Skip to content

What to automate first

Tell us about one process that feels too manual.

Answer 12 quick questions about how the work happens today. We'll tell you whether the simplest useful next step is automation, integration, software, AI — or nothing yet.

  • Takes about 2 minutes
  • No email needed to see the results

This tool needs JavaScript. Email us the process instead and we will answer the same questions ourselves.

Email us

We won't recommend AI just because AI exists.

Describe the process, not the sensitive data inside it. No passwords, customer names or account details.

How it decides

12 questions about one process, and a set of rules we wrote down. No model, no score out of a hundred: the same answers always produce the same recommendation, and every line of the result traces back to an answer you gave.

  1. How many hours a month does it really cost?

    How often it runs, times how long one case takes. Two hours a month is a different decision from two days a week — whatever the technology could do.

  2. Is the process settled enough to automate?

    Clear rules, data everyone trusts, a workflow that held still this quarter. Automating an unsettled process makes it run faster, not better.

  3. How often does a case need a person?

    That share decides it: automate everything, or automate the routine and send the rest to someone who can judge it.

What it can tell you

There are 9 possible answers, and 3 of them are reasons not to build anything yet. That is deliberate: most automation advice has one answer and finds a way to reach it.

Don't build yet3

Leave it alone
The manual effort is small next to the cost of building an automation and keeping it working. The result says what would change that — more volume, longer cases, or a mistake that starts costing money.
Clarify the process first
The steps are still moving or different people apply different rules. Automation would make an unclear process run faster rather than better, so the first step is agreeing how the work should go.
Settle the data first
Systems disagree about the same record, or nobody is sure which copy is authoritative. Synchronising data nobody trusts spreads the problem faster.

Use what you already have3

A rule-based automation
Repetitive work, clear rules, structured data. A rule engine or a scheduled job is enough — this is the case where custom software and AI are both the wrong starting point.
Integration or workflow
Your applications each do their job; the friction is the same information being carried between them by hand. The fix is a connection, not a new application.
Automate the routine, route the exceptions
Most cases follow rules and a meaningful share do not. The routine runs itself, and each exception reaches a person with the reason it was flagged, the context that decides it, and clear actions.

Build something new3

AI-assisted workflow
There is enough unstructured material — emails, documents, free text — for AI to classify, extract and prepare. The decision stays with a person where it carries money or a relationship.
Custom software
The tools cannot hold the process you actually run, and the workarounds have become the process. This is the case where building is the smaller job.
Possible agentic workflow
Rare, and worded carefully. Volume, stability, tool access and genuinely variable paths all present — and even then a deterministic workflow is simpler if the sequence can be written down in advance.

Common questions

How do I know which process to automate first?

Start with the one that repeats most and takes the most hands-on time, then check three things before building: whether the rules are clear enough to write down, whether everyone agrees which system holds the correct data, and what share of cases needs human judgement. A process that fails the second test should be settled before it is automated.

How much manual work justifies automating a process?

As a rough line, under about an hour a week rarely repays the cost of building and maintaining an automation — unless mistakes are expensive, the work is done by someone costly, or a delay reaches the customer. Above roughly a day a week, the case usually makes itself.

Do I need AI to automate a business process?

Usually not. If the inputs are structured — fields, numbers, statuses — and the rules can be written down, rules and integrations are cheaper, faster and easier to verify. AI earns its place when the work involves reading emails, documents or free text, and even then its useful role is often to prepare a decision rather than make it.

What is the difference between automation and integration?

Automation performs a task nobody has to do by hand. Integration moves information between systems that already exist. If your complaint is that the same data is typed into several applications, the answer is usually integration: the applications work, the connection between them is missing.

Should I replace my systems or connect them?

Replace when the current system genuinely cannot support how you work — missing capabilities you need daily, no maintenance, or it cannot handle your volume. Connect when each application does its job and the friction is only the manual carrying between them. A migration changes how everyone works; an integration changes almost nothing visible except the copying.

Is this tool free, and do you need my email?

Free, and no. The questions, your answers and the result stay in your browser; nothing is sent to us while you answer. An email address is only asked for afterwards, if you want the written assessment sent to you.