Council Post: Every Business Process You Run Is Already Legacy

Jakob Freund is the CEO of Camunda, a software company innovating end-to-end process orchestration and automation with agentic AI.

getty

If you designed your most critical business process from scratch today, knowing what AI can actually do, would you build it the way it currently runs? For almost every process in almost every enterprise, the honest answer is no.

Every business process running in your organization today was designed before modern AI could reason, act and coordinate across systems without human intermediaries. For years, you built approval chains, documentation requirements and handoff sequences around a single assumption: humans would need to fill every gap software couldn't close. That assumption is now outdated. The processes, however, remain.​

Age and inefficiency alone do not make a process legacy. A process becomes legacy when its design assumptions no longer reflect what is possible. Every organization that wants to lead in the next decade must move past simply deploying AI tools to make tasks more efficient. Leaders have to reconsider whether certain tasks should exist at all, who should do them and how the whole operation should function now that AI agents can take action.

Adding AI To Existing Processes Creates An Automation Ceiling

The natural response to AI has been to add it to existing processes. For example, an AI assistant or chatbot might be layered onto a customer support workflow already straining under its own complexity. With this approach, productivity improves in isolated pockets, and some teams might feel initial momentum. Then the ceiling appears.

When systems were never designed to share context, processes become impossible to coordinate. Exceptions accumulate, and the cost of handling them reappears in new forms. A field experiment from INSEAD and Harvard Business School, published earlier this year, studied 515 companies given identical AI tools, budgets and technical training, and found a significant divergence depending on whether companies restructured their operations around AI or deployed AI within existing processes. The firms that restructured generated nearly twice as much revenue and required 40% less capital to achieve the same growth. The researchers found that mapping AI onto a process designed without it produces a faster version of the old process, not a better one.

Think About Re-Engineering, Not Optimization

Most organizations are stuck in this optimization phase of AI. Optimization asks how a step can run faster or cheaper. Re-engineering asks whether the step should exist at all. Older processes often carry redundant approvals simply because an earlier system failed to earn trust. Some preserve unnecessary handoffs because integrating systems a decade ago cost too much to justify. Analyzing a process through an AI-first lens almost always reveals human involvement that adds no judgment (and ultimately adds delays).​

At Camunda, we tested this on our own operations. We put our Quote-to-Cash process through a full re-engineering cycle: a process spanning five departments, six systems and up to 80 manual handoffs per contract. A team of four people completed the full discovery, re-engineering, build and deployment cycle in four weeks. Discovery historically required two to three months of stakeholder interviews, but the new approach completed it in 10 days. Manual touchpoints per contract fell from as many as 80 to just two or three. Error rates dropped by roughly 80%, and the process saved approximately 6,000 person-hours per year.​

What CIOs Should Do Right Now

A typical enterprise runs more than 500 core business processes. Traditionally, re-engineering one took approximately 12 months. AI advances in cycles of weeks. CIOs who wait for a comprehensive enterprise program before starting will find their competitors have already built organizational memory that compounds with every process they complete.

Here are four steps CIOs can take starting today.

1. Start with the process everyone already knows is broken. Begin with a visible process instead of a low-stakes pilot. The goal is to spend organizational energy on an effort that both drives change and gains leadership buy-in.

2. Separate the discovery phase from the build phase. AI-driven discovery can now reconstruct how a process actually runs, as opposed to how documentation says it runs, in one to two weeks. Treating discovery as a distinct, bounded exercise accelerates stakeholder alignment and speeds the path to production.

3. Appoint someone who owns delivery with no departmental bias. Re-engineering initiatives stall when they accumulate stakeholders who protect their slice of the existing process. A single owner accountable to outcomes clears that obstacle faster than any technology choice.

4. Build organizational memory deliberately. Every process you re-engineer encodes institutional knowledge about your systems, integration patterns and decisions. That knowledge makes the next re-engineering faster and more precise.

Come back to this most important question, “If I built this process today, knowing what AI can do, would I build it this way?” The organizations that answer honestly and act first will set the baseline the rest of the market chases.

Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?