Airline revenue management has always been a discipline defined by complexity. Volatile demand, shifting competitive dynamics, narrow decision windows, and legacy infrastructure have kept many carriers locked in a cycle of reactive, manual optimization. And yet, while many of today's airline revenue management systems were built for a different era, the tools available to revenue management teams are now undergoing significant changes — and they are changing fast.

In a recent conversation on the Airline Tech Podcast, hosted by Garth Lund, Charles Pierre, Wiremind's CTO, explored how airlines can move from batch-updating, rules-driven revenue decisions to real-time airline pricing that is powered by powerful and transparent AI. From demand modelling and contextual data to ancillary pricing and agentic workflows, here is what a best-in-class airline revenue management system looks like today, and where it is heading.

Why Airlines Need a New Generation of Revenue Management Systems

The core pain point for most airlines is straightforward: the tools they rely on were built for a different era. Many airline revenue management systems require significant manual intervention, struggle to keep pace with real-time market signals, and treat seat inventory, ancillary revenue, and overbooking strategy as separate problems to be solved by different systems.

At Wiremind, the philosophy is different: CAYZN, an end-to-end AI-powered revenue management solution for transport operators, is designed around two principles: smarter AI models and real-time optimization. Rather than running periodic optimization cycles, the platform continuously monitors new bookings, cancellations, competitor moves, and market signals, making it possible to re-optimize decisions on the fly, without having to wait for someone to trigger the process.

Equally important is the scope of what is being optimized. Seat inventory, ancillary pricing, and overbooking strategy are deeply interconnected problems. But treating them in silos leaves revenue on the table. CAYZN addresses this with its unified platform, giving airlines a coherent view of their revenue performance across all dimensions.

Explore the sections below to see what this shift looks like in practice, from contextual data and causal forecasting to ancillary pricing and agentic workflows.

Contextual Data in Airline Pricing

Contextual data has become central to airline pricing strategy, with competitor fare data now a standard input to inventory and pricing decisions.

The most impactful signal in use today is real-time booking data on the flight itself. The latest booking activity on a departure acts as a self-correction mechanism: when something unusual happens — an event, a competitor capacity change, a sudden demand spike — the booking signal reflects it almost immediately, capturing disruptions that structured data sources may not yet have recorded. Getting that signal right, and acting on it fast, is where best practice sits today.

Look-to-book data — understanding search and browsing behaviour before a conversion — holds significant promise as well, but its value depends on infrastructure. Integrating look-to-book data properly requires tight coordination between the PSS (Passenger Service System), the RMS, and all sales channels. While many airlines are not yet, technically speaking, in a position where this integration is clean enough to make the signal reliable, the potential is nonetheless real.

The next step is event and calendar data. While this has always been a known input, building and maintaining a high-quality event calendar, including public holidays, school breaks, sporting events, at a city level and across dozens of markets has historically required significant manual effort and is never fully current.

Wiremind is actively building an LLM-powered layer to scrape, structure, and continuously enrich event data without manual intervention. Within a few years, this will be a standard component of how the best RMS platforms ingest the world around them.

Ancillary Pricing Optimization and the Path to Total Revenue Management

While ancillary revenue is one of those areas where the industry has made huge commercial progress over the past decade, most airlines are still pricing ancillaries with relatively basic tools at hand that utilize fixed price points and basic segmentation. At Wiremind, however, the same causal inference methodology underpinning CAYZN's AI-driven pricing optimization can be applied to ancillary pricing, and the results are significant.

At the core is an attachment rate estimator that predicts, for any given ancillary product, the probability that a customer will purchase it, and how sensitive that probability is to price changes.

Ancillaries also offer something seat pricing cannot: as there is no legacy fare ladder to contend with, and no class structure inherited from decades of GDS architecture, pricing can be naturally continuous, and personalized. Meaning, a passenger travelling on a fully flexible business ticket has a different willingness to pay for a seat upgrade or extra baggage than a passenger on the cheapest available fare. CAYZN models purchase probability by those customer segments, enabling dynamic ancillary pricing that adapts to each traveler's propensity to buy.

This capability is already live in production. Wiremind's Ancillary Pricing Optimization (APO) enables analysts to manage both fare and ancillary pricing from the same interface and with the same business rule logic — bringing the same rigour to ancillary decisions that best-in-class carriers have long applied to seat pricing.

But APO is only the first step. The more transformative direction is joint optimization: a single model that recognizes a passenger likely to purchase multiple ancillaries may be more valuable at a lower seat price than one generating higher seat revenue with no ancillary spend. This fundamentally reframes the optimization problem — shifting from maximizing seat revenue in isolation to maximizing total trip revenue per passenger.

There is also a third direction that brings the logic full circle: applying the same segmentation and continuous pricing approach back to the seat itself. If you can infer a booking’s customer segment at the moment of purchase, you can move beyond availability-driven pricing and toward more personalized seat pricing, informed by each segment’s willingness to pay. This is a longer-term evolution, but the causal modelling foundations we have built for ancillaries make it a credible next step.

Inside CAYZN's Two-Stage AI Optimization: Demand Modelling and Optimization

The technical foundation of CAYZN's AI pricing engine is a two-stage approach that separates demand modelling from optimization (”Optimizer”), and solves one of the most persistent problems in airline revenue management along the way.

The demand modelling challenge: Understanding how demand responds to price sounds simple. In practice, it is extraordinarily difficult for two reasons. First, data scarcity: for any given flight on any given date, you only observe one price per day to departure, with limited variation over time. Second, and more fundamentally, a classic statistical trap: because airlines already know which flights are popular, they price up high-demand departures and price down lower-demand ones. A naive predictive model trained on this data will conclude that raising prices increases demand, which is obviously wrong, but is exactly what the historical data appears to show.

CAYZN addresses this through a neural network architecture called TSMixer, embedded within a causal inference framework. Rather than modelling demand as a single function of price, the model decomposes the problem into three components:

• M0: the predicted demand accounting for all factors except price, such as seasonality, booking pace, competition, and other exogenous effects.
• P0: the predicted price a skilled analyst would have set in this context.
• theta: the elasticity parameter capturing how demand shifts in response to deviations from that expected price.

Together, these three components produce a full demand matrix grounded in causal inference rather than simple correlation to illustrate the expected demand at any price across a continuous range.

The optimization stage then deploys a greedy heuristic that continuously re-runs as new data arrives, searching through the demand matrix for the pricing path that maximizes revenue over the remaining booking window. The algorithm is also fully steerable: analysts can configure the balance between revenue and traffic objectives, impose price stability constraints, and set hard floors and ceilings — all of which the Optimizer respects as it searches for the best path.

The next evolution takes this further. At Wiremind we are now building an AI agent operating on top of this optimization layer, enabling analysts to express commercial intent in plain language, such as "apply an early booking strategy," "undercut competition where possible", and have the agent select the optimization configuration most aligned with that intent. The analyst retains strategic control; the system handles execution.

Business Rules and AI: Complementary Layers, Not Competing Approaches

The framing of business rules versus AI comes up often, and it’s worth reframing slightly — because in practice, these two models aren’t in opposition at all. They’re complementary layers, and the most sophisticated operators use them together very deliberately.

At the most basic level, business rules can enforce constraints: promotional fares on specific routes, booking class cuts at defined day-to-departure thresholds, alert flags for unusual flight performance. These are guardrails working around the optimization, not substituting it.

More sophisticated use cases would be business rules actively steering the optimization's behaviour based on context. Have a look at below examples from Wiremind's most advanced customers to better illustrate this use case:

• Price stability tuning: The optimization has a parameter controlling how frequently it changes prices. Many airlines prefer stability well in advance of departure but want the system to become more reactive closing-in on D-0. Business rules automatically lower the stability constraint as the day-to-departure count crosses defined thresholds. Hence the optimization becomes progressively more aggressive without manual intervention.
• Dynamic price latitude on outperforming flights: When a flight is tracking ahead of its expected booking curve, rules detect this and automatically widen the pricing range available to the Optimizer on those departures, giving the AI more room to push prices where the context warrants it.
• Mixed objective management: The optimization's balance between traffic and revenue objectives can be shifted by season, route type, or commercial period via business rules, without having to touch individual flights manually.

The result is a system where business rules provide control and confidence, while AI provides analytical intelligence. The value lies in how they are interconnected.

How CAYZN Solves the Black Box Problem: From Model Transparency to Agentic Controlled Optimization

Transparency is a valid concern for revenue management teams evaluating next‑generation systems. At Wiremind, we address it on multiple levels — not just by explaining outputs, but by making the system transparent and controllable in practice:

Transparency for Forecasting Models:

At the modelling layer, customers have full visibility into forecast outputs and error metrics within the CAYZN application, with the ability to export predictions and accuracy data by route, departure, and day to departure. Each model retraining is accompanied by a structured review session with CAYZN’s AI team covering accuracy by origin-destination pair, implied price elasticity, and a direct comparison of how the model would have priced historical flights versus what the analyst actually did. On top of that, a forecast accuracy dashboard is currently being rolled out, built around concrete KPIs (e.g., MAPE targets, centered error distributions, and checks that recommended prices stay within typical analyst ranges) to give teams a shared language for model performance.

When everyone is aligned around the same metrics, the conversation shifts from "do I trust the model" to "here is exactly where the model performs well and where it needs more attention."

Transparency for Optimization via Agentic Steering:

At the optimization layer, the answer is a more structural one. The right long-term response to the black box problem is not just better, explicative dashboards — it is rethinking how humans and systems interact. The agentic steering layer described above achieves exactly this: an agent operates on behalf of the analyst, based on a previous expressed intent. Analysts engage with the agent the way they would engage with a skilled colleague, providing strategic directions rather than parameter-level instructions.

What Airlines Are Really Buying When They Choose Wiremind and CAYZN

The competitive landscape for RM technology is changing. What separates the leading airline revenue management systems from the rest is not the ability to ship code: it is scientific depth, cross-functional product alignment, and the unwavering commitment to keep evolving.

At Wiremind, that means a dedicated research team working on fields that have not yet been explored. It also means product teams with a deep knowledge of and passion for revenue management who can translate that research into real production capabilities. And it is also marked by an approach to the technology landscape that treats the arrival of large language models and agentic AI not as buzzwords but as genuine tools that are already changing what is possible in today’s passenger industry.

Airlines choosing a next-generation RMS are not just buying software. They are entering a long-term partnership with a team that treats revenue management as a never-ending scientific problem, and that has no intention of stopping solving it.

Listen to Charles go deeper on causal inference forecasting models, agentic workflows, and the future of airline revenue management in the full episode here.

Discover how you can take advantage of our end-to-end AI-powered Revenue Management System, CAYZN, today. Reach out to us for a demo here.

‍