When the workload grows, the default answer is another hire. Sometimes that is exactly right. But often the new person inherits the same manual process, and six months later the team is asking for another.
The question is not “people or technology?” It is “which parts of this work need judgement, and which parts are just movement?”
Start by splitting the work
Take the process that is straining the team and break it into steps. For each one, ask:
- Is it repetitive? Does it happen the same way many times a week?
- Is it rule-based? Could you write down how to do it for a new starter?
- Does it involve moving information from one place to another?
- Does it need judgement, relationships or context that lives in someone’s head?
Steps that are repetitive, rule-based and about moving information are strong candidates for automation. Steps that need judgement are where people create value, and where you want your team spending more of their time.
Where AI changes the picture
Until recently, automation stopped wherever information was messy: free-text emails, varied documents, customer messages. Modern AI can read, classify and draft from exactly that kind of input. That moves a new set of steps into the “can be automated” column: sorting incoming requests, extracting details from documents, drafting first responses for a person to approve.
The pattern that works is usually AI prepares, a person decides. It keeps quality high while removing the slow, repetitive part.
When hiring is the right answer
- The work is mostly judgement, relationships or expertise.
- The process changes so often that automating it would be rebuilding it constantly.
- The volume is genuinely low, and the cost of building exceeds the cost of doing.
In those cases, a good hire is the better investment, and a good system helps them be effective from their first week.
A simple test
If you described the job to a new hire and most of it sounded like “copy this from here to there, check it matches, then tell someone”, you are about to hire a person to do a machine’s work. Fix the process first, then hire for the part that needs a human.
This is the kind of decision we work through in the Scale path, before anything is built.
A worked example
Imagine an operations team that receives service requests by email, WhatsApp and a web form. Today, one person reads each request, copies the details into a spreadsheet, checks the customer’s record in another system, assigns the job and sends a confirmation. Volume is growing, and the team wants a second hire to keep up.
Split the work and the picture changes:
- Collecting requests from three channels is movement. It can be automated.
- Extracting the details from free-text messages used to need a person. AI can now do it reliably, with a person checking unusual cases.
- Looking up the customer record is movement. An integration can do it instantly.
- Deciding who should take the job, when it depends on skills, location and priorities, is judgement. It stays with a person, now with all the information in front of them.
- Sending the confirmation is movement. It can be automated once the job is assigned.
The team may still need to grow eventually. But the next hire spends their day making decisions and serving customers, not copying data, and the organisation can handle far more volume before that hire is needed.
Five questions before you decide
- How much of this role is movement, and how much is judgement?
- Would the process still make sense if volume doubled?
- What happens to the work when the person who knows it is away?
- How often does the process change?
- If we automated the movement, what would we want people to do with the time?
What good automation feels like to the team
Automation fails when it is done to a team rather than with them. The versions that stick share a few traits:
- People can see what happened. Every automated step leaves a clear record, so nobody has to wonder whether a message was sent or a record updated.
- Exceptions go to a person, with context. When something does not fit the rules, it arrives on someone’s desk with the information needed to decide.
- It can be switched off. If something goes wrong, the team can fall back to the manual process while it is fixed.
- The team helped design it. The people who did the work know which steps are truly routine and which only look routine.
Common mistakes
- Automating a broken process. If the manual process is confusing, automating it produces confusion faster. Simplify first.
- Trying to automate everything. The last ten percent of unusual cases often costs more to automate than the first ninety. Leave them to people.
- Forgetting maintenance. Tools change, forms change, suppliers change. Someone needs to own the automation after launch.
- Measuring only speed. Fewer errors, fewer dropped requests and less stress on the team are often worth more than time saved.
Where to start
Pick the process that is growing fastest and frustrating people most. Split it into movement and judgement, automate one or two of the heaviest movement steps, and measure. That single, visible improvement usually makes the next decision, automate further or hire, much easier to take.
Thinking about cost the right way
The comparison is rarely “a salary versus a software bill”. A fair comparison looks at the whole picture over a year or two:
- A new hire brings salary and on-boarding time, but also judgement, relationships and flexibility. They can handle work nobody predicted.
- Automation brings a build cost and some ongoing maintenance, but it scales with volume, works every hour of the day and makes the same step the same way every time.
- Doing nothing has costs too: overtime, burnout, slower responses, errors and the growth you cannot take on.
In many organisations the best answer is both, in the right order: automate the movement first, then hire people whose time goes into the work only people can do. That hire becomes more valuable, and easier to justify, because they are not spending their week as a human copy-and-paste function.
What to tell your team
Automation projects make some people nervous about their jobs. Be clear and honest about the goal: removing the tedious parts of the work so the team can handle more, respond faster and spend time on what they are good at. Involve them early, credit their knowledge of the process, and show them the results. Teams that help design their automation usually become its strongest supporters.
Quick answers
Will automation replace my staff? In most growing organisations it changes what staff spend their time on rather than removing roles. The usual outcome is handling more volume with the same team.
What should we automate first? The step that happens most often, follows clear rules and frustrates people most. It gives the quickest, most visible return.
Do we need AI, or just automation? Use plain automation for steps that follow fixed rules. Add AI only where the input is messy, such as free-text messages or varied documents, and keep a person reviewing its output until it has earned trust.
Not sure what should be automated?
