Where automation makes a practical difference
Documentation and evidence. Screen records, specifications, test reports and change documentation: work that recurs with every release and eventually sets the pace. We automate recurring parts of its creation so the evidence keeps up with the application.
User-generated content and moderation. A campaign, competition or community produces a daily flow of material your brand is responsible for: photos, comments and posts. We build assessment and screening around your criteria and define which cases people handle.
We explain these two tasks through examples below. The same approach also helps with other processes:
- Data reconciliation between systems. A list keeps two systems in sync. An interface can take over this work and make transfers and missing data visible. - Approval and review workflows. Who submits what to whom, which version was reviewed and what was the result? We model the workflow and its history. - Recurring analysis and reporting. Figures from defined sources are combined, with traceable origins and calculations. - Notifications and deadlines. The right message reaches the responsible group without anyone regularly checking lists. - Migration and data maintenance. Data is transferred, checked and cleaned. The result shows what was imported and what needs further attention. - Monitoring. Failures and changes become visible and reach the people who can respond.
How we approach a process
Look at what actually happens. Many workflows have evolved over years. Some steps exist only because of an earlier system boundary.
Separate decisions from routine work. Gathering, sorting, formatting and passing information on can often be automated. For decisions, we clarify the criteria, consequences of an action and when a person intervenes.
Look for the shorter route. Sometimes a step can be removed altogether. Then it does not need automating.
Make results visible before they have consequences. New automated decision-making can initially run in shadow mode. The responsible people see how it judges before automatic actions are enabled.
Identify open questions. Assumptions and unresolved cases remain visible. This makes it easier to refine the rules together.
Document app screens automatically
An interactive application has many states. For specialist approval, these must be available in a fixed, organised format clearly linked to the relevant version.
For Takeda's ID+ patient app, we generate screen documentation automatically: several hundred screen states for iOS and Android, organised by the app's workflows. Every app version therefore has its matching review documents.
With manual documentation, often only changed screenshots are replaced. But when a colour, font size or spacing changes across the app, all screenshots may need updating.
After app changes, we regenerate the screen documentation. The documents do not need updating manually, screen by screen. The client handles specialist review and approval.
There is no measured comparison here against fully manual creation. The benefit lies in repeatability and avoiding extensive manual updates to screen documentation after changes.
Assess images against your own criteria
Around half a million people are registered in AIDA Cruises' guest community. They upload photos of their trips. Some of this material is suitable for highlighting in the community or using for marketing.
Our image assessment considers criteria for this selection. Around one in seven images is presented to the editorial team for a decision. Aesthetic criteria play a part; submission does not automatically imply a suspected rule violation.
The editorial team confirms nineteen out of twenty submitted images unchanged. This statement applies to the submitted selection. It does not describe measured quality across all other assessments.
Assessment runs on our own hardware and can now consider more complex subjects than the former cloud service, which supplied keywords. We can refine the criteria and control where images are processed.
Controlled automatic actions
A model provides an assessment. The application uses its rules to determine the resulting action. This separation is crucial in AI-assisted automation.
Our moderation platform Modary considers conversational context and logs its decisions. It can initially run in shadow mode. Once enabled, it performs permitted actions automatically; human corrections take priority. Modary is currently in testing.
Work on these applications is a second level of our AI use. In our own development, AI also supports analysis, implementation and documentation. Results are checked and enter the agreed approval process.
Less manual work on recurring tasks
Documentation keeps pace with new versions. Content can be assessed against defined criteria. Recurring data work needs fewer manual handovers. People can concentrate on exceptions and improving the process.
The first step
Describe the process and what you would like to change. We will explain the approach we see, what can be automated and which decisions need separate clarification.