A commercial team can create three versions of one customer address before lunch, given enough forms. The formatting changed slightly each time. By the end, the organisation had three versions of one address and a lively debate about which floor the customer occupied.

Duplicate data entry is usually described as inefficient. That undersells it. Re-entry creates multiple versions of truth, delays handoffs and invites mistakes precisely where users assume the information is already known. It is a quiet tax charged on every transaction.

The interface may make each form easy. The workflow remains absurd.

Repetition signals a broken boundary

When users copy data between systems, the problem is not always a missing integration. It may be unclear ownership. Which system owns the customer record? Which stores contract terms? Which can amend an invoicing address after approval?

Before connecting systems, map the objects and authorities. Otherwise integration merely moves disagreement faster. A field flowing automatically from CRM to finance is helpful only if everybody agrees that CRM contains the right value at the relevant moment.

I use a simple model: source, consumer, transformation and feedback. The source owns the value. Consumers use it. Transformations adapt it. Feedback corrects errors at the appropriate source. This prevents every system from becoming an editable copy of every other system.

Copying is hidden decision-making

Re-entry looks mechanical, but users often make choices while copying. They shorten a company name, select a billing contact, convert a date or omit a note that seems irrelevant. Those choices may be sensible and invisible.

Automation must discover them before removing the manual step. Observe what people change and why. Some variation should become a rule. Some should remain an explicit decision. Some is accidental inconsistency the product can eliminate.

In the Aerios platform work, integrations with CRM, finance and cargo systems were strategically important because charter data crosses commercial, operational and financial boundaries. The opportunity was not simply fewer keystrokes. It was a consistent record moving through the lifecycle with deliberate ownership.

The cost appears downstream

The time spent typing is easy to see. The larger cost arrives later: mismatched references, rejected invoices, duplicate customers, incorrect taxes and support investigations.

Measure correction and reconciliation, not only entry time. Ask how often teams compare systems, which fields generate disputes and how many exceptions trace back to stale or transformed data. A five-minute copy task can create hours of downstream archaeology.

This is why data quality is part of user experience. Users lose trust when the product displays two answers to a question that should have one. They then build verification rituals, usually involving spreadsheets, which further multiply the answers.

Integrate around events

Point-to-point field syncing can become fragile when systems update one another continuously. Event-based thinking is often clearer: customer confirmed, quote accepted, contract signed, invoice issued.

Each event defines what data is stable enough to transfer and what should happen next. It also makes failures visible. If contract creation fails after acceptance, the workflow can show that specific gap rather than leaving users to wonder why the finance system feels unusually quiet.

The technical implementation varies, but the product experience needs consistent signals: pending transfer, successful transfer, failed transfer and retry. Do not display data as final while the relevant integration is still negotiating with reality.

Preserve provenance

Users need to know where important data came from and when it was last updated. Provenance is especially valuable when information cannot be edited in the current system.

“Customer address — from Salesforce — updated yesterday” explains both authority and route to correction. A disabled field without explanation looks broken. An editable copy creates divergence.

Provenance also supports audit and compliance. It should be exposed proportionately; not every screen needs a family tree for each value. Consequential documents and exceptions deserve more detail than routine list views.

Design the correction loop

Integrations fail partly because correction is treated as somebody else’s problem. If a user spots an incorrect address in a contract, where should they fix it? Will the change propagate? Does the document need regeneration? Are previous versions retained?

A good correction loop points to the source, explains the impact and confirms propagation. If automatic updates would create risk, require an explicit refresh or new version. Preserve historical documents rather than silently rewriting what was previously agreed.

The GOV.UK guidance on maintaining live services stresses operational responsibility and continuous improvement. Data flows need the same ownership. An integration without a correction process is a launch plan, not a service.

Do not automate uncertainty

Some duplicate entry exists because teams do not trust the upstream data. Connecting the systems without addressing trust can make the downstream product faster and less correct.

Show validation, freshness and completeness. Allow review at consequential transitions. If confidence is low, present the imported value as a suggestion rather than fact. Automation should reduce unnecessary decisions, not conceal unresolved ones.

AI extraction can help with emails and documents, but extracted values need source links, confidence and correction. A plausible company name is not the same as the contracted legal entity. The interface must respect the distinction even when the model sounds very sure of itself.

Prioritise by pain and leverage

Not every duplicate field deserves an integration programme. Start with high-volume, high-consequence information that crosses stable boundaries. Customer identity, core request data, approved commercial terms and invoice references often have more leverage than occasional free-text notes.

Estimate frequency, time, error cost and downstream reuse. A value entered once and consumed by five processes is a strong candidate. A value copied twice a year by one specialist may be better supported with a clear manual step.

Measure one source of truth

Success means more than fewer keystrokes. Look for reduced reconciliation, fewer duplicate records, faster document creation, lower rejection rates and greater confidence in displayed data.

Interview users after automation. If they still keep a parallel spreadsheet, ask what trust or flexibility it provides. The spreadsheet is evidence, not disobedience.

Remove the organisational copy-and-paste

Duplicate entry is where fragmented products expose fragmented ownership. Fixing it requires object modelling, system authority, visible transfers and robust correction—not merely a faster copy button.

Map the source. Understand the decisions hidden in manual work. Integrate around meaningful events. Preserve provenance and design failure as part of the workflow.

When users type the same address three times, the problem is not that they need better autocomplete. The organisation needs one answer and a product capable of carrying it.