AI Planning:
Designing for Human Deliberation
in the Age of Intelligent Systems

The new starting point for travel

Travel once began inside the airline’s domain. Customers opened an app, compared within a brand’s frame, and made a choice.
That frame has dissolved.

Today’s travelers begin their journeys with AI copilots that propose, refine, and recompose itineraries across airlines, hotels, and ground transport. In doing so, they have become curators of their own reasoning—assembling, labeling, and revisiting options until they form a coherent story about what the trip should be.

This shift from choosing to curating marks a profound behavioral change. Planning has become a form of reasoning carried out through technology, and the airline no longer owns the opening of that reasoning process. By the time travelers arrive in its ecosystem, they are already mid-conversation—with goals, trade-offs, and preferences partly shaped by AI.

The business question is clear:
When deliberation begins elsewhere, how can an airline remain a trusted partner in it?

Why this research mattered

AI-assisted planning is redrawing the boundaries of customer interaction. If travelers plan through external systems, the airline risks losing influence over the formation of preference and the definition of value.

This study explored how those emerging behaviors reshape expectations once travelers enter United’s channels—how AI planning affects trust, coherence, and perceived control—and how the company’s AI and digital strategy should evolve in response.

At its core, the work addressed a pivotal strategic decision:

Should United invest heavily in front-facing conversational AI—building or branding copilots that interact directly with customers?
Or should it focus primarily on back-end intelligence—integrating with external AI platforms while ensuring that, once travelers arrive, the airline delivers unmatched continuity, provenance, and coherence?

Understanding how travelers reasoned with AI before reaching United was essential to deciding where differentiation would matter most: at the interface or in the intelligence beneath it.

In short, the study asked how AI changes not just what customers choose, but how they come to believe a choice makes sense—and where the airline should participate in that reasoning.

Questions we asked

How do travelers’ prior AI interactions influence their expectations of airline interfaces and explanations?
What cues of provenance, reasoning, and pacing preserve a sense of continuity across tools?
And how should digital channels adapt when planning no longer begins—or ends—within them?

Approach

To capture the texture of human and machine reasoning intertwined, the team combined three complementary methods:

• In-depth interviews and journey probes tracing recent AI-assisted planning sessions, including screenshots, prompts, and the handoff points between AI tools and airline sites.

• A mobile diary study documenting how travelers verified, revised, and finalized plans across platforms.

• Survey and vignette testing that varied presentation style—point estimates vs. ranges, declarative vs. conversational tone, provenance vs. none—to measure effects on trust and booking confidence.

Together, these methods revealed not a linear funnel but a distributed deliberation—a reasoning process that unfolded across tools, moments, and modes of attention.

What we learned

1. AI reshaped goal formation

Planning became a co-authored act of aspiration. AI systems didn’t merely help travelers reach their goals—they helped them form them.
Participants described moments when prompts reframed what they were seeking: new trip lengths, alternate airports, different balances between cost and flexibility.
As one put it, “It made me realize what kind of trip I actually wanted.”

These micro-shifts blurred agency lines: AI surfaced what was feasible and, in doing so, quietly reframed what was desirable.

2. Continuity became the new trust

Confidence no longer hinged on price or speed but on shared reasoning.
Travelers trusted systems that carried forward the logic of prior deliberation—remembering assumptions, explanations, and trade-offs. When reasoning broke, trust broke; when it held, the experience felt “guided” rather than “controlled.”

3. Deliberation depended on pacing

The most trusted interfaces slowed down at the right moments.
When travelers could see trade-offs—time, cost, convenience—clearly laid out, they felt more in control even with identical outcomes.
AI that “thought with me” rather than “thought ahead of me” created confidence.

This respect for the deliberative interval—the pause that allows reflection—proved as important as technical accuracy.

Design principles

• Explain, don’t declare. Conversational rationales (“$642 because it includes a flexible change policy and departs earlier”) outperformed static prices.

• Maintain provenance. Timestamping and data-source labeling preserved both continuity and integrity. In a deliberative system, data provenance isn’t just technical—it’s ethical: it lets users retrace the reasoning that led to a recommendation.

• Design for curation, not just recommendation. Support saving, comparing, and revisiting reasoning steps, not merely selecting options.

• Reveal goal evolution. Make visible how the system’s suggestions reshape what the traveler’s goal is becoming.

• Respect the deliberative interval. Confidence grows not from speed but from rhythm—the right tempo of explanation and choice.

Strategic implications

This research reframed United’s AI strategy around deliberation as a service—recognizing that value now lies in how well the brand supports the human work of reasoning before, during, and after booking.

It provided a framework for evaluating where to compete in the AI ecosystem:

1. Front-facing copilots: investing in proprietary or branded AI assistants that mirror travelers’ external experiences and preserve brand presence within the conversation itself.

2. Back-end intelligence: focusing on integration, provenance, and continuity—ensuring the airline’s data, pricing logic, and reasoning remain consistent and intelligible when surfaced through external AI intermediaries.

The findings suggested that trust and differentiation stem less from owning the interface than from owning coherence. Supporting deliberation—ensuring the customer’s reasoning holds together across contexts—may deliver greater long-term value than replicating tools that already dominate the front-end space.

In this light, the research didn’t just reveal customer preferences; it helped clarify where AI investment creates strategic leverage for the business.

Reflection: living with decisions

Every booking is a small act of commitment—an intention projected into the future.
AI now participates in forming those intentions, but good design ensures they remain intelligible, defensible, and revisable.

The true measure of intelligence in this context is deliberative integrity: whether a system can remember, explain, and repair its own reasoning as travelers move from plan to action.

Designing for that integrity ensures planning remains something we do with AI, not something AI does for us—and, in turn, secures the airline’s place as a partner in human reasoning rather than a destination at the end of it.