The Reckoning in Revenue Management

Two things are simultaneously true about what AI is doing to the airline revenue management analyst role, and they sit together in tension. This tension must be reckoned with, regardless of how uncomfortable we may be with the conversation.

The first: AI and automation are reducing the need for certain tasks that currently define the RM analyst job. This is already happening, and it will only gain momentum as the technology matures.

The second: the analysts who remain will be the ones who earn that place because of what they bring that AI cannot replicate; these analysts will have more consequential, better-defined, and arguably more interesting work than the current job description asks of them.

Holding both of these simultaneously, without softening either, is the “let’s be real here” framing the industry’s leaders owe the conversation. What follows is my attempt at it.

The task bundle

Economists have a useful way of thinking about jobs not as fixed roles, but as bundles of tasks. Any job (a nurse, an engineer, an analyst) can be decomposed into the specific things someone does throughout the course of a day. Technology does not eliminate jobs wholesale; it changes the bundle. Some tasks leave the bundle and new tasks arrive. Some tasks get augmented: the human does them better or faster with assistance from the technology. Some tasks get automated entirely and are no longer done by a human.

Anthropic recently published an economic futures model built on this framework. It lays out three scenarios for how AI reshapes the broader economy. Which scenario plays out, they argue, will depend on two things: how capable AI becomes, and how fast it is adopted. Their model ranges from modest impact (likened to the internet) to a fundamental restructuring of the knowledge economy.¹ The task-bundle lens is the most tangible way to describe what is happening today in RM.

The RM analyst’s task bundle

Having been an analyst myself, and having commiserated with my fellow analysts from airlines around the world, I know firsthand the tasks that comprise a typical workday.

Around a third of an analyst’s working day goes to sifting: dashboards open in multiple browser tabs and supplemental data pulled from different systems, all an effort to find the important signals that matter amidst everything else that doesn’t. Another third goes to taking actions: logging into the RMS, making configuration changes, adjusting going class or availability based on what the dashboards surfaced. The rest of the day is a steady current of Teams messages, emails, and meetings: alignment calls, briefings, handoffs, etc. These exist primarily because the systems do not share context with each other, so the humans have to.

That is the task bundle for an experienced analyst. A new analyst must also add a significant amount of time in their first few months on the job to learn complex, niche systems as well as the unique characteristics of the subset of markets for which they will be responsible. A significant portion of this task bundle is not what these analysts were hired to do. It is the infrastructure cost of getting to the actual work that is maximizing revenue for the airline.

telos Signals and the task bundle

telos Signals was built to detect issues and opportunities in the data, assigning a revenue impact, identify the most likely cause, and surface the insight directly to the analyst. No dashboard-sifting required. The problem comes to the analyst rather than the analyst hunting for the problem.

That is the augmentation layer. Today, the analyst reviews the issue, applies judgment, and takes action inside the RMS. Alternately, they may send an email to Network Planning or Marketing to make the handoff of responsibility if there is nothing they can do to resolve the issue. Human in the loop. Human making the call.

What is coming very soon is the automation layer. Once sufficient trust has been established through the human-review stage, Signals will be able to write changes directly to the RMS. When that happens, the RMS user interface (the place where analysts currently spend part of their day making configuration changes) ceases to be where the work happens. The analyst’s relationship to the RMS changes entirely.

One more layer deep is collaboration: a multiplayer workspace where an analyst can begin investigating a problem, add context, and pass it to another team member for their input before a decision is made. The handoff meeting (or ad hoc Teams messages and emails) becomes a structured, traceable transfer of ownership inside the tool.

Today, humans pass the information. What is available is that specialized AI agents do the passing, routing issues to the relevant expertise while the human operates at the level of orchestration. At this point the human is responsible for something far more important than sifting through disparate data or creating ad hoc graphs in excel. They are now responsible for configuring rules according to the airline’s business strategy, defining how the company values revenue impact, designing the agent workflows, and reviewing the outputs to ensure that the LLM is staying on track.

That is a materially different RM job than the one currently being done.

This is uncomfortable

I think our industry deserves directness on this.

Not all of the people doing the current RM analyst job will be a good fit the next one.

The analysts who are most likely to thrive in the orchestration model are the ones who are deeply curious about the commercial problem. They care about understanding why a market is behaving the way it is, strive to learn the underlying tech infrastructure well enough to know when an agent’s output should be questioned, and find the work of defining rules and designing workflows intellectually engaging. These people already exist on every RM team I have encountered. They are often underutilized by the current job structure because so much of their time goes to the tasks that Signals will absorb.

The analysts who are just doing the job because it pays decently well and provides flight benefits, or who are using gut feel or manager-facing KPIs to override a sophisticated RMS without engaging with the data science are not well positioned for what comes next. To be direct about the specific harm of that pattern: overriding a well-calibrated RMS on instinct or under pressure to hit short-term, narrowly-scoped metrics is not neutral behavior. It costs revenue. The systems are not perfect, but they are the product of serious data science, and treating them as an obstacle to be worked around rather than a tool to be understood and interrogated is its own problem. Signals addresses this, in part, by surfacing where the system got it wrong and giving analysts a better-informed basis to decide what to do about it.

The workforce

The Anthropic economic model identifies what it calls job reallocation: in more transformative scenarios, more workers have to change occupations. That transition is not painless. Changing jobs is hard, especially if the change is not by choice. It takes time. The workers who are displaced from knowledge-work tasks do not automatically land in the roles where demand is rising.

The airline industry has an interesting version of this problem. Some of the jobs that have been most systematically underfunded and understaffed in commercial aviation are the ones that require humans, and always will. Gate agents, aircraft cleaners, check-in staff, ground operations, maintenance, catering—these roles have operated on thin budgets precisely because the revenue-generating functions of the business consumed the investment. If AI in commercial operations generates meaningful productivity gains, the question of where those gains go is not purely economic. It is a very human decision and should be approached with the utmost of care.

The Anthropic model notes that in the substantial scenario, where AI does half of all knowledge work by 2030, wages for knowledge workers stagnate while the gains flow disproportionately to capital. That outcome is not inevitable; it is a function of what organizations choose to do with the productivity they recover. Companies that reinvest those gains into appropriate staffing and living wages for the roles that require human presence are making a different decision than companies that treat the efficiency gains purely as margin expansion.

Where do I land? I hope this shift prompts a soul-searched revaluation of what every employee earns. The people cleaning aircraft between turns, throwing bags, and staffing gates at 5am have always been (and will always be) essential to the operation, but they have not always been compensated as though they were.

What the board needs to decide

The Anthropic framework presents three scenarios for how AI reshapes the economy: modest, substantial, and extreme. It notes that the outcome depends on how AI capabilities develop and, critically, how companies and workers choose to adopt them. That second variable is a strategic decision entirely within organizational control.

For an airline commercial leadership team, that decision has a few specific dimensions.

What is your talent strategy for the RM function as automation of routine tasks accelerates? Are you hiring (or promoting) for the orchestration role: curious, analytical, systems-minded? Or are you still hiring for the data sifting and niche RM system mastery that platforms like Signals will absorb?

What happens to the productivity recovered when analysts no longer spend half their day on dashboards? Does it go to more flights managed, more markets covered, and better decisions made with the same headcount? Or does it become a headcount reduction conversation, and if so, what is the plan for the people affected? Dedicated employees deserve a plan.

Lastly, what is your definition of a good outcome? Not just for the commercial operation’s revenue performance, but for the people inside and adjacent to it whose jobs are changing?

The tension is difficult to cleanly resolve. That’s the point. It requires a reckoning, not a roadmap. The organizations that engage with it honestly and early will make better decisions than the ones that let it arrive as a surprise.