Automate Data Entry: find the places your team types the same thing twice

To automate data entry, find every point where a person re-types information the business already holds, then remove the second keystroke rather than speeding it up. South African law treats that second copy as a new record, not the original, so duplicate typing is an accuracy and evidence problem before it is a time problem.

The question is not how fast you type, it is how many times

Most guides on this keyword are tool comparisons. They rank the same five products and assume the job is to make typing faster. That is the wrong end of the problem.

Every South African small business has a short list of places where one fact gets keyed in more than once. A customer gives their company registration number on a quote form. Someone types it again onto the invoice. It gets typed a third time onto the delivery note, and a fourth time into the accounting package at month end. Four keystrokes, one fact, four chances to be wrong, and four records that can now disagree with each other.

The useful question is not "which tool types faster". It is "where does this business hold the same fact in more than one place, and which copy is the one we would rely on".

What South African law says about the second copy

This is the part that almost never appears in coverage of this topic, and it is the part that decides how much a duplicate keystroke actually costs you.

POPIA makes accuracy an obligation, not an aspiration

Section 16(1) of the Protection of Personal Information Act 4 of 2013 states:

"A responsible party must take reasonably practicable steps to ensure that the personal information is complete, accurate, not misleading and updated where necessary."

Section 16(2) adds that in taking those steps you must have regard to the purpose for which the information was collected or further processed. Read plainly, the standard scales with what you are using the information for. A misspelled surname on a mailing list and a wrong registration number on a tax invoice are not the same risk, and the Act expects you to know the difference.

Duplicate data entry is the most ordinary way a business fails this. Nobody decides to hold two versions of a customer's details. It happens because two people typed the same thing into two systems on two different days.

Section 24 turns every duplicate into work you have to repeat

Section 24(1)(a) lets a person ask you to correct or delete personal information that is "inaccurate, irrelevant, excessive, out of date, incomplete, misleading or obtained unlawfully". Section 24(2) requires you to act as soon as reasonably practicable.

Section 24(3) is the one that matters for this topic. Where a correction changes information that has an impact on decisions taken about that person, you must, if reasonably practicable, inform every person or body you disclosed that information to.

So a correction is not one edit. It is one edit per copy, plus a round of notifications. Every place your team typed the same thing twice is a place you now have to go back to. The duplicate keystroke does not cost you thirty seconds. It costs you thirty seconds now and an unbounded amount later.

ECTA decides which copy counts as the original

The Electronic Communications and Transactions Act 25 of 2002 answers the question people assume is unanswerable. Section 14(1) says that where a law requires information to be presented or retained in its original form, a data message meets that requirement if the integrity of the information has passed the assessment in section 14(2), and the information can be displayed or produced to the person it must be shown to.

Section 14(2)(a) sets the test:

"by considering whether the information has remained complete and unaltered, except for the addition of any endorsement and any change which arises in the normal course of communication, storage and display".

Complete and unaltered. A record that a person read off one screen and re-typed into another has not remained unaltered in any meaningful sense. It is a new record that resembles the first one. Usually it matches. Occasionally it does not, and the whole point is that you cannot tell which occasion you are in without checking.

And section 15 decides how much your record is worth in a dispute

Section 15(2) says information in the form of a data message must be given due evidential weight. Section 15(3) lists what that weight is assessed against:

1. the reliability of the manner in which the data message was generated, stored or communicated;
2. the reliability of the manner in which the integrity of the data message was maintained;
3. the manner in which its originator was identified; and
4. any other relevant factor.

The first two are a direct description of your data entry process. "Somebody read it off an email and typed it into a spreadsheet" is an answer to both questions, and it is a weak one. A system that carried the value forward without a human re-keying it is a stronger one.

Section 15(4) is worth knowing as well. A data message made in the ordinary course of business, or a copy or printout certified correct by an officer of that business, is admissible on its mere production and is rebuttable proof of the facts it contains. That is a real advantage, and it rests on the record having been made in the ordinary course of business, which is a statement about your process rather than about your software.

None of this is exotic. Section 14 and section 15 have been in force since 2002, and section 16 of POPIA since 2013. They are simply absent from the way this topic is normally written about, which is why the standard advice stops at "use a tool".

The Excel answer, and where it stops working

If you search this topic, most of what you find is about Excel: macros, formulas, importing from PDF into a sheet, pushing from a sheet into a web form. That is genuinely where most people start, and for a single repeated task inside one spreadsheet it often works.

It stops working at the point where the second system appears. A macro can fill a sheet quickly. It cannot make the sheet and the invoicing system agree about what a customer's address is. The moment a fact needs to exist in two places, the problem changes from typing speed to record keeping, and a faster keystroke does not touch it.

This is the honest reason to be careful with the framing. Automating data entry inside a spreadsheet makes one person faster. Removing data entry between systems makes the business consistent. The second one is worth much more and is almost never what the tool comparisons are selling.

How to find your own duplicate keystroke points

This does not require software to diagnose. It requires an hour and a list.

1. Pick one fact, not one process. A customer's VAT number. A job reference. A delivery address. Follow that single fact through a whole month of normal work.
2. Write down every screen it is typed into. Not every screen it appears on, every screen a human types it into. The distinction is the whole exercise.
3. Count the systems, not the steps. Three keystrokes inside one system is a workflow problem. Three keystrokes across three systems is a record keeping problem, and it is the expensive kind.
4. Ask which copy you would produce. If a customer disputed that fact next year, which record would you hand over, and could you show it had remained complete and unaltered? If the answer is "whichever one I find first", you have located a real risk rather than an inefficiency.
5. Rank by how often, times how bad. A fact typed twice a day that only affects an internal label is a low priority. A fact typed twice a month that appears on a tax invoice is not.

Most businesses that do this find between four and eight of these points. They are rarely the ones people expected, and they are almost never the ones a tool demo is aimed at.

What actually changes when this gets automated

This is the part that has to be concrete, so here it is in the terms that matter rather than as a feature list.

What gets captured once. Identity and reference data: the registration number, the VAT number, the billing address, the contact person, the job or quote reference. These are captured at the first point of contact and never asked for again. The test of whether you have done this properly is simple: your team should stop asking customers for information the business already holds.

What gets reused instead of re-typed. The quote becomes the job, the job becomes the delivery note, the delivery note becomes the invoice. The same stored values carry forward. Nobody reads a number off one document and types it onto the next, which is precisely the step section 14(2) is unhappy about.

What stops being re-keyed at month end. The handover into the accounting package is usually the single largest cluster of duplicate keystrokes in a small business, and it is concentrated in the few days when everyone is busiest. That is a bad combination and it is the usual place to start.

Where the reminder comes from. Once a fact lives in one place, the follow up attaches to the record rather than to a person's memory. An unpaid invoice raises a dated task because the invoice knows it is unpaid. It does not depend on someone remembering to look.

What stays measured. Not keystrokes saved, which is easy to count and tells you very little. Measure how often two systems disagree about the same fact, and how long a correction takes to reach every place that fact is held. Those are the numbers that map onto section 16 and section 24(3), and they are the ones that get worse quietly.

If you want the wider picture of what manual admin costs a South African business before any of this is automated, we set that out in the real cost of manual admin for South African SMEs. The related question of getting more done without adding headcount is covered in how to increase team output without hiring.

What not to automate

Three things are worth leaving alone, at least at first.

A fact that is genuinely entered once. If it is captured in one place and read everywhere else, it is already solved. Automating it is work with no return.

A process nobody agrees on. If two people in the business do the same task differently and both think they are right, automating it just makes the disagreement faster and harder to see. Settle the process first. This is one of the more common reasons automation projects fail, which we go into in why AI projects fail.

A judgement call wearing a data entry costume. Deciding which discount band a customer falls into, or whether a job is complete, looks like data entry because it ends with someone typing a value. It is a decision. Automate the capture around it and leave the decision with a person, particularly where it has legal or financial consequences for the customer.

Where this connects to the rest of your paperwork

Data entry sits next to two things worth reading if this is the problem you are solving. The first is signatures: once records carry forward without re-typing, the question of whether the resulting document is validly signed comes up quickly, and the answer is in the same Act. See whether electronic signatures are legal in South Africa. The second is invoicing, where duplicate keystrokes are most concentrated and where SARS has specific requirements about what the document must contain: see tax invoice requirements and how to automate the invoicing process.

If personal information is involved, and for most of these facts it is, the POPIA compliance checklist covers the surrounding obligations.

Frequently Asked Questions

Does automating data entry mean replacing my accounting or CRM system?

No. Most duplicate keystrokes happen between systems that are each fine on their own. The usual fix is to connect them so a value captured once carries forward, rather than to replace either one. Replacement is a much larger decision with different reasons behind it.

Is a re-typed copy of a document still valid evidence in South Africa?

It can be admissible, but section 15(3) of ECTA weighs it against the reliability of how it was generated and how its integrity was maintained. A record carried forward by a system, rather than re-keyed by a person, is on stronger ground under the section 14(2) test of complete and unaltered.

How many places should one piece of information live in?

One place that is authoritative, with everything else reading from it. That is the practical reading of POPIA section 16, and it is what makes a section 24 correction request a single edit rather than a hunt through every system the fact was typed into.

Where do most South African small businesses start?

The month end handover into the accounting package, because it is the largest cluster of duplicate typing and it lands in the busiest few days. Quote to invoice is the usual second, since the same customer and job details are typically typed at least twice.

Do I need to automate everything for this to be worth doing?

No. Removing the two or three highest frequency duplicate keystrokes usually accounts for most of the benefit. The list from the exercise above is normally between four and eight points, and they are not equally valuable.

Where to start

The diagnosis above is something you can run yourself with a pen and a month of normal work. If you would rather have it done properly, with the duplicate points ranked and the record keeping risk assessed alongside them, that is what a Business Autopsy produces. You can also book a discovery call to talk through which of your systems are disagreeing with each other.

Sources

• Protection of Personal Information Act 4 of 2013, sections 16 and 24. Government Gazette No. 37067, 26 November 2013.
• Electronic Communications and Transactions Act 25 of 2002, sections 14, 15 and 16. Government Gazette No. 23708, 2 August 2002.

Statutory text in this article was read from the published Government Gazette versions of both Acts and cross checked against two further copies of each. Verified 16 August 2026. This is general information about how these provisions read, not legal advice on your specific circumstances.