Gartner’s AI Cost Prediction Is Wrong – But It’s a Symptom of a Broken Industry Model

By Alasdair Miller, Head of Development and R&D at Sabio Group

Gartner’s latest prediction that GenAI resolution costs will exceed $3 by 2030 – outstripping offshore human agents – has contradicted many of their previous CX industry predictions.

But before we collectively panic and shelve our AI ambitions, let’s be clear on an inconvenient truth: Gartner isn’t predicting a technology failure. They’re predicting a partnership failure.

Read between the lines of Gartner’s research and you’ll find the culprits hiding in plain sight: vendors flogging consumption based pricing whilst organisations “consume a lot more than they expect.” Partners promising that AI will “make sense of your mess” when it patently won’t. Technology providers taking their fee regardless of whether their solution actually delivers a single automated resolution.

This isn’t a GenAI problem. Instead, it’s a commercial model problem, and it’s been baked into the CX technology industry since long before large language models (LLMs) entered the chat – something that we’ll be discussing as part of our upcoming Disrupt programme of events later this year.

The traditional playbook is simple: sell the dream, implement the technology, invoice for consumption and then move on. If costs spiral? That’s the client’s problem. If automation rates disappoint? Well, perhaps you need more consultancy days.

Is it any wonder organisations are haemorrhaging money on AI investments that deliver nothing but inflated cloud bills?

The Accountability Gap

Gartner rightly identifies the hidden costs; specialised AI talent, unpredictable usage patterns, infrastructure demands. But here’s what they’re missing: these costs only become catastrophic when your technology partner has no stake in controlling them.

When a vendor’s revenue grows as your AI consumption grows, they have precisely zero incentive to optimise your solution. When their fee arrives regardless of outcomes, why would they care if your automation rate hits 30% or 3%?

The entire model is designed to transfer risk onto the client whilst the partner counts their consumption fees.

A Different Approach

At Sabio, we’ve taken a fundamentally different path. Our recent partnership with a large telecommunications organisation operates on a ‘risk reward’ model where our compensation is directly tied to measurable business outcomes. We earn a fee per completed interaction – with additional incentives only when resolutions stick and don’t generate repeat contacts.

In other words: if our AI doesn’t work, we don’t get paid!

This isn’t altruism; it’s accountability. When your commercial success depends on your client’s operational success, you suddenly become very interested in getting the implementation right, optimising token usage, and ensuring the technology actually resolves customer queries.

The Real Prediction

So here’s my counter prediction to Gartner: AI costs will indeed soar for organisations who continue choosing partners with misaligned incentives. They’ll spiral for those seduced by consumption based pricing that sounds cheap until the monthly invoice arrives. They’ll explode for anyone who believes a technology vendor when they promise AI will magically sort their data chaos.

But for organisations who demand outcome based accountability from their partners? The economics look rather different.

The future of AI in customer service isn’t about choosing between expensive automation and cheap offshore agents. It’s about choosing partners who are willing to put their money where their mouth is.