Most growing companies do not have an automation problem. They have a "where do we start" problem. The systems work well enough on their own, the team is capable, and yet every month the same hours disappear into copying data between tools, chasing approvals, and rebuilding the same report. You know there is waste. The hard part is deciding which of the dozen candidates to tackle first, before you commit budget to a build.

This guide gives you a way to decide. We build automations for mid-sized companies for a living, and the pattern that separates a project that pays for itself from one that quietly fails is almost never the technology. It is the choice of what to automate, and in what order.

What process automation actually means for a growing company

Business process automation is the use of software to carry out a repeatable business process with little or no manual effort. That is the textbook definition. The part that matters for a company of fifty, a hundred, or two hundred people is the word "process," because at your scale the most expensive work is rarely a single task. It is the hand-off.

A task is one person doing one thing: entering an order, sending an invoice. A process is the chain: a quote becomes an order, the order updates the warehouse, the warehouse triggers an invoice, the invoice lands in the accounts. The places where that chain breaks are the joints between systems and departments, where someone exports a file from one tool and re-keys it into another. This manual shuttling between systems that should talk to each other has a name: the swivel-chair problem. It is where mid-sized companies lose the most time, and it is invisible on any single org chart.

So when you ask which processes to automate, the sharper question is: where does work stall, get re-entered, or fall through the cracks as it crosses a boundary?

Start with the whole picture, not a task list

The most common mistake is to start by listing tasks that feel annoying. Annoying is not the same as expensive, and a task that irritates one person may matter far less than a silent hand-off that costs three departments an afternoon a week.

Before you score anything, map how work flows end to end. You do not need a formal modelling exercise. Walk one real example through the business: take a single order or project from first contact to paid invoice, and write down every system it touches and every point where a human moves it along by hand. Do the same for two or three of your core flows. Two patterns will jump out immediately: the steps where data gets copied between tools, and the steps where work waits for someone to notice it.

This is the part competitors skip, and it is the part that decides everything. A list of tasks tells you what is irritating. A map tells you what is costing you. It also shows you the processes that should not be automated at all, which we will come back to.

The signs a process is worth automating

A process is a strong automation candidate when it is repetitive, rule-based, and high in volume, and when it currently relies on manual effort that is slow or error-prone. The clearest candidates share most of these traits:

• It happens often. Frequency multiplies the return. A five-minute task done two hundred times a month beats a two-hour task done once.
• It follows rules. The steps are consistent and the decisions can be written down, rather than depending on judgement that changes every time.
• It has clear inputs and outputs. You can describe exactly what goes in and what should come out.
• It crosses systems or people. Every manual hand-off between two tools or two teams is a place automation removes delay and error at once.
• It is error-prone today. Manual data entry breeds mistakes, and the time spent finding and fixing one wrong figure often dwarfs the time the task itself took.
• It is slow in a way customers feel. A quote that takes three days, an enquiry that waits a day for a reply. In higher-value sales, the first credible response often wins the deal.

A useful shorthand from the automation world is the five D's: work that is dull, dirty, dangerous, difficult, or done repeatedly is work to hand to a machine. For most office-based companies, "dull" and "done repeatedly" are where the money is.

Judgement-heavy work is the opposite. Negotiating, designing, advising, managing people: these resist automation because the rules keep changing. McKinsey Global Institute estimated that only about 5 percent of occupations can be fully automated with current technology, while around 60 percent have at least 30 percent of their activities that could be. You are not automating jobs. You are automating the routine slices inside them.

The processes most growing companies should look at first

In our work, the same handful of processes come up again and again as the highest-value, most feasible places to start. If you are not sure where to look, check these:

• Lead and enquiry handling. Capturing every enquiry from every channel in one place and acknowledging it instantly, so nothing sits unseen and no hot lead goes cold.
• Quote and order processing. Turning a "yes" into a recorded order and a generated quote or invoice without re-keying, which is usually the biggest single source of swivel-chair work.
• Invoicing and reconciliation. Generating invoices from the source record and matching payments automatically, instead of by eye at month-end.
• Reporting. Pulling the few numbers leadership actually steers on into a live view, rather than assembling them from exports every week.
• Onboarding, whether of customers or staff. The same checklist of folder, contract, welcome, and setup, run the same way every time.
• System synchronisation. Keeping the ERP, the CRM, and the finance tool holding the same truth, so a change in one appears in the others on its own. An automation layer such as n8n is well suited to this connective role, sitting between systems and reacting to events without replacing what you already run.

None of these is exotic. That is the point. The best first automations are boring, frequent, and rule-based, not clever.

Eliminate, simplify, connect, then automate

Here is the rule that saves the most money and that almost no guide states plainly: do not automate a process you should be removing. Automation makes a process faster and more permanent. Apply it to a broken or redundant process and you have just paid to entrench the problem.

Before you automate anything, run each candidate through four questions in order.

By the time a process reaches the fourth step, it is lean, it is necessary, and it is the right thing to build. Skip the first three steps and you automate chaos. Newer McKinsey research makes the same operator's point: their 2025 analysis found that currently demonstrated technology could in theory automate around 57 percent of US work hours, but turning that ceiling into real gains takes workflow redesign and change management, not just tools. The redesign comes first.

How to prioritise: impact against effort

Once you have a clean list of genuine candidates, prioritise them on two axes: impact and effort. Impact is what the process costs you today in hours, money, errors, lost revenue, and risk. Effort is what it will take to automate, including cost, complexity, and disruption to live operations.

Plot each candidate roughly into one of four groups:

• High impact, low effort: do these first. These are your quick wins, the proof that earns trust and frees time to tackle the rest.
• High impact, high effort: plan these as projects. These are the strategic plays that change how the company scales, and they deserve a proper build.
• Low impact, low effort: fit them in when convenient. Pleasant, not urgent.
• Low impact, high effort: leave them. Being willing to say "not worth it" is part of doing this well.

You do not need a precise ROI model to start. A rough estimate of hours saved per month, multiplied by a loaded hourly cost, against a rough build cost, is enough to rank the list and spot the obvious wins. Refine the numbers on the few candidates that make the shortlist.

Build a roadmap you can actually execute

Turn the prioritised list into a sequence, not a wish list. A roadmap that a busy team can follow has three phases.

Phase one is quick wins: the high-impact, low-effort automations you can ship in days or weeks. Lead them, because early relief builds the internal confidence that carries the harder work.

Phase two is foundations: the connective tissue. Connecting the core systems, establishing one source of truth, automating the reporting. Medium effort, large and lasting payoff.

Phase three is the higher-value plays: the end-to-end, organisation-wide automations that let you grow without adding headcount for every increase in volume.

For each item, name the outcome in plain terms, a rough effort level, and who owns it. Some of this your own team can start. Some needs a specialist to design and build well. Being honest about which is which is what keeps a roadmap credible and keeps you from stalling halfway.

Common mistakes when choosing what to automate

The failures we see are almost always choices, not code.

Automating the wrong thing: picking the loudest annoyance instead of the costliest hand-off. Automating chaos: building on top of a broken process you should have fixed or removed. Tool-first thinking: buying a platform and then hunting for something to use it on, rather than starting from the work. Over-engineering: a custom build where switching on a feature you already pay for would have done. Ignoring the people: shipping an automation no one was trained for or bought into, which quietly gets worked around.

Avoid these and you avoid the large majority of automation projects that disappoint.

How to know what to fix first, with confidence

Identifying which processes to automate comes down to three moves: map how work really flows, find the repetitive rule-based hand-offs that cost you most, and sequence them by impact against effort so the quick wins come first. The discipline that matters most is the one to resist: do not automate a process before you have asked whether it should exist, be simplified, or simply be connected.

If you want the full picture before you commit budget, this is exactly what our Innovation Scan does. We map your operation across its systems and departments, show you honestly where the time and money leak, and hand you a prioritised roadmap with the low-hanging fruit first. It is complimentary, and it is yours to act on with or without us.

Frequently asked questions

How do you identify processes for automation?

Map how work flows across your teams and systems, then look for steps that are repetitive, rule-based, high in volume, and reliant on manual effort, especially the hand-offs where data gets copied between tools. Score each on impact against effort, and start with the high-impact, low-effort candidates.

How do you determine if a process can be automated?

A process can be automated when its steps are consistent, its decisions follow rules you can write down, and its inputs and outputs are clear. If a step depends on changing human judgement, it resists automation. Repetitive, predictable, rule-based work with defined inputs is the strongest candidate.

What are the five D's of automation?

The five D's are dull, dirty, dangerous, difficult, and done repeatedly. They are a quick filter for work to hand to a machine. For office-based companies, "dull" and "done repeatedly" point to the highest-value targets: the routine, frequent, rule-based tasks that quietly consume the week.

Which processes should you automate first?

Automate the processes that are high in cost today and low in effort to change. For most growing companies these are instant enquiry handling, quote and order processing, invoicing and reconciliation, reporting, and keeping core systems in sync. Quick wins first builds the confidence to tackle bigger projects.

Should you automate a process that is not working well?

No. Fix or remove it first. Automation makes a process faster and more permanent, so applying it to a broken or redundant process entrenches the problem. Run each candidate through eliminate, simplify, connect, then automate. Only automate what remains once the process is lean and necessary.

Do you need to replace your systems to automate?

Usually not. At mid-market scale the answer is more often to connect and automate what you already own than to rip and replace. An integration layer can sit between your existing tools and move data automatically, which is faster, cheaper, and less disruptive than a new platform.