Learn by doing
How to stop entering the same data twice
When the same details are typed into several places, start by asking why each copy exists. A shared record or an existing tool’s feature may solve the problem before a new integration is needed.
By Max O’Sullivan · Guide and worksheet
Find the copy that creates extra work
Choose one recurring handover: a request moving into a job list, an agreed job becoming a task, or a weekly update being assembled from several places. Watch the steps and note which details someone types or pastes again.
For each copied item, ask whether the next person actually needs it. If the second list is only a view of work already held elsewhere, a shared view or link may be enough. Removing an unnecessary copy can be simpler than automating it.
Separate three different problems: copying a fact into another tool, recording the same item twice in one tool, and finding conflicting versions. They can appear together, but each needs a clear rule.
Agree which record is the one to trust
Decide where each important fact should be updated and who is responsible for it. The same place doesn’t have to own everything: a job list may own the agreed start date, while a different tool owns an approved invoice number.
Explain what should happen if the information changes after it has been copied. Should the receiving record update, wait for review or stay as a dated snapshot? Don’t leave two people editing the same detail in different places without an agreement.
Also decide how to recognise the same item. A job or request reference can be more useful than a customer name, because one customer can have several jobs. Check that the reference you choose is present and unique for the records involved.
A worked example: from a request to a job list
Fictional training example — not a client project.
Imagine a small team that keeps approved requests in one tool and copies them into a shared job list. Use a made-up request, reference R-42, for this exercise.
- What starts the handover?
- A person marks the request as approved. A new message arriving alone is not approval to create a job.
- What needs copying?
- The request reference, agreed work and assigned person may be enough. A link back to the request can avoid copying the full conversation into another place.
- What if it arrives again?
- The job list should recognise reference R-42 and follow the agreed update rule. A repeated input must not silently create an unwanted second job.
- What if information is missing?
- A request without an assigned person goes to the owner for review. It doesn’t guess a name or disappear from view.
These rules describe the useful behaviour. Whether an existing feature, a connection or a small custom tool can provide it depends on the tools and access you actually have.
Exercise: map one data handover
Copy the worksheet into your notes. Use made-up records while working out the rules, then ask the people doing the work whether the map matches the real process.
Your data-handover map
Copy these prompts into your notes and fill in one example.
- Trigger and destination
What starts this handover, and where does the information need to go?
- Information actually needed
Which fields must move, and which can be removed or linked instead?
- Source and owner
Where is each fact updated, and who is responsible for it?
- Matching rule
How will you recognise the same record when it arrives again?
- Changes and exceptions
What happens after an edit, missing data or a failed transfer?
- A useful check
Who compares the result with the source and checks whether the whole task is easier?
Mark each field as needed, not needed or unclear. Remove the fields that don’t need copying. Resolve the unclear ones with the responsible person before building a connection.
Choose the smallest change and check it
First check whether your current tools already offer a shared view, an export/import or a supported connection. Check what information it carries, who can access it and whether its limits fit the handover. A new subscription is not the starting point.
For a consistent set of fields, ordinary rules may be enough. If the task involves interpreting varied messages, use the AI-versus-automation guide to consider where review is needed. AI shouldn’t invent missing values or decide which conflicting record is correct.
Before relying on a change, test a normal record, missing information, a repeat, a later edit and a failed transfer in a separate test destination. The automation testing guide explains how to keep a test record and an exception route.
Compare the complete task before and after, including checking and correcting the result. Less typing is useful only if the work as a whole becomes easier. IBM’s introduction to data integration gives background on bringing information from different systems together; it doesn’t choose a tool for your business.
If the steps and responsibilities aren’t agreed yet, document the process before deciding how to move its data.