As AI-native firms build around automation from day one, established companies may be forced to revisit the much-criticised idea of business process reengineering.
Artificial intelligence is moving from experimentation into the operating core of many organisations. It is often framed as a productivity tool, a source of innovation and a way to imagine new business models.
The deeper question is whether AI can deliver real value while businesses keep the same processes, hierarchies and assumptions that shaped the pre-AI enterprise.
Business process reengineering, or BPR, was one of the defining management ideas of the 1990s. Michael Hammer and James Champy argued that companies should not merely automate existing work. They should rethink it from first principles. The phrase often associated with Hammer was “Don’t automate, obliterate”, a deliberately provocative way of saying that technology should not be used to make broken processes run faster. It should be used to redesign the process itself.
For many executives, BPR later became a warning rather than an inspiration. It was associated with disruption, failed transformation programmes, job cuts and consulting-led restructuring that sometimes promised more than it delivered. The idea was not always wrong. The practice was often brutal, oversimplified or poorly implemented.
History matters because AI is now reviving a familiar temptation. And this time, the tools are more capable, more accessible, more adaptive and more deeply embedded in everyday work.
But the difference is that AI-native firms do not begin with legacy workflows. A startup built around AI can design its operating model differently from day one. Customer support may be designed around automation before a large contact centre ever exists. Software development may be organised around small teams using AI coding assistants from the outset. Sales, finance, compliance, onboarding and analytics may all be designed around data flows, APIs and intelligent systems rather than departmental handovers.
This does not make startups invincible. Many will fail. Some will mistake automation for strategy. Others will discover that trust, regulation, distribution and capital still matter. But AI-native firms may enjoy one important advantage. They do not have to persuade an old organisation to stop defending old processes.
Why incumbents cannot simply add AI
Large firms face a harder problem because they already have systems, roles, reporting lines, governance forums, risk controls, performance metrics and budgets. These structures were built over years to manage scale, accountability and control. In banks, insurers, retailers, logistics firms, telecoms operators and public institutions, processes often exist because someone once had to reduce risk, satisfy a regulator, manage fraud or keep a complex service running.
AI transformation is slower than AI adoption.
An employee using a chatbot is adoption. A department deploying an AI assistant is adoption. A company redesigning how decisions, accountability, workflow and customer service operate is transformation. Those are different things.
This is the point at which BPR becomes relevant again.
The old BPR promise was that technology could enable radical redesign. In the 1990s, that technology was often enterprise software, databases and networked computing. Today it is generative AI, machine learning, process mining, robotic process automation, APIs and increasingly agentic systems. The tools are different. The question is familiar.
Should companies automate the process they have, or redesign the work they need?
Where AI redesign meets the real organisation
This is where the AI transformation story becomes less glamorous.
The first test is data. AI depends on usable, trusted and accessible information. Yet many large organisations still operate with fragmented systems, duplicated records and inconsistent definitions of the customer, product or asset. AI can make messy information look neat, but it cannot make unreliable data true. A model cannot compensate for years of poor data discipline.
The second test is architecture. Intelligent automation requires cloud capacity, integration, cybersecurity and workflow design. In South Africa, this also meets familiar constraints such as energy reliability, skills availability, legacy technology estates and uneven digital maturity. A clever AI tool cannot fix a supply chain, contact centre or claims process if the systems around it remain fragile.
The third test is people. People do not change simply because software improves. Teams protect familiar routines. Managers defend reporting lines. Professionals worry about judgement, accountability and status. Workers worry about job security. Customers may still want a human being when the stakes are high.
Intelligent automation also requires new human competencies. People need to question AI outputs, redesign workflows, manage exceptions and exercise judgement when machines are involved. AI transformation is therefore not only an IT programme. It is a social negotiation inside the firm.
Ironically, many large companies may use AI to preserve the very complexity that AI should help them question. They will deploy copilots into bloated workflows. They will summarise unnecessary meetings. They will automate approvals that should perhaps not exist. They will keep the same roles, incentives and decision rights, then wonder why smarter tools have not produced a smarter organisation.
That may produce savings. It may even improve service in places. But it will not necessarily create an AI-native enterprise.
Human cost and radical redesign
The uncomfortable possibility is that Hammer and Champy were, in one important respect, ahead of their time. Their language was, in retrospect, too forceful for today’s more careful management culture. The BPR era often underestimated the human cost of radical redesign. Yet their central warning still carries weight. Automating a bad process can deepen the problem.
For established firms, the challenge is not to revive BPR as a corporate hammer. That would repeat old mistakes. The challenge is to recover the useful part of the idea without the destructive theatre that damaged its reputation.
This means asking harder questions before buying another AI platform such as:
- Why does this process exist?
- Which decisions genuinely need human judgement?
- Which controls reduce real risk, and which ones merely preserve habit?
- Where does work pass between teams because of organisational history rather than customer need?
- Which tasks should be automated, which should be redesigned and which should remain human because trust matters?
AI-native competitors may increasingly force these questions onto the agenda in some markets. Their advantage will not only come from better models. It may come from simpler processes, smaller teams, tighter feedback loops and fewer inherited beliefs about how work should be done.
Large firms are not doomed. They have customers, brands, capital, regulatory experience and operational knowledge that startups often lack. In South Africa, established firms also operate in markets where trust, resilience and local knowledge matter deeply. These are not small advantages.
But they may not be enough if AI is treated as another digital layer.
The next phase of AI transformation may therefore look less like software deployment and more like organisational redesign. Not reckless obliteration. Not nostalgic BPR theatre. Rather, a disciplined rethinking of how work should happen when intelligent automation is part of the operating fabric.
Technology expands what organisations can do. It does not automatically teach them what to stop doing. For CEOs, COOs and CIOs, this may be the harder question.
It may also be the question that separates firms that merely adopt AI from those that are genuinely transformed by it.
