Automation Checklist for Israeli Small Businesses: What to Fix First

A business asking for automation almost always knows what annoys it. Far fewer know what it costs them. Those are not the same thing, and the difference decides whether the project pays for itself.

Step 1: measure before deciding

Before discussing tools, answer three questions about each recurring task:

1. How often does it happen per month?
2. How long does it take each time?
3. What happens when it is done wrong?

The third is the important one. A task that takes a minute but where a mistake costs a working day is worth more, in automation terms, than an hour-long task that never breaks.

A concrete example: manually keying payment details onto an invoice. Takes two minutes. But a wrong digit in the amount, or matching to the wrong transaction, produces a broken bank reconciliation, a customer chased for a debt they paid, and hours of investigation. The task is small; the cost of the error is not.

The threshold

Below a few hours a month, do not automate.

This sounds counterintuitive but holds: every automation carries a build cost, a maintenance cost when something on the other side changes, and the cost of the moment it breaks and nobody notices for a week. A two-hour-a-month task does not cover that.

The only exception: a task cheap in time but expensive in errors - like the example above.

Step 2: fix before automating

This is the most common mistake.

If the manual process is unclear - two people do it differently, there are edge cases nobody has decided, someone "always checks something manually" - automation will not fix that. It will execute the confusion faster and with perfect consistency.

The test: can you write the process as a list of steps a new person could follow without asking questions? If not, that is what to do first, and it is not a technology project.

Step 3: the order that pays back fastest

In a typical Israeli small business, this is the order I have seen pay back soonest:

First: the money flow

Quote to invoice to payment to reconciliation to collection reminder.

Why first: errors here cost cash, not just time. An invoice that never went out is revenue never collected. A reminder sent to a customer who already paid is relationship damage.

The most common gap: payment arrives in one system while the invoice is issued in another, and someone keys between them.

Second: what the customer sees

Order confirmation, status update, appointment reminder, first response.

Why: the cost here is not hours but customers. An enquiry unanswered for two days is a customer who went elsewhere.

In Israel this is almost always WhatsApp rather than email - and that changes the technical choice.

Third: moving data between systems

The thing someone does with an intermediate spreadsheet.

The tell: "I download from here and upload to there". That is exactly what integration solves, and it usually has a known cost.

Fourth: reporting

The report someone assembles by hand each month.

Why last: because it is usually asked for once a month, and the total time is smaller than it feels. Worth automating - but after the three above.

Step 4: what to leave manual

A short list of things worth not automating:

• Judgement. Approving an unusual discount, deciding on a refund, handling a complaint. You can automate gathering the information for the decision; the decision itself, no.
• What happens twice a year. No payback.
• A process about to change. If you are considering swapping a supplier or system in the next six months, the automation goes out with it.
• What nobody really understands. If the manual process is knowledge in one person's head, that has to come out first.

Step 5: the questions before building

For each automation that survived the filter:

1. Do the systems you need to connect have an API? If not, there are routes, but they change the budget.
2. What happens when it fails? Who knows? What happens to what did not go through? This is not an end-of-project question - it is half the work.
3. Who maintains it? Vendors change interfaces. Without an answer here, the automation works until it doesn't.
4. What is the cost against the saving? Including maintenance, not just build.

Step 6: check the ground is not fighting you

Two Israel-specific points that derail projects mid-way:

• Hebrew and RTL. Names with double spaces, invisible directional characters, inconsistent encoding. This is not an aesthetic problem - it breaks string comparison and therefore record matching.
• The system of record. If everything rests on a spreadsheet, the automation will rest on it too. Sometimes the right answer is not to automate but to move the foundation.

The checklist

1. Measure: frequency x time x cost of error. Not a feeling.
2. Filter: under a few hours a month, skip - unless errors are expensive.
3. Fix: a process you cannot write as a list of steps is not ready to automate.
4. Order: money, then customer, then data, then reporting.
5. Exclude: judgement, rare, temporary, not understood.
6. Ask: API? failure? maintenance? cost?
7. Check the ground: clean Hebrew, a stable system of record.

A map of the systems common in Israel helps establish what connects to what before deciding.