Qatreh AI Logoقطره
Back to blog
AI Adoption and Management

Why an AI Budget Does Not Turn Into Profit

Two studies from MIT and McKinsey reached one conclusion: what separates the organisations that profit is redesigning the work, not the technology they bought.

تیم قطرهSeptember 6, 20265 min read
Why an AI Budget Does Not Turn Into Profit

Two recent studies from two independent institutions reached the same conclusion: what separates organisations that get a return from AI from those that do not has almost nothing to do with technology.

The short answer

In McKinsey's Global Survey on AI for 2026 — 1,719 respondents across 97 countries, fielded in May and June — the organisations reporting the largest financial impact were about three times more likely than the rest to have fundamentally redesigned their workflows. Nearly three-quarters of them said they had, against roughly one-quarter of everyone else. The deciding variable is the redesign of work, not the choice of model.

Two numbers that only mean something together

The first comes from the MIT Media Lab's Project NANDA and its report The GenAI Divide: State of AI in Business: in the sample studied, around 95 percent of enterprise generative-AI pilots produced no measurable financial return.

The second is McKinsey's: the share of organisations saying AI has contributed to their EBIT stands at 37 percent, essentially unchanged from a year earlier. Only 6 percent attribute more than five percent of EBIT to their use of AI.

In fairness, the 95 percent figure is contested — its methodology has been criticised, and "measurable return" is a demanding definition. But the headline number is not the valuable part of this research. What matters is its account of what the successful minority did differently.

The real distinction: individual speed versus institutional capability

The pattern both studies describe is this. Most organisations bought a tool that makes one employee faster at one task, and called it transformation.

That has value, but the value is personal and never appears in the accounts. The employee drafts the letter faster; the number of letters, the approval path, the waiting time between stages and the unit's total capacity are untouched.

Institutional capability is a different thing: knowledge that settles into the process and accumulates. When a person leaves, individual speed leaves with them; institutional capability stays.

Why adding a tool to an old process does not work

Suppose a process has seven stages, and stage three takes three days because it is waiting for an approval. Put an intelligent tool on stage two and take it from two hours to ten minutes, and the process still takes about as long as it did. The constraint was somewhere else.

This is what McKinsey means by redesign. The first question is which stage genuinely consumes time and capacity — and whether, given information now available, that stage is still necessary in its current form.

The management point follows: this is not a technical question. Someone must have the authority to change the process itself. A project owned only by the IT department is not permitted to remove an approval stage, and so it necessarily arrives at "adding a tool".

What this means for an organisation in Karaj

Most organisations in Karaj and Alborz province — factories, training centres, service businesses — do not have the budget for repeated trial and error. That constraint is also an advantage: when you cannot make everything intelligent at once, one specific process has to be chosen, and that forced choice is exactly what the research recommends.

Our practical advice is one process, one numeric measure, one short period. If the measure does not move, the project does not grow. How to choose that process is in how to choose your organisation's first AI project, and how to redesign it is in redesign the process, do not add a tool.

Our team at Alborz Science and Technology Park runs this session in person and against the real process, because a constraint cannot be identified from a diagram — it takes sitting beside the person who does the work.

Frequently asked questions

Does this mean AI does not pay off for organisations?

No. It means the return comes from redesigning work rather than from buying tools. In the same survey, the successful organisations were nearly three times more likely to have redesigned their processes; their technology was not markedly different from anyone else's.

What is the smallest organisation that can start?

Size is not the determinant; having a repeatable, measurable process is. A ten-person company entering a hundred orders a day by hand is a better candidate than a two-hundred-person company doing non-repeating work.

Why has the financial impact not moved in the global figures?

Because investment has grown faster than redesign. Buying a tool takes weeks; changing an organisational process takes months. EBIT figures reflect the second, not the first.

How do we tell whether our project is in the 95 percent?

A simple test: if you cannot write, before starting, "which number has to move from x to y by what date", the project has no measure — and will not be assessable at the end either.

Should we wait for better models?

On the evidence of both studies, the model is not the binding constraint. An organisation that redesigns its process today will benefit more from tomorrow's better models; one whose process stays untouched will benefit from none of them.

What does it cost to start?

It depends on scope, and no figure is meaningful before the problem is defined. Our approach is to begin with a prototype in a narrow domain, so real effect is measured before any larger commitment.

Related services from Qatreh