← All insights
DEVELOPMENT4 MIN READ

An internal tool succeeds when it fits the people who use it. Before connecting systems or writing code, we study the work: where it starts, where it waits and where somebody has to do the same thing again.

01 / Follow the work from beginning to end

A team rarely describes its workflow in the same way as an organisation chart. Real work moves between documents, messages, spreadsheets and informal decisions. We ask people to walk through a recent example rather than describe the ideal process. What arrived first? Who made the next decision? Where did information get copied, checked or clarified?

This investigation should include the less visible work: finding an attachment, confirming which version is current and reminding someone that approval is waiting. Those small actions may not appear on a task list, but together they shape the day. We document the sequence and its exceptions so that the team can recognise the process in the description.

02 / Choose one useful place to start

A complete operating system for the company is an ambitious brief. A reliable way to prepare and approve one type of campaign is a much clearer first release. We look for a repeatable workflow with a visible outcome, people who can give feedback and a practical way to assess whether the change is useful. The first version should solve a real problem on its own.

We also distinguish between unnecessary work and necessary judgement. Copying an approved value between systems may be a good candidate for automation. Deciding whether a sensitive communication is appropriate may require a human review. Designing that boundary explicitly helps the team trust the tool. Automation should support responsibility rather than make it harder to see who is making a decision.

Automate a clear process. Do not make a confusing process run faster.

03 / Create a shared language for the process

Before building screens, we define the key objects and states. What is a campaign? How does a draft differ from an approved version? Can something be ready for review while still waiting for an asset? A small, consistent vocabulary prevents different parts of the interface from describing the same thing in incompatible ways.

The visual design should make those states easy to read. People need to know what requires attention, what is blocked and what they can do next. A dashboard is useful when it supports a decision, not when it simply contains many charts. We favour clear status, ownership and next actions over the appearance of complexity.

04 / Connect systems without hiding the edges

Integrations need clear rules about which system owns which information. We agree on the source of truth for contacts, assets, approvals and reporting before moving data between tools. This avoids a familiar problem: two interfaces display slightly different versions of a record and nobody knows which one to trust.

We design for interruption as well as success. What happens if an external service is unavailable? Can a failed action be retried safely? Is the person told whether an operation completed or is still waiting? Meaningful feedback and a visible activity history make problems easier to investigate. These details are less glamorous than a new feature, but they matter when the tool becomes part of daily work.

Make workflow status visible

  1. CompletedRecord what ran and make the output easy to find.
  2. Awaiting approvalShow who needs to approve the next step.
  3. Needs attentionShow what failed, who owns it and how to retry.
Make completed, waiting and failed actions visible, with an owner and a safe way to retry.

05 / Make ownership and control visible

A useful internal application has an understandable permission model. People should know which information they can change and which actions require approval. Sensitive operations deserve a deliberate interaction, while frequent low-risk tasks should remain straightforward. The right amount of friction depends on the consequence of the action.

We also consider handover and maintenance. Someone on the team needs to understand how templates are updated, where an integration is configured and who receives an error notification. Clear documentation and a manageable administration flow help the tool remain useful after its first launch. Operational ownership is a design requirement, not only a technical concern.

06 / Measure the workflow in practice

We establish a practical baseline before making claims about improvement. Depending on the task, that might include the time from brief to approval, the number of manual handoffs or the frequency of avoidable corrections. A small set of relevant measures is more useful than an elaborate reporting system that nobody has time to interpret.

The first release should be used with real work and a manageable group of people. We observe where they hesitate, where they leave the system and what they still need to do manually. That feedback guides the next iteration. The long-term goal is a calmer, clearer way of working: fewer duplicated actions, visible responsibility and more time for the decisions that benefit from human attention.

What to take away

The best automation feels like a natural extension of the team. It makes the work easier to understand, easier to complete and easier to improve.

Start a project ↗