maintenance on the HVAC units.”
The equipment wasn’t necessarily the problem. The operating routine around it was.
For hotel operators, that distinction is important. AI is making it possible to analyze far more guest feedback than a team could reasonably read on its own. But the real opportunity isn’t finding more complaints.
It’s using those complaints to understand what’s happening inside the hotel.
Hotels already have the data
As JMI evaluates hotels, Bridges uses AI agents to analyze different sources of information. Guest feedback is one of them.
The opportunity, as he sees it, comes from information hotels have been collecting for years but haven’t always been able to use effectively.
“The verbatims, the text part of those guest service scores, really would always end up going into a black box,” Bridges says. “Nobody would end up looking at them. There were just too many of them.”
That’s a problem most hotel operators will recognize.
A GM can read yesterday’s reviews. A department head can look through comments related to housekeeping or breakfast. Teams can discuss recurring complaints in morning meetings.
But a hotel can accumulate thousands of comments over time. Once the volume gets large enough, it becomes difficult to see anything beyond the most obvious themes.
Traditional text analysis helped somewhat. Bridges remembers tools that could identify frequently mentioned words.
If “smell” suddenly appeared more often, for example, you knew something deserved attention.
But that still left the operator with the important question:
What exactly is happening?
“With AI now you can get into a much larger and deeper level of detail,” Bridges says. “You’ve got a high-volume genius who can plow through a thousand verbatims and really tell you in a very intelligent way what is wrong and what’s right.”
The value isn’t the summary.
It’s knowing where to investigate.
The elevators really were broken
At the Hilton Garden Inn Austin Downtown, elevator complaints kept appearing in guest feedback.
JMI’s analysis suggested they weren’t isolated incidents or noise in the data. There was a physical problem that would need to be addressed.
“We found that out pre-closing,” Bridges says. “So we were able to factor that into our underwriting and carry that cost.”
JMI was looking at the issue as a prospective owner, but the operating lesson applies more broadly.
Repeated guest complaints about a physical part of the hotel can be an early warning system.
A single complaint about an elevator might not tell you much. A pattern over time can reveal problems around reliability, wait times or outages that deserve a closer look.
The same could be true of hot water, mattresses, plumbing, Wi-Fi or room condition.
That doesn’t mean guest feedback replaces inspections, work orders or an engineering team’s knowledge of the building.
It tells you where those other sources of information deserve more attention.
The HVAC wasn’t
Air conditioning also appeared repeatedly in the Hilton Garden Inn’s guest feedback.
This time, the diagnosis was different.
“The more we dug into it, the more we realized, hey, this actually isn’t a problem,” Bridges says.
What JMI found, according to Bridges, was a preventative maintenance problem. The operating routine needed to include regular HVAC maintenance by the hotel’s engineering team.
That’s a very different fix from replacing equipment.
And it illustrates the limitation of taking guest complaints at face value.
A guest knows the room is too hot. They don’t know why.
Maybe the equipment is reaching the end of its useful life.
Maybe filters aren’t being changed frequently enough.
Maybe the unit isn’t being inspected properly between stays.
Maybe engineering response times are too slow.
Maybe the problem is something else entirely.
The guest’s experience is still valuable evidence. But the complaint describes a symptom, not necessarily the cause.
That’s where AI analysis becomes more useful for an operator.
It can help move the team from “guests keep complaining about air conditioning” toward a more specific set of questions about where, when and how those failures are occurring.
The hotel team still has to answer those questions.
Don’t stop at the theme
Most hotel teams already categorize guest feedback in some form.
Cleanliness.
Service.
Breakfast.
Noise.
Maintenance.
Those categories are useful, but they’re only the first layer.
If AI makes it possible to analyze thousands of comments quickly, operators can start asking more useful questions of the same feedback.
Take a rise in cleanliness complaints.
Instead of stopping at cleanliness scores are down, look at what guests are actually describing.
Are complaints concentrated in bathrooms?
Are guests finding rooms that appear not to have been inspected?
Is the issue showing up at arrival?
Are particular room components mentioned repeatedly?
Do complaints cluster around certain days or operating conditions?
Those patterns don’t automatically identify the root cause. They give the operations team better places to investigate.
The same approach works with maintenance.
If guests repeatedly report that rooms are too warm, the next question isn’t necessarily How much will new HVAC units cost?
It might be What does our preventative maintenance history look like for the rooms generating these complaints?
That’s essentially the distinction JMI found in Austin.
One recurring complaint pointed toward the physical asset.
Another pointed toward the way the hotel was being operated.
Connect guest feedback to what is happening inside the hotel
This is where the process becomes more interesting than simply feeding reviews into an AI model.
Guest feedback is one source of operational evidence.
Hotels have others.
Work orders can show how frequently equipment fails.
Preventative maintenance records can show whether required work is actually being completed.
Room inspection data can reveal recurring deficiencies.
Staffing information can help explain whether service problems correspond with particular shifts or periods.
Employee observations provide context that guest surveys won’t.
A useful operating workflow is:
Guest feedback → pattern → possible cause → operational validation → action
AI can accelerate the first few steps, especially when the volume of feedback makes manual analysis impractical.
But it shouldn’t be allowed to skip the validation step.
If AI identifies air conditioning as a recurring problem, go to engineering.
Look at the units.
Look at the work orders.
Look at the preventative maintenance schedule.
Talk to the people doing the work.
Then decide what needs to change.
That might mean replacing equipment.
It might mean changing a process.
It might mean changing staffing, training or accountability.
The point is to get closer to the actual cause before choosing the intervention.
Use AI to answer an operating question
There is a temptation with new technology to start with the tool.
Bridges’ approach suggests a more useful starting point for hotel operators: start with a question you need to answer.
Why are room-condition complaints increasing?
What is driving our breakfast scores down?
Why does the same maintenance issue keep appearing in guest feedback?
Which recurring complaints represent physical problems?
Which ones could be fixed through better operating routines?
Then give AI enough feedback to look for patterns that would be difficult to identify manually.
Bridges is using Claude Code and has built agents and sub-agents to do this work at JMI. A hotel operator doesn’t need to replicate that technical setup to adopt the underlying approach.
The useful part is the sequence.
Start with an operating question.
Use AI to interrogate a larger body of guest feedback.
Identify patterns worth investigating.
Validate those patterns against what is actually happening at the hotel.
Then act.
The opportunity isn’t better summaries
Hotels don’t need AI simply to produce nicer summaries of their guest reviews.
The more interesting opportunity is connecting what guests say to how the hotel operates.
Bridges describes hotels as “living, breathing” operating businesses. That complexity is exactly what makes guest feedback valuable.
Every day, hundreds of operating decisions shape what guests eventually experience: whether preventative maintenance gets completed, whether a room is inspected properly, whether an elevator is reliable, whether an employee has enough time to solve a problem.
Guests see the results of those decisions.
They have been documenting them in surveys and reviews for years.
AI makes it easier to examine those observations together.
The Hilton Garden Inn example shows what can happen when you do.
Guests complained about the elevators.
They complained about the air conditioning.
The complaints sounded similar: something in the building wasn’t working properly.
But one required investment in the physical hotel.
The other required better preventative maintenance.
For an operator, finding that difference is far more useful than another guest satisfaction score.
This article draws on Josiah Mackenzie’s conversation with Drew Bridges of JMI Realty on the Hospitality Daily Podcast. Listen to the full interview.
