Do we need a new system before we can automate anything?

No. Many automations connect the tools you already have: your accounting software, email, spreadsheets that become a simple database, and your bank or payment provider. A new system is only needed when the data underneath is too scattered to work with.

Will automation replace our staff?

In our experience it replaces tasks, not people. The copying, chasing and checking disappears; the people who did it move to work that needs judgement, such as customers, exceptions and decisions.

How do we measure whether an automation worked?

Measure it before you start: how long the process takes from start to finish, how many hours people spend on it each week, and how often it goes wrong. Measure the same three numbers a month after it is live.

Which processes to automate first

Automation has a reputation problem. Leaders hear about it, approve a project, and six months later the organisation has an expensive tool that one person understands and nobody trusts. The technology was rarely the problem. The choice of what to automate was.

Here is the method we use with clients to choose the first processes. It takes an afternoon.

List the work that repeats

Ask each department head one question: what does your team do every week that follows the same steps every time? Write every answer down without judging it. In most organisations the list looks something like this:

Typing the same order into the sales system, the warehouse sheet and the accounts tool
Chasing approvals for purchase requests, leave or payments
Matching payments received against invoices
Building the weekly or monthly management report
Answering the same customer questions about status, prices and documents
Reading invoices, receipts and forms and typing what they say

Score each one

Give each process a score from 1 to 5 on four things:

Volume. How often does it happen? Daily scores high; once a quarter scores low.
Pain. How much time does it take, and how much does it frustrate people or customers?
Errors. How often does it go wrong, and what does a mistake cost?
Clarity. Could you write the steps on one page? Processes that depend on judgement at every step score low.

Multiply volume by pain, add errors, and keep clarity as a filter: anything that scores 1 or 2 on clarity is not ready to automate yet. It needs to be simplified first. The top three on the list are your first projects.

The four kinds of work that pay back fastest

Across sectors, the same four kinds of work almost always rise to the top.

Approvals

A request that waits on someone’s desk is the most common delay in any organisation. Routing requests automatically, by amount, department and type, to the right person, who can approve from a phone, often turns days into hours. Every step is recorded, which auditors appreciate.

Reconciliation

Matching payments to invoices, bank statements to the ledger, or deliveries to orders is high in volume, low in judgement and costly when it goes wrong. Automated matching handles the routine cases and puts only the exceptions in front of a person.

Reporting

If someone rebuilds the same report every week from the same sources, the report should build itself. Once the data is connected, a dashboard is always current and a summary can arrive by email before the Monday meeting.

Documents

Invoices, receipts, application forms and identity documents can now be read by AI models with good accuracy and turned into structured data. A person checks what the model is unsure about instead of typing everything.

Where AI fits, and where it does not

AI is useful where the input is messy: reading documents, sorting incoming messages, drafting replies from your own policies, and summarising long records. It is not a substitute for clear rules. If the process is “approve purchases under a set amount automatically”, write the rule; do not ask a model to guess.

Two rules keep AI safe in a business process. First, it works from your own data and policies, not the open internet. Second, a person stays in the loop wherever a decision affects money, health or someone’s rights.

Why automation projects fail

Automating a bad process. If the steps do not make sense on paper, software only makes the confusion faster. Simplify first.
No owner. Every automation needs a named person in the business who decides how it should work and notices when it does not.
Everything at once. Start with one process, put it live, measure it, then move to the next. Confidence spreads faster than any announcement.
Invisible failures. An automation that fails silently is worse than none. Every automated flow should report what it did and alert someone when it cannot finish.

Start small, measure, repeat

Pick the top process from your list, measure how it runs today, and put an automated version live within a few weeks. If you would like help scoring your list or building the first one, send us a brief. We will tell you which process we would start with, and why.