Visible value: AI has to earn its budget like everything else

By Oana Savu, Chief Growth Officer, Ink Innovation

Airlines make thousands of decisions every day. Some are repetitive, while others depend on experience, context, and judgment. AI is starting to change how many of those decisions are supported and, in some cases, how much human intervention they require.

That creates a much bigger question than where AI can be used.

Where should airlines actually invest?

The challenge is not finding ideas. The harder part is deciding which ones can make a meaningful difference to the economics of the business, the operation, or the customer, and which ones will remain interesting experiments.

As for me, this distinction is becoming much more important as airlines are under pressure to manage costs while still improving reliability, customer experience, and revenue. AI, therefore, must compete with many other investment priorities, and it cannot be treated as a separate innovation agenda in which every promising idea deserves a pilot.

At some point, AI has to compete for capital like everything else, and perhaps that is a sign of maturity.

Start with the problem, not the technology

We still tend to start many AI conversations with the technology.

What can we predict? What can we automate? Where can we use an agent?

I would start somewhere else. Is the problem important enough?

A process performed thousands of times a day, involving repetitive work or avoidable customer effort, may be a much better place to invest than a sophisticated use case solving an interesting but relatively small problem.

This sounds simple, but it changes the discussion.

The question becomes less about whether AI can do something and more about whether changing that particular process or decision creates enough value to matter.

Another trap is that making an existing process faster does not necessarily make it better.

If AI simply adds a recommendation on top of the same process, the same approvals and the same way of working, the result may be technically impressive without changing very much.

The same is true from the traveller’s perspective.

A faster internal decision does not create a better experience if the passenger still has to repeat information, switch channels, or figure out what to do next.

Sometimes the real opportunity is not to automate the process we have today, but to rethink why it works that way in the first place.

This is one of the harder parts of AI adoption because it moves the conversation beyond technology.

It becomes an operating-model and customer-experience question.

The role of people will change too

Many airline roles still involve monitoring information, checking what has changed, reconciling different inputs and deciding where attention is needed.

AI will increasingly take on some of that work.

But I do not think the interesting question is whether AI replaces people.

It is what we want people to spend their time doing instead.

If technology can handle more routine tasks, people can focus more on exceptions, judgment, trade-offs and the moments when human interaction genuinely matters.

That matters for travellers too. The best use of AI may not always be the most visible one. Sometimes the real benefit is that an employee has the right context at the right time, can resolve an issue faster and does not need to send the passenger elsewhere.

If AI makes the operation more efficient but makes the customer experience more fragmented or less human when it matters, we have probably optimised the wrong thing.

Where are we comfortable allowing technology to act? Where do we still want human approval? And are we keeping a person in the loop because they genuinely add value, or simply because that is how the process works today?

There will not be one answer.

But if AI changes the technology and not the work around it, I suspect much of the potential value will remain on the table.

Scaling is where the real work starts

A successful pilot proves something can work, but it doesn’t automatically mean the use case should scale.

Before scaling, we should look beyond model performance. Are people using it? Is behaviour changing? Is the outcome actually improving?

A highly accurate model can still create limited business value. Equally, a relatively simple capability can have a significant impact if it removes repetitive work, improves a high-volume process or reduces traveller effort at scale.

Part of AI maturity is knowing which use cases create enough value to justify scaling and which do not.

That is simply part of making more disciplined choices about where to invest and scale.

And ultimately, the value has to be visible

This matters even more in the current environment, where AI has to compete with many other priorities for investment.

That doesn’t mean every use case needs a simple cost-saving number. Value can show up in different ways: lower costs, protected revenue, better use of resources, improved reliability, less effort for travellers, or simply giving employees more time to focus on work that matters.

But we should be clear from the start about what we are trying to improve.

Otherwise, it is easy to end up measuring the AI rather than the outcome — how many models were deployed, how many recommendations were generated, how accurate they were, how often they were used.

Those measures can be useful, but they do not tell us whether the airline, the employee or the traveller is actually better off.

For me, the more important question is:

What is different now that we have introduced this capability?

Did the airline make a better decision? Did the employee spend less time on repetitive work? Did the traveller have to do less?

Ideally, we should be clear from the start about what success looks like and how we will know whether the use case is creating value.

The next phase is about harder choices

AI will keep moving quickly. The technology will improve, new use cases will emerge, and what looks difficult today may soon become much easier.

That makes experimentation important, but it also makes judgment more important.

Airlines cannot pursue every opportunity, nor should they try.

The more mature organisations may not be the ones with the longest list of AI projects, but the ones making the clearest choices about where AI can genuinely make a difference.

Which problems are worth solving differently? Where does human judgment still matter? Which use cases create enough value for the traveller and the business to justify scaling?

For me, that is where the next phase of AI in aviation becomes more interesting.

Less about proving that AI can do more, and more about deciding where it should make a real difference and being disciplined enough to prove that it did.

To learn more, click here.