Discussing with some clients that attended the recent ICRS 2026 in Dallas, I understood that some may be blurred by “AI as the solver of all my pains” and “the tool approach will cure all my pricing”.
In our industry, there is a prevailing narrative: that the US market represents the "gold standard" for Revenue Management. The perception is often that because markets like the US benefit from technology leadership (which is still right in a way), high levels of rate parity—where shopping at a major OTA often gives you the full picture—pricing is inherently easier, and expertise is commoditized. It is also driven by the fact that the US was the origin of Revenue Management, created by airlines in 1980.
But after years of experience at WeYield, I’ve learned that the reality on the ground is far more nuanced.
Whether you are dealing with net rates in Europe or disparate local channels, the challenge is not just "having the tech." It is about the deep and easy access to a reliable data, a granular mapping of channels, understanding the specific margin strategies for each, set a fleet plan and all strategic decision-making that goes beyond a simple rate update.
We’ve seen firsthand that a tool is only as good as the expert driving it. Technology cannot replace the necessity of training in-house talent or the strategic guidance required to translate data into actionable revenue growth.
The Reality of the "AI Silver Bullet"
There is an ongoing temptation to believe that AI will eventually automate every tactical decision in revenue management. However, we must be realistic about the current boundaries of these tools.
AI is not a magic solution for every tactical problem. It frequently hits a wall at the smallest granularity of elements, or in scenarios where the available dataset is simply not big enough to satisfy the requirements of statistical significance dictated by mathematical law.
Example 1
- You have 500 cars split in 5 car categories distributed among 5 different locations to make my computation easier. Thus we can assume that basically, each location will get 20 cars per category per location per day on which a decision is made.
- But it might be that only 50% of this capacity is booked by the individual segments (individual direct, brokers, OTA’s, Tour operators, eventually partners like hotels, airlines) etc that I call “yieldable demand”. So the other 50% are not impacted by any pricing decisions at all
- So the “super AI” will have 10 cars to optimize per day. If I take two bookings more than last year for a given location, day, car category, does it mean that my demand is stronger by +20%. Yes
- But is it enough to detect a trend that will trigger the use of one model or another ? Clearly not as our dataset is only 10.
Example 2 : do your own maths
- Take your average capacity dedicated to short term
- Apply on it the share that is yieldable to remove the non-daily pricing dependant one (usually corporate segments, replacement, assistance)
- Split it by car category (usually, you may have between 8 to 10 different car categories to optimize)
- Apply total activity share by location (eventually an airport will have a bigger share in your activity than a city station).
And you will get the reality of the data available for your “super AI” to play with. If it is below 200-300 at the smallest, the AI will be not reliable as the dataset will be far too small.
In these instances, relying solely on automated systems without human oversight can lead to suboptimal outcomes and surely noise.
True revenue management excellence is not just about the software—it is about the synthesis of robust technology associated with the human skills and knowledge that understands when the data is sufficient, and when the human strategy needs to take the wheel. The tool is good to compute and execute. Don’t believe it will replace you.
.jpg)
