Think back to the last time you called to change a travel-related reservation. What sounds like it should be such a simple request can involve verifying an itinerary, finding alternative plans, navigating seat or cabin availability, updating a booking, processing a payment or applying a credit, and confirming the change.
From the customer’s perspective, that is one interaction, perhaps a frustrating or time-consuming one.
For a CX leader, it is also a workflow: a series of steps an agent has to navigate, systems they may need to access, decisions they have to make, and points where the conversation can follow the expected path…or veer somewhere else entirely.
Now multiply that workflow across all its possible variations and across the thousands of customers reaching out to a major travel or hospitality company every day. What looks like a relatively straightforward request at the individual level becomes a significant operational challenge at scale.
For customer experience leaders evaluating where AI can take on more of this workload, the temptation is naturally to look first at sheer volume: find the interactions consuming the most agent capacity, automate them, and keep it moving.
Our latest analysis shows that may be far too simple an approach, and could instead lead you to overlook some of the strongest automation opportunities.
A fundamental question for enterprise automation strategies emerges:
What actually makes a conversation a good candidate for automation?
In this latest installment of Cresta IQ, we analyzed more than one million customer conversations across anonymized travel and hospitality deployments spanning airlines, cruise lines, and hotels.
We wanted to understand where automation could be applied, as well as where the underlying conversation data suggests it has the strongest foundation for success.
Methodology
Our analysis looked across conversation reasons and, where available, common customer questions, along with signals including conversation volume, AHT, sentiment, and resolution.
We paired those conversation-level signals with AI-generated conversation flow maps that show how customer interactions typically progress—from the core stages and expected ‘happy path’ to branching points, deviations, and workarounds.
These maps allowed us to examine not just what customers contact companies about, but how those interactions actually unfold: the steps agents take, where conversations diverge from the expected path, and the systems and tools agents rely on along the way.
From there, we assessed each flow’s Automation Readiness. Rather than looking at conversation volume or intent alone, Automation Readiness considers the structure and complexity of the underlying workflow to identify flows that may be stronger candidates for automation.
But readiness alone does not tell us where automation could have the greatest business impact. A workflow can be highly suitable for automation without representing a meaningful share of contact center demand, and a high-volume interaction can offer significant potential impact while being much harder to automate reliably. To identify the strongest opportunities, we therefore looked at automation readiness alongside the scale and potential efficiency gains associated with each conversation theme.
Looking at these dimensions together allowed us to distinguish between a workflow that is simply common and one where readiness and business impact converge.
What the Strongest Automation Opportunities Have in Common
Across the analysis, multiple signals combine to help explain why certain themes rise above others with regard to automation potential.
Conversation volume establishes the potential scale of the opportunity. ‘Booking & Reservation Changes’ demonstrates this most clearly: even automating a portion of the category could shift a meaningful amount of demand away from live agents, freeing up time and resources for more nuanced and strategic work.
Resolution performance tells us whether human agents can already handle the workflow reliably. ‘Seat & Cabin Management’, ‘Credits, Refunds & Payments’, and ‘Service Requests & Support’ all show resolution rates in the mid-to-high 70% range.
Average handle time helps establish the potential value of automating each interaction. In this installment’s analyzed data, several of the highest-priority individual use cases combine handle times well above the overall median with high readiness.
Flow structure helps explain whether that performance is repeatable. Across the analyzed travel and hospitality flows, workflows generally contained five to seven phases and 14-16 steps. Optional branches and exceptions existed, but the core structures remained relatively compact.
Deviation rate captures how often conversations leave those expected paths. Roughly 75% of analyzed flows landed at approximately 19% deviation or below, placing them in the high-readiness range.
Tool complexity was similarly contained. The analyzed flows generally required one to three distinct tools, limiting one potential source of automation complexity.
Taken together, those characteristics resulted in three of four analyzed flows receiving a High Automation Readiness designation, with the remaining flow receiving a Medium designation.
The important point is that none of these measures tells the whole story on its own. The strongest opportunities emerge when those signals begin to reinforce one another.
Here are our takeaways across industries:
Where Scale and Automation Readiness Converge
The four largest cross-industry themes account for roughly 80% of all classifiable travel and hospitality conversations in our analysis. But looking at volume alone obscures an important finding: these themes represent different types of automation opportunity.
Some offer significant conversation volume that could be deflected away from the contact center, while others stand out because agents already resolve them highly consistently. And some offer an unusually large amount of agent time back for every interaction that can be automated.
‘Booking & Reservation Changes’ is the clearest example of an opportunity driven by scale. It represents approximately 45% of classifiable volume and was the top contact theme across every travel and hospitality company we analyzed. The concentration varies considerably by business, from the teens at the low end to more than half of conversations at the high end, but the theme is universal.
More importantly, that volume is not attached to an obviously unstable workflow. These interactions average more than 12 minutes of handle time, while the underlying flows tend to remain relatively compact at roughly five to seven phases and 14 to 16 steps. Resolution rates exceed 73% in the underlying analysis, and the relevant flows generally show high automation readiness. This is not only the largest source of demand, but one where agents are also already following relatively consistent processes to reach successful outcomes. Automating even a portion of it could therefore address a meaningful share of total agent workload.
The ‘Credits, Refunds & Payments’ category presents a different kind of opportunity. At approximately 15% of classifiable volume, it is substantially smaller than the booking category, but again appears across every company in the analysis. These conversations include interactions around flight credits, cruise deposits, refunds, and payments, exactly the kinds of issues where customer frustration might suggest a difficult automation environment.
The data complicates that assumption. We see evidence of negative sentiment in portions of these conversations and elevated handle time in some refund-related clusters, yet the underlying workflows remain comparatively structured. Resolution is strong, and deviation rates remain in a range associated with higher readiness. In other words, a frustrating customer experience may not always be indicative of an unpredictable operational process; customers may be frustrated precisely because they are navigating a procedural process that takes time, not because the path to resolution is unknowable.
‘Service Requests & Support’ stands out for reliability. These interactions account for approximately 11% of classifiable volume and include use cases such as dining reservations, accessibility assistance, and tour inquiries. Their individual handle times tend to be more modest, often approximately six to thirteen minutes, but they pair that efficiency with some of the strongest resolution performance in the analysis, at 75% or higher, and relatively low tool complexity.
Agents already handle these interactions consistently, the work is highly procedural, and the technical footprint is relatively contained. They therefore represent the kind of repeatable service work where the path from human-assisted to automated resolution may be comparatively straightforward.
Then there is ‘Seat and Cabin Management’, which illustrates why the largest opportunity and the highest-scoring opportunity may not always overlap.
‘Seat and Cabin Management’ represents only a small share of overall classifiable volume, yet it emerged as the highest-scoring theme in our overall automation opportunity assessment. Its strength comes from the combination of signals underneath that smaller volume: high resolution, low deviation, limited tool complexity, high automation readiness, and substantially longer interactions. In the underlying data, seat- and flight-management interactions can extend from the high teens to more than 30 minutes.
Where ‘Booking & Reservation Changes’ offers the greatest opportunity through sheer scale, ‘Seat and Cabin Management’ offers the greatest potential time savings each time automation successfully handles an interaction. Neither is definitively the “better” use case; they solve for different dimensions of the business case.
Automation opportunities are not one-dimensional. A high-volume workflow can create enormous aggregate value even if another theme scores better on structural readiness. A lower-volume workflow can still deserve priority when each interaction consumes significant agent time and follows an unusually stable path. And a relatively short interaction can be attractive when high resolution and low complexity make it easier to automate reliably.
The strongest automation roadmap isn't necessarily a ranked list from most conversations to fewest. It is a portfolio of opportunities that balances scale, potential time savings, existing resolution performance, and the predictability of the underlying work.
A Deviation Is Not Always Evidence of a Broken Workflow
The readiness analysis surfaced another finding that deserves a closer look.
At first glance, deviation rate appears straightforward: common sense may suggest that the more often a conversation leaves the happy path, the harder it should be to automate. To be fair, broadly, that logic holds.
But when we examined why conversations deviated, there was more to the story.
In one high-readiness travel flow with approximately 12% deviation, more than half of the deviations were due to off-topic or casual conversation, while nearly a third were call transfers. In another flow with approximately 15% deviation, nearly three-quarters of deviations were casual or personal conversation.
Often, these casual, off-topic conversations occur while agents are waiting for an internal system or process to respond. The deviation may point to something beyond the automation readiness of the conversation itself: an opportunity to improve the workflows and systems agents are interacting with behind the scenes.
Another analyzed travel flow had a deviation rate around 19%. More than half of those deviations were transfers outside the core flow to specialized departments, while roughly one-fifth reflected communication or technical issues. Again, the deviation is real, but its cause reveals more nuance.
The medium-readiness hospitality flow told a different story.
Its deviation approached 30%, with nearly half associated with voicemail or automated interactions and roughly a quarter with special-handling requests. Here, the composition of the contact mix creates a substantially noisier environment.
Deviation rate tells you where to investigate; deviation type helps tell you what the number actually means.
A 15% deviation rate dominated by casual conversation is not equivalent to a 15% deviation rate caused by procedural exceptions. The former may leave the underlying workflow relatively intact. The latter may expose genuine automation complexity.
That is precisely where human analysis remains essential. A readiness score can identify the flows worth investigating, but understanding why conversations deviate is what turns a metric into an actionable recommendation.
Resolution and Handle Time Reveal Another Layer of Readiness
The relationship between AHT and resolution provides another useful lens.
Across the analyzed hospitality interactions, the densest concentration of conversations sits in a relatively favorable part of the distribution: shorter-to-moderate handle times paired with comparatively high resolution rates. As handle time increases, resolution generally trends downward.
Now, that doesn’t mean that long conversations are inherently poor automation candidates. The theme-level results demonstrate the opposite: ‘Seat & Cabin Management’ combines the longest AHT among the major themes with one of the strongest overall automation assessments.
Instead, the relationship helps distinguish valuable duration from problematic duration.
If an interaction takes 20-plus minutes but still resolves reliably, follows a repeatable path, and uses a limited number of tools, that handle time represents potential capacity that automation could free. If a similarly long interaction has low resolution and repeatedly leaves its expected path, the duration may instead be evidence of complexity.
Consider ‘Disruption & Complaint Handling’. It represents approximately 9% of classifiable volume and roughly 12 minutes of AHT, enough on both measures to attract attention. Its deviation rate, at approximately 19%, is also not dramatically different from several higher-ranked categories.
But resolution falls to roughly 60%, substantially below themes such as ‘Service Requests & Support’ or ‘Credits, Refunds & Payments’.
That single difference materially changes the picture. The interaction may follow a recognizable path, but agents are considerably less likely to bring it to resolution. Automating it therefore presents a different challenge than automating a similarly structured workflow that humans already resolve nearly eight times out of ten.
The broader lesson is that automation readiness cannot be inferred from efficiency metrics in isolation. Handle time, resolution, deviation, volume, and workflow structure tell different parts of the story.
Travel and Hospitality Is Automation-Ready
The overwhelming majority of analyzed flows for which readiness could be assessed landed in the high-readiness range. The workflows themselves were relatively compact, generally required only a handful of tools, and showed lower deviation.
At the same time, readiness is not uniform; the same broad conversation category can account for dramatically different shares of demand across different travel businesses. And even a relatively low deviation rate can mean very different things depending on what is causing the deviation.
That is why a data-backed automation strategy needs to move through several layers of analysis.
Rethinking Automation in Travel & Hospitality
For travel and hospitality leaders, perhaps the most important finding from this Cresta IQ analysis is also the simplest: frequency and readiness are not the same thing.
‘Booking & Reservation Changes’ represents the largest pool of opportunity in the dataset, and its combination of scale and relatively strong readiness makes it an obvious part of the automation conversation.
But the data also surfaced smaller themes with stronger structural signals, including workflows that combine higher resolution, lower deviation, and substantially more agent time per interaction.
That creates a more useful, nuanced way to think about automation strategy: “Where does customer demand intersect with a workflow that is stable, repeatable, and easily resolved enough to automate?”
And in an industry where a single customer experience can touch everything from reservations, payments, loyalty, and service requests, to accommodations and disruptions, understanding that distinction may be the difference between deploying more automation, and deploying automation where it can actually make the greatest impact.
To learn more, download our practical playbook for airlines, hotels, cruise lines, timeshares, and other guest experience brands to help you drive revenue, improve loyalty, and scale customer experience.
Frequently asked questions
