Before You Automate, Understand the Work
Automation works best when it follows a clear understanding of how work really happens, including the judgement, exceptions and friction hidden by the formal process.
Automation projects often begin with a deceptively simple question: what could this technology do for us?
It is an understandable place to start. New tools make new possibilities visible, and a working demonstration creates momentum. But it also puts the proposed solution ahead of the work it is meant to improve. The result can be an efficient implementation of the wrong process, or a system that performs well until it meets the exceptions that occupy much of an organisation’s real attention.
Before deciding what to automate, it is worth understanding what the work actually is.
The documented process is only a partial account
Most organisations can describe a process as a sequence of activities. A request arrives, somebody reviews it, a decision is made and an outcome is recorded. That description may be accurate, but it is rarely complete.
The work also includes the email sent to recover missing information, the spreadsheet maintained because two systems do not agree, the colleague consulted when a case does not fit the rules, and the judgement used to decide whether an apparent anomaly matters. It includes waiting, rework and the informal coordination that keeps the formal process moving.
These details are easy to dismiss as noise around the process. In practice, they often reveal where the organisation creates value and where it carries risk. They show which decisions depend on context, which handovers lose information, and which controls exist for reasons that are no longer visible.
Automating the documented sequence without examining this surrounding work does not remove the complexity. It hides it inside a new system, pushes it downstream, or leaves people to manage it through another set of workarounds.
Efficiency is not the same as improvement
Automation is usually justified through speed, consistency or cost. All three can matter. None is enough on its own to show that the work has improved.
A faster approval process is not better if it increases avoidable errors. A consistent response is not better if the cases require different treatment. A lower handling cost is not better if customers have to make repeated contact to achieve the outcome they need.
The useful question is not simply whether an activity can be performed with less human effort. It is whether changing that activity improves the performance of the system as a whole.
That requires a wider view. What demand is the process trying to meet? Where does information lose meaning as it moves? Which decisions affect the eventual outcome? What causes work to return, wait or escalate? How will the organisation know that the change has helped?
Answers to these questions create a basis for choosing technology. Without them, a team may be able to measure how much has been automated while remaining unable to say whether anything is genuinely better.
Judgement should be made visible
The rapid development of AI makes this especially important. Software can now classify, summarise, recommend and act across tasks that once appeared resistant to automation. That expands the opportunity, but it also increases the need to understand where judgement sits in the work.
Judgement is not a mysterious quality that must always remain human. Nor is it an inconvenience to be removed wherever possible. It is a set of decisions made with particular information, experience, constraints and consequences. Some of those decisions can be supported effectively by AI. Some can be delegated within clear boundaries. Others should remain with a person because the context is uncertain, the consequences are significant or accountability cannot sensibly be transferred.
Making judgement visible allows those distinctions to be designed deliberately. It helps a team decide what the system may do, what evidence it must provide, when it must ask for help and who remains responsible for the outcome.
A better sequence
A more reliable approach begins with the work and moves towards the technology:
- Observe how the work happens, including exceptions, delays and informal coordination.
- Identify the demand being served, the outcomes that matter and the evidence available.
- Improve obvious sources of friction before encoding them in a system.
- Decide where automation, AI assistance or human judgement best fits.
- Introduce the change with measures that reveal its effect on the whole flow of work.
This sequence need not become a long programme of analysis. A small amount of disciplined observation can challenge assumptions that would otherwise shape months of implementation. The aim is not to delay action, but to make action more purposeful.
Technology can perform an increasing share of organisational work. The enduring management task is to ensure that this capability serves the right purpose, operates within sensible boundaries and improves outcomes that matter.
Understanding the work first is what makes that possible.