The question we hear most often from growing businesses is which of their processes to automate. The honest answer is that it depends on more than how tedious the process is. Some processes automate cleanly and pay back quickly. Others look like ideal candidates and turn into a maintenance burden within months. The difference is mostly technical, and it can be assessed before anyone spends money.
The first question is whether the process follows clear rules. If a person can write down exactly how a decision is made, a system can follow it. If the decision relies on judgement and context, automation will either get it wrong or need a person to check every case, which removes most of the benefit.
The second is the quality of the inputs. Structured data such as a form, a database record or a standard file is easy to work with. Unstructured input such as free text emails, scanned documents and photographs is harder. Modern AI based tools can read many of these, which widens what is possible, but they still need a checking step and a measured error rate before you rely on them for anything that affects money or customers.
The third is how the automation connects to the systems involved. Some systems offer an API, which is a stable, documented way for one piece of software to talk to another. Others offer no such connection, and the only option is software that operates the screen the way a person would, clicking through fields and buttons. That approach is often called robotic process automation. It can work, but it is fragile, because a small change to a screen layout after a software update can stop it working without warning.
The fourth is the exception rate. If most cases follow the standard path and a few do not, automation handles the majority and people handle the rest. If a large share of cases are exceptions, you end up with a system that still needs constant human attention, and often more coordination than before.
The fifth is stability. A process that is about to change, because of a system replacement, a reorganisation or new regulation, is not ready. Automating it means paying to build something that will have to be rebuilt.
Technical readiness is only half the picture. The business case needs three inputs that are often left out. The first is volume, meaning how many times the process runs and how long each run takes now. A process that runs twenty times a year is rarely worth automating however tedious it is. The second is the cost of error, since automating something where mistakes are expensive can be more valuable than automating something that is merely slow. The third is the ongoing cost.
Every automation needs an owner, monitoring and maintenance. Somebody must know when it fails, and there must be a manual fallback for the day it does. If nobody has been assigned that responsibility, the automation will quietly break and the team will quietly return to doing the work by hand, which is how many automation projects end up abandoned.
We score each candidate process from one to five on volume, clarity of rules, quality of inputs, stability of the systems involved, exception rate and cost of error. The scores are not a formula that gives a final answer. They make the discussion concrete. A process with high volume, clear rules and clean inputs is a strong candidate. A process with high volume but a high exception rate probably needs to be redesigned first. A process with low scores across the board is best left manual, and it is useful to record that as a deliberate decision.
Consider two processes that both look like automation candidates. The first is entering supplier invoices into an accounting system. If most invoices arrive in a small number of standard formats, the fields are predictable and the rules for coding them are clear, this scores well. The exceptions, such as invoices with no purchase order or unusual tax treatment, can be routed to a person, and the volume makes the saving real.
The second is triaging customer complaints. It runs often and feels repetitive, but every case depends on context, the wording is unstructured and the right response varies with the customer’s history. It scores poorly on rule clarity and input quality. A better first step is to standardise how complaints are logged and categorised, which improves the data and may allow limited automation later.
Both examples are common patterns rather than specific client cases, and they show why the scoring is more useful than instinct. The second process is more annoying to staff, but the first is the better investment.
A low score does not mean the process is a lost cause. It usually points to a specific piece of work. Where rules are unclear, the answer is to agree and document them. Where inputs are messy, the answer is to standardise how information is captured at the start. Where the process has too many exceptions, the answer is to understand why they occur and remove the causes. Where staff do the work differently from one another, an SOP that sets out the agreed way often delivers the largest single gain, and it is a prerequisite for automation later.
One pattern is consistent. Processes that have been mapped and simplified automate well, and processes that have not been mapped tend to carry their confusion into the automation. Steps that only existed to work around an old limitation get automated along with everything else. That is why we start with the current state map and remove unnecessary steps before deciding what to automate.
We run automation readiness assessments as part of our digital transformation and process improvement work. We map the processes, score them against these criteria, and deliver a ranked roadmap that shows what to automate now, what to redesign first and what to leave alone, with the reasoning written down. We are vendor agnostic, so tooling recommendations follow from the requirements and not from a preferred supplier.
If you are weighing up where automation would actually pay back in your business, a free 30 minute consultation through groowfuse.com is a good place to start.