Peak booking season does not create governance problems. It exposes them.
When demand picks up, there are more campaigns to manage, more offers in market, more changes to inventory and availability, and more pressure to respond to traveler behavior quickly. Add AI-led personalization, and suddenly far more decisions are moving through the same operating model.
A process that works for a handful of personalized experiences can break down when AI can generate and adapt hundreds of them. Every approval becomes a potential bottleneck. Every unclear decision right becomes a delay. Every rule that lives in someone’s head instead of the workflow becomes harder to enforce consistently.
The problem is not that the organization suddenly needs governance. It is that its governance was never designed for this volume.
The time to fix that is before peak season starts.
Applying the same approval process to everything doesn’t scale
When teams become concerned about AI-generated experiences, the natural response is often to add review. Someone needs to check the copy. Someone else needs to approve the offer. Legal should probably see it. Maybe revenue management should weigh in, too.
Some personalization decisions deserve that level of scrutiny. Most do not.
Consider the range of decisions an AI-enabled experience team might make. Changing supporting copy on a destination page is not the same as changing which package a traveler sees. Personalizing an email based on destination interest is different from introducing urgency messaging tied to availability. An itinerary recommendation based on browsing behavior carries different implications than an offer that could affect margin.
Putting all of those decisions through the same approval process does not create stronger governance. It creates a queue. And as personalization volume grows, that queue grows with it.
The better question is where human judgment actually adds value.
Define what AI can and cannot decide
Before scaling personalization, teams need a shared understanding of how much autonomy AI has across different types of experience changes.
A useful way to organize those decisions is into three categories:
- Autonomous within guardrails: Lower-risk changes AI can make when predefined conditions are met. This might include adapting destination-page copy to reflect a traveler’s known interests or changing supporting content within approved brand and messaging parameters.
- Human review required: Changes where AI can generate or recommend an action, but a person needs to approve it before it goes live. Offers, certain availability or urgency messages, and experiences that could materially affect revenue or brand perception may belong here.
- Human-led: Decisions where the business impact, sensitivity, or ambiguity is high enough that AI should support the work rather than make the decision. Pricing strategy, exceptions to established policies, or decisions involving significant legal or privacy considerations are obvious candidates.
Those boundaries will look different for every organization. The important part is deciding them before teams are forced to make judgment calls under peak-season pressure.
Once the boundaries are clear, governance can reflect the actual risk of the decision instead of treating everything AI touches as equally consequential.
Put the rules where the work happens
This is where governance becomes operational.
A policy document that says what AI should and should not do is useful. A workflow that knows those boundaries is much more useful.
Brand standards, approved claims, privacy requirements, escalation rules, and other operating constraints should inform how agents, prompts, playbooks, and personalization workflows are designed. When work stays within those boundaries, it can keep moving. When it crosses them, the workflow should know where it needs to go next.
That does not mean asking AI to decide whether something is legally compliant. It means translating established policies and business rules into a workflow that can recognize when a task is within its approved scope and when a person needs to step in.
For example, a workflow might adapt destination content using approved source material and established brand guidance without requesting approval for every variation. Ask that same workflow to introduce an unapproved promotional claim or materially change how an offer is presented, and the output can be routed for review.
That distinction is what allows governance to support speed rather than compete with it. Teams stop spending time reviewing low-risk work simply because AI touched it, while the decisions that genuinely require judgment still get the attention they need.
Road Scholar shows what governed automation can look like
Road Scholar’s itinerary review is not a personalization use case, but it shows what happens when established rules become part of the workflow itself.
The organization had a highly structured process for reviewing programs against six compliance policies. Each reviewer spent eight to nine hours completing the review.
Nansen redesigned that process as a workflow powered by Optimizely’s Agent Platform. The same review now takes about seven minutes, a 98.5% reduction.
The efficiency gain is significant, but the way it was achieved matters just as much.
Road Scholar did not remove its policies to make the process faster. Those policies became part of how the workflow operated. AI was not being asked to make an open-ended judgment about whether an itinerary was acceptable. It was applying standards Road Scholar had already established and producing a structured output for the team.
That same operating principle applies to personalization. The clearer the rules around a decision, the easier it becomes to determine what AI can handle independently, what should be routed for review, and what should remain human-led.
Three questions to answer before peak season
Experience teams cannot anticipate every personalization scenario that will arise during peak season. They can make sure routine decisions do not turn into debates once volume starts climbing.
Three questions are especially important.
1. Who has decision authority?
Personalization rarely stays neatly within marketing.
A marketer may want to promote a package because it is generating engagement while revenue management is trying to shift demand toward different inventory. An experience change that looks like a straightforward optimization can quickly affect another team’s goals.
Teams should know who makes the call when personalization touches pricing, inventory, revenue, brand, privacy, or another function’s responsibilities.
If ownership only becomes a question after the workflow reaches one of those intersections, the workflow stops while everyone figures out who is allowed to say yes.
2. Are the rules actually built into the workflow?
Defining what AI can and cannot do is only useful if those boundaries carry through to execution. Teams should know whether approved claims, brand standards, privacy requirements, and escalation rules are reflected in the workflows themselves.
Pressure-test a few common scenarios. Can a low-risk content adaptation move without unnecessary review? Does an offer that could affect margin automatically reach the right person? What happens when a request falls outside the approved scope?
If the workflow still depends on someone remembering the rules and manually deciding where work should go, the governance model will become harder to maintain as volume increases.
3. Can the workflow handle peak volume?
A review process can look perfectly reasonable at normal volume and become completely impractical when the number of personalized experiences increases five- or tenfold.
Pressure-test that process before peak season.
If every additional experience creates another manual approval, how many can the team realistically support? Where will work start to queue? Which reviewers become bottlenecks? What happens when one of those people is unavailable?
Those answers reveal where clearer guardrails, decision rights, or escalation rules need to be built into the workflow before volume arrives.
Good governance should give the team more room to move
Governance should tell teams where they can move quickly, not simply where they need to stop.
During peak season, teams may need to respond to changing demand, promote different destinations as availability shifts, adapt lifecycle messaging, or change what content is emphasized across the booking journey. If every variation waits for the same group of people to approve it, the experience can only respond as quickly as that approval queue moves.
A better operating model establishes decision authority, autonomy boundaries, and escalation paths before volume increases, then builds those rules into the workflows themselves. Routine changes can move within established parameters, while people focus their attention on decisions where context, judgment, or business tradeoffs actually matter.
The result is more capacity without giving up control: less time manually checking work that already falls within established boundaries and more time for the decisions that still need a person behind them.
FAQ: AI governance for peak-season personalization
Why does peak season expose AI governance gaps that were not visible during the rest of the year?
Peak season puts more decisions through the same operating model. Processes that work when personalization volume is low can become bottlenecks as campaigns, offers, content variations, and experience changes increase. Testing the operating model before peak season helps teams identify where ownership, approval, or escalation rules need to be clearer.
Why doesn’t adding more approval steps solve an AI governance problem?
Some AI-driven changes should require human approval, but applying the same review process to every change does not account for differences in risk. Lower-risk changes may be able to operate within established guardrails, while higher-risk decisions still require human judgment. This keeps review focused where it provides the most value.
What does it mean to build governance into an AI personalization workflow?
It means defining the rules, decision rights, guardrails, and escalation paths that determine how the workflow operates. When a change falls within approved boundaries, it may be able to proceed automatically. When it falls outside those boundaries or reaches a defined risk threshold, the workflow routes it to the appropriate person for review.
Which AI personalization decisions should still require human review?
The answer depends on the organization and its risk tolerance, but changes involving pricing, offers, significant revenue implications, sensitive customer data, legal requirements, or material departures from approved brand standards are more likely to require human oversight. Teams should establish those boundaries before scaling personalization rather than deciding them case by case.
What does Road Scholar’s experience show about governed AI workflows?
Road Scholar applied AI to a structured itinerary review governed by six established compliance policies. Each reviewer previously spent eight to nine hours completing the review. Nansen redesigned the process as a workflow powered by Optimizely’s Agent Platform, reducing review time to about seven minutes, a 98.5% reduction. While it is not a personalization use case, it demonstrates how clearly defined rules can make repeatable work easier to automate without removing the standards governing the process.
What should travel experience teams establish before scaling AI personalization for peak season?
Teams should know who has authority when personalization affects other parts of the business, which decisions AI can make within established guardrails, which require human review, and whether the review process can handle peak-season volume. Establishing those rules ahead of time makes it easier to scale personalization without creating unnecessary approval bottlenecks.
Nansen helps travel and hospitality teams define the governance, decision rights, and operating workflows needed to scale AI-led personalization without putting every experience through the same manual gate. Connect with us to assess where your organization is ready to scale and where stronger guardrails are needed before peak season.
