Why seven in ten transformations still fail, and what operators do differently

Every executive has seen the pattern. A transformation is announced with a clear ambition, a programme office, a steering committee and a technology partner. Twelve months later the new system is live, the dashboards are green, and the organisation works almost exactly as it did before. Meetings run the same way. The same people hold the same information. The spreadsheet that was supposed to be retired is still the real source of truth.

This is not a failure of effort. It is a well-documented failure of design, and the evidence has been consistent for a long time.

~70%

of corporate transformations fail, per BCG (May 2026)

75%

of senior IT executives expect operating models to change within 12–18 months (Deloitte, 2026)

89%

of leaders say AI only delivers ROI if it understands how the business runs (Celonis, 2026)

A failure rate that has not moved

In May 2026, BCG's Julia Dhar, Kristy Ellmer and Philip Jameson revisited a number that has circulated in boardrooms for years: roughly 70% of corporate transformations fail. Their central observation is not the number itself but its stability. Society has made large technical advances over the same period, yet organisations have not become meaningfully better at helping groups of people change how they work. BCG frames the cause as a gap between what leaders intend and what employees actually experience.

That framing matters because it moves the problem out of the project plan and into the organisation. A transformation can be perfectly scoped, funded and governed and still fail, if the people who must work differently on Monday morning do not experience a clearer, easier, more rewarding way to work.

The technology is rarely the constraint

Deloitte's 2026 Global Leadership Technology Study makes the same point from the technology side. More than 80% of senior IT executives said they were confident their organisations could deploy and govern AI capabilities at scale. Yet 75% believed their operating models and processes would need to change within the next 12 to 18 months to capture more value. Deloitte's Anjali Shaikh read this as evidence that IT is not overconfident, and that the bottleneck sits in the wider business rather than in the IT function.

Celonis's 2026 Process Optimization Report, based on 1,649 senior business leaders worldwide, points the same way. The leading hurdle to AI success was a lack of expertise (47%), followed by misalignment between departments and the difficulty of getting AI to understand business context such as rules, KPIs and benchmarks. Only 6% named resistance to change as a major hurdle to scaling. In Celonis chief executive Carsten Thoma's words, "For AI to truly work for the enterprise, it needs more than just data—it needs context."

What the surviving transformations have in common

From our experience leading technical operations, service delivery and customer-facing organisations, the transformations that hold have a recognisable shape. None of it is exotic. All of it is difficult because it is unglamorous.

1. They start with how work actually flows

Before choosing a platform, they map how a request, an order or an incident moves from the first touch to the last: who handles it, where it waits, which decisions need a person and which do not. The map is built with the people who do the work, not for them. Most organisations discover that the real process is different from the documented one, and that the difference is where the cost sits.

2. Every process has a named owner with authority

A process without an owner is a habit. The owner is accountable for how it performs, can change it, and reports on it in a regular forum. Where ownership is shared across departments, performance is shared too, which in practice means nobody holds it. Celonis's finding on misalignment between departments is this problem seen from the data side.

3. They measure a small number of things, consistently

Good operators resist the urge to instrument everything. A handful of measures that reflect the customer's experience and the cost of delivering it, defined once and reported the same way every period, will change behaviour more than a hundred dashboards. Consistency matters more than sophistication, because trends are only visible if the definition does not move.

4. They build an operating rhythm, not a launch

Launch is an event. Rhythm is what is left afterwards: a weekly operational review, a monthly performance conversation, a quarterly look at where the model needs to change. In large service organisations this cadence, including structured executive business reviews, is what turns a one-time improvement into a permanent capability. Without it, performance decays to its previous level within a few quarters.

5. They sequence technology after the process is clear

Technology is the multiplier. It amplifies a clear process and it amplifies a confused one. Organisations that automate before they simplify tend to digitise their complexity and then pay to maintain it.

A practical test for your own programme

If you are sponsoring or running a transformation, five questions will tell you more than a status report:

• Can the person doing the work describe, in one sentence, what is different for them since the programme started?
• Does every process in scope have a named owner who can change it?
• Are there fewer than ten measures, defined once, reviewed on a fixed schedule?
• Is there a forum where operating problems are surfaced and decided, not just reported?
• If the programme team left tomorrow, would performance hold for two quarters?

If the answers are weak, the remedy is rarely more technology or more reporting. It is usually a slower, more disciplined return to process, ownership and rhythm.

Where this leaves emerging and growth markets

For growing organisations in Rwanda and across Africa, the lesson is encouraging. Many do not carry decades of legacy systems and entrenched departmental habits. They can design the operating model deliberately, with clear ownership and a small set of measures from the start, and then choose technology that fits it. The risk is the opposite one: buying a platform to solve a process problem nobody has yet defined.

That is how we approach engagements at NURBIX. We begin by understanding how the organisation actually works, because the evidence says that is where the outcome is decided.