The cost of automating a process comes down to a handful of factors: how many processes are involved and how complex they are, how many exceptions they have, which systems need connecting, the state of the data and whether AI is required. That is why an honest figure can only come from sizing the real process. What you can work out before asking for a proposal is what the process costs you today without automation, and that is the benchmark for your decision.
What drives the cost of automating a process
What sets the cost of an automation project is the work the process demands, not the name of the technology. These are the factors that weigh most:
| Factor | What makes it more expensive | What makes it simpler |
|---|---|---|
| Number and complexity of processes | Several processes at once, with many steps and departments | One well-bounded process with a clear start and finish |
| Exceptions and business rules | Decisions that rest on one person’s judgement, frequent special cases | Stable rules, written down and agreed by the team |
| Systems to integrate | Many systems, or legacy applications without an API | Few systems, with documented APIs |
| Data quality and dispersion | Data scattered across spreadsheets, inboxes and paper, with duplicates | Digital data held in a single system of record |
| Volume | Sharp peaks that call for more resilience and monitoring | Steady, predictable volume |
| Artificial intelligence | Highly varied documents, no past examples, costly errors | Rules that suffice, or plenty of historical examples |
| Infrastructure | Environments with their own policies to adapt to | Infrastructure already in place, yours or ours |
| Security | Sensitive data, audits and undefined access profiles | Clear requirements, written down at the outset |
| Evolution after go-live | A process whose rules change often | A stable process and a well-documented system |
No factor is good or bad in itself. Together they explain why two processes that look identical, such as invoice processing at two companies, can call for very different amounts of work.
Volume deserves a note of its own: it can add technical demands, but it is also what multiplies the saving.
Why a tool’s price is not a project’s cost
A licence pays for software. It does not pay for the work of making that software solve your process. Between the tool and the result sit five tasks that someone has to carry out:
- Analysing the process. Understanding how work gets done today, which exceptions crop up and what should change before automating.
- Integrating. Connecting the tool to your ERP, your CRM, email or the spreadsheets you already rely on.
- Testing. Running the system against real cases, the awkward ones included, until it behaves as it should.
- Supporting adoption. Training the team and adjusting whatever is needed so the system actually gets used.
- Maintaining and evolving. Adapting the system as the business, its systems or its volume change.
If a supplier does not do this work, your team does, and their time has a cost too. A tool that nobody integrates or uses carries a cost and no return. So always compare the total cost of the project, internal hours included, rather than the licence alone.
What it costs not to automate
The cost question has a second half that is often overlooked: what it costs to leave the process as it is. That cost never appears on an invoice, but you pay it every month:
- Hours. Repetitive work that could go to higher-value tasks.
- Errors. Corrections, returns, lost revenue or penalties.
- Delays. Customers kept waiting, month-end closes that drag on, decisions made on stale data.
- Reliance on key people. A process that stalls when someone is on holiday or leaves the company.
Example, with round, illustrative figures based on your own team’s internal cost:
- Three people each spend 40 hours a month keying emailed customer orders into the ERP: 120 hours a month, 1,440 a year.
- At an internal cost of €30 an hour, the process costs €43,200 a year.
- Each month 20 orders contain errors, and each takes an hour to put right: another 240 hours a year, or €7,200.
- In total the process costs €50,400 a year, before counting delays or the dependence on those three people.
That figure does not tell you what automation should cost. It tells you what not automating costs today, every year, and it is the benchmark against which to judge any proposal. One caveat: released hours are only a real saving if they go to other valuable work.
How to estimate the return before you decide
Once today’s cost is measured, the decision becomes a calculation. The annual saving is the difference between the cost of the process before and after. Subtract the system’s annual running cost and compare the result with the initial investment.
Everything hinges on the starting point. If today’s hours, errors and lead times are estimates, the return will be an estimate too. If they are measurements, the return is a figure your finance director can check. We set out the formula, the metrics and a worked example in how to measure automation ROI.
And ask for any proposal to state the indicators that will be used to measure the result.
What the cost of AI implementation depends on
AI is not a separate line item. It is one more part of the process, and its cost depends on the same factors, plus three further questions:
- What data it needs. If there are past examples of how similar cases were resolved, there is less work than if they have to be built.
- How much human review it requires. The higher the cost of an error, the more checks and testing are needed.
- Where the data is processed. On your servers or ours, according to your security policy. Either way, get it in writing that your data will not be used to train third-party models.
AI pays off where rules fall short: emails, scanned documents or free text. Where rules are enough, adding it only makes the project more expensive.
How we size a project at Neji
Before putting forward a figure, we need to understand the process. We do that in the first two steps of our method:
- First conversation. We get to know your company and identify where we can help.
- Specifications. We define the chosen process in detail: steps, exceptions, systems, data, volume, security requirements and success indicators.
With that information, we prepare a proposal shaped around that process rather than a catalogue. The project is turnkey: your team explains how it works and validates the solution, and we handle everything technical. The steps that follow, from development through to continuous evolution, are described in how it works.
How to reduce the investment
The same analysis that sets the cost shows where to bring it down. Five decisions help:
- Start with the process that offers the best ratio of impact to effort. Our article on which processes to automate first sets out five questions for choosing it.
- Standardise before you automate. Remove unnecessary steps and agree the rules. Automating a messy process only gets you a faster mess.
- Reuse the systems you already have. Integrating your ERP, CRM or email usually costs less than replacing them.
- Automate the common case. Rare exceptions can stay with a person, with the system preparing the information for them.
- Measure the starting point. Knowing what the process costs today stops you investing where the return is small.
Where to start
Pick the process that consumes the most repetitive hours and work out what it costs you today, using the same calculation as the example. That figure will tell you whether it is worth going further. If you would rather do it with us, request a first conversation so we can get to know your company and see where we can help.