Welcome to the Era of Perpetual Obsolescence
Technology has gone from moving fast to moving at an absurdly fast pace with ever-increasing velocity. I don't just see this, I live it, every day running my own business while working alongside different organizations that move very differently. Something new launches, improves, disappears, gets replaced or suddenly becomes possible. If you blink long enough you could miss something. That creates a very real problem for airlines trying to build long-term technology strategies around AI.
This isn't a single argument so much as a set of interconnected ones. Obsolescence, architecture, economics and people all show up in this piece, and I think that's the point — airlines and IFEC suppliers don't get to solve these one at a time. They're pulling on each other in real time, and treating them as separate problems is part of how organizations end up behind.
The Speed Problem
When Seth Miller at PaxEx.Aero tested the live Delta Concierge product in August 2026, some of the use cases reported immediately struck me as far less interesting than they would have in early 2025, simply because the technology around them had moved so quickly. When Delta introduced Delta Concierge as a GenAI-powered concept at CES in January 2025, it sounded genuinely forward-looking at the time.
What was innovative in 2025 is in many cases now obsolete. In some cases, what looked innovative three months ago is already becoming commodity technology. That should make airlines think very carefully about how far in advance they announce AI functionality and how tightly they attach themselves to a particular vision.
There is value in showing vision, but vision alone is no longer enough. In AI, the ability to test quickly, decide whether something works, abandon it if it does not, and move on may now matter just as much as early adoption. The uncomfortable part for many is accepting that changing your mind does not necessarily mean the original strategy failed. It means the factors and inputs have changed, and you must be willing to change your assumptions, strategies, process and so on with them.
Why Airlines Can't Move Like Startups
The reality is that airlines are enormous organizations operating complex, safety-critical businesses with legacy systems, regulatory obligations, cybersecurity requirements, privacy constraints and layers of governance. It is not easy for airlines to change direction at the same speed as the technology they are trying to deploy. There will absolutely be efficiencies in organizations with some being substantial. However, getting from an impressive AI demonstration to enterprise-wide operational savings really is a completely different exercise, and the technology can continue evolving faster than the organization can absorb it.
That reality helps explain why products sometimes arrive with less functionality than originally envisioned. PaxEx.Aero found that the initial Delta Concierge deployment falls short of several of the more ambitious use cases discussed when the concept was introduced. Delta Concierge is one example, but it's a useful one, because the gap between what was discussed and what shipped is exactly where these bigger structural questions tend to surface.
It would be easy to conclude that somebody simply did not think the original concept through properly for a use case. We will never know for sure if that is the case, but we must also take into consideration that there are plenty of other plausible explanations. Ambitious AI concepts still have to get through information security, privacy, legal, data access, architecture, integration and operational governance, and some ideas will not survive that process in their original form.
Delta may well be shipping an MVP and iterating as they gather data, which is a legitimate approach. But MVP thinking assumes you have time to iterate your way to relevance. I've watched that assumption fail in real time. Teams ship a lean version, start iterating, and the market moves past the gap before round two even ships. The bar an MVP needs to clear just to feel current keeps rising before the iteration even happens.
The Personalization Test
Personalization is where this gets much more interesting. PaxEx.Aero reported that although Concierge integrates some third-party travel documentation data, questions that were framed around what the logged-in passenger personally needed still produced general responses. That demonstrates the difference between a chatbot and a genuinely intelligent personal travel assistant.
Knowing who somebody is, understanding their itinerary, accessing loyalty information, using third-party travel requirements, applying the right rules and then giving personalized advice requires access to a lot of data, much of it personal. Giving AI that level of access safely is definitely not trivial. Quite frankly, if I were seated at the decision table, even I would be voting to move cautiously knowing full well what could go wrong and the risks. Every additional dataset raises questions around permissions, accuracy, privacy, cybersecurity and liability and these all have to be addressed as the data passes through various gate keepers.
Architecture Is The Real Bottleneck
This is why I increasingly believe architecture, not AI itself, may become one of the biggest determinants of success. Airlines have accumulated technology over decades through different vendors, platforms, business units and generations of architecture. Dropping increasingly powerful AI on top of a higgledy-piggledy technology foundation, something hardly unknown in aviation, does not magically turn that foundation into a modern enterprise.
AI will, in many cases, expose weaknesses faster. A model might be capable of producing an answer in seconds, but if the data sits across disconnected systems, access requires multiple approvals and ownership is fragmented across departments, the capability means very little. The technology may be ready long before the people and the organization are.
Who Pays, Who Owns, and Why IFEC Feels It First
The other issue that gets far less attention is the economics of an AI project and how it can change before the project even launches. A capability that justified significant development investment twelve months ago may now be bundled into another platform, delivered through an inexpensive application programming interface (API) or surpassed by something that did not exist when the original business case was written.
That makes traditional business cases increasingly risky. The value proposition may need to be challenged continuously because the external technology market is moving faster than most corporate budgeting cycles. A project can still be technically viable yet strategically obsolete by the time it reaches the customer.
Then there is the internal organization itself. AI cuts across digital, IT, customer experience, operations, commercial, loyalty, cybersecurity, legal and data. That’s a lot of stakeholders that need to be onboard, aligned and clear on where the project or organization is trying to go.
Who owns the AI capability, who controls access to data, who funds integration and whose approval is required can determine whether an initiative moves quickly or quietly dies inside the organization. Internal service level agreements (SLAs), competing priorities and organizational ownership can become bigger barriers than the technology itself, resulting in many potentially valuable projects losing momentum.
This becomes particularly relevant in inflight entertainment and connectivity (IFEC) and especially seatback inflight entertainment (IFE). Suppliers are increasingly innovating to give airlines more flexibility, faster software updates and greater control over their digital passenger experience. However, giving an airline the technical ability to change quickly achieves very little if the surrounding technology, particularly legacy systems, processes and governance, still moves at the pace of traditional IFE programs.
Seatback IFE has historically operated around long aircraft programs, certification requirements and multi-year product roadmaps. Those realities are not disappearing, but we know the digital layers are changing at a completely different speed and that creates a mismatch.
That mismatch gets sharper the more granular the deployment gets. If an airline effectively gets one shot at a given application for years before the next update cycle, does it make sense to keep investing in incremental releases, or does the smarter move become skipping ahead and building for what's next instead? One signal is which product issues linger instead of getting fixed.
People, Not Just Platforms
This isn't just happening in IFEC. I'm seeing it firsthand across the industry and beyond. I've had this play out with a few projects recently, both in and out of IFEC where the solution under consideration was trending towards obsolescence just as the decision was being made. That's not a criticism of any particular person or organization, but rather, it's a warning about the environment we are all now operating in.
Architecture is half the equation. The other half is who's making the decisions because people still play a critical role, and I remain a steadfast supporter of Human in the Loop (HITL). If I were developing an airline IFEC strategy again today, I would approach it differently. I’d start by hiring team members differently because the skillset and experience needed today has changed from what it once was. The leaders of those teams must also evolve from what we typically see if they are going to have a chance of keeping up. Deep industry and technical knowledge still matters, but so does the ability to constantly question what you thought you knew six months, six weeks or even six days ago.
Executive leadership also needs a different mindset. They will need to be more intellectually flexible, comfortable being challenged, open to new ideas, prepared to test their own assumptions and willing to change direction when the evidence changes.
Being attached to a strategy because six months of work went into creating it could become extraordinarily costly if it is not the right strategy moving forward. The traditional approach of spending months developing a twenty-page business case before finding out whether an idea actually works is increasingly feeling incompatible with the speed of technology evolution. It does not mean abandoning governance, discipline or proper investment decisions. It means recognizing that by the time an organization has finished proving why it should pursue an initiative, the initiative it spent months evaluating may have already changed. So, strategy has to also assume that some of today's answers will be wrong tomorrow.
Back to Delta
None of this means Delta is wrong to build Concierge. They have been clear that this was meant to make travel easier for customers, and there is no reason to doubt that. The harder question is whether building individual tools remains the right mindset when the entire tool-building environment has gone on steroids.
AI is changing what, how quickly, and how cheaply something can be built. It is also changing what customers will consider innovative by the time something finally reaches them. The lesson from Delta Concierge is therefore much bigger than whether the current product is impressive.
Airlines, and the IFEC industry supporting them, may need to rethink how technology is developed, funded, governed and even announced. The winners will not necessarily be the companies making the biggest AI promises today. They'll be the ones willing to scrap last quarter's roadmap when the evidence says it's already stale, staying comfortable changing their minds to keep delivering something current and relevant, however long that window lasts.
Disclaimer: This article reflects the independent perspective of Corinne Streichert and IFECtiv, informed by consulting work across the airline and IFEC industry, broader business consulting engagements, and publicly available information. It is intended for general discussion and should not be interpreted as specific legal, financial, or technical advice. References to prior airline experience are included for context only, and the views shared should be considered in light of each airline's individual circumstances.