Skip to content
The Best Workflow Is Not Always the Shortest One

There is a natural tendency in automation to optimise for fewer steps: remove clicks, remove handovers, remove people, make the workflow shorter. Sometimes that improves the process. Sometimes it only improves the diagram.

A process can become faster while losing useful checks, context, judgement or control. It can look more efficient on paper while becoming more fragile in the real world. The objective should not be to create the fewest possible steps, but to get the work done in the best possible way.

Every step should earn its place

Some steps exist because nobody has questioned them for years, and those should absolutely be challenged. But other steps exist for a reason.

A person may review something because experience matters. A second system may be checked because the first does not contain enough information. An approval may exist because the financial or operational consequence justifies it. A validation step may only catch a small percentage of cases, but those cases may be the expensive ones.

Removing those steps can make a workflow look cleaner without making the process better. Good automation is not about preserving unnecessary work. It is about understanding why each step exists before deciding whether it should stay, change or disappear.

Automating a person away is not automatically an improvement

There is still a tendency to describe automation as a journey towards removing human involvement. We do not think that is the right way to look at it.

Sometimes AI can perform a step faster and more consistently. Sometimes a rule or API is clearly the better option. But there are also steps where a person is simply the best participant.

That may be because of experience, empathy, commercial understanding, negotiation, accountability, or simply because the work does not happen often enough to justify building something more complex around it. Replacing that person with automation can increase cost, complexity or risk rather than reduce it.

The better question is not, “Can we automate this step?” It is, “Who or what should do this step?”

Different participants are good at different things

Modern processes are rarely purely human or purely automated. They may involve AI, workflows, APIs, business systems, rules and people, sometimes several of each. That is not a weakness in the process. It is often what makes the process work well.

AI may classify or interpret information. An API may retrieve data. A workflow may coordinate the work. A person may handle a conversation, make a decision or complete a task. Another system may perform the final transaction.

The value comes from choosing the right participant for each part of the work and making sure they can operate together. Trying to force the entire process into one technology usually creates compromises somewhere else.

Optimising for speed can create expensive mistakes

A workflow that finishes in thirty seconds is not necessarily better than one that takes two minutes. The extra time may contain the most important part of the process.

Perhaps one step validates a customer request before money is moved. Perhaps another checks whether an automated answer is supported by the available information. Perhaps a person reviews a high-value case before an irreversible action is taken.

These steps create friction by design, and that is not always a bad thing. The question is whether that friction protects something valuable.

There is a big difference between unnecessary friction and useful control. Good process design knows the difference.

Complexity should exist where it creates value

This does not mean workflows should become unnecessarily complicated. Quite the opposite. Every step should have a reason to exist, and if it does not contribute to the outcome, remove it.

But simplicity should be judged from the perspective of the business outcome, not from the number of boxes in a workflow diagram. A process with twelve well-designed steps may be much simpler operationally than a five-step process that constantly creates exceptions, manual investigation and rework.

The shortest workflow can become the longest process once something goes wrong.

This is how we think about automation in Cention

In Cention, the goal is not to force everything through AI or remove people wherever possible. The goal is to execute the work using the participants that make the most sense.

That can mean workflows, AI, APIs, MCP, business systems, rules or people. They can participate at different points in the same case, and they can change as the process evolves. The case keeps the work together while those participants do their part.

That gives organisations the freedom to design the process around the outcome rather than around the limitations of a particular technology.

If a human step creates value, keep it. If AI can do it better, use AI. If a simple rule is enough, do not use AI at all. If a system can complete the work directly, let the system do it.

The architecture should support the process, not dictate it.

The best workflow is the one that gets the work done properly

There is nothing wrong with shorter workflows, but there is also nothing inherently good about them. The same is true for automation rates, AI usage and the number of human steps in a process. They are useful measurements, but they are not the outcome.

The outcome is whether the work was completed correctly, efficiently and with the right level of control.

Sometimes the best workflow is shorter. Sometimes it is not.

The important thing is knowing why.

Workflow Automation Process Automation Enterprise Automation Case Management AI Automation Human-in-the-Loop Digital Transformation
← Back to Blog