Skip to content
All apps

Delivery Race

About this app

Classical delivery against agent delivery on two lanes: implementation time collapses to minutes, the understanding block stays equally long on both lanes — and decides the quality.

The idea

AI agents cut programming time dramatically. Whether a tool ends up fitting, though, depends on someone having understood the process it is meant to support. The app shows this as a race. Classic delivery runs through understanding, specifying, tendering and waiting, building, testing and release. Agent delivery runs through understanding, describing, “agent builds”, reviewing and shipping. The model’s claim: build time collapses, time to understand does not. Durations are illustrative and meant to show orders of magnitude.

What you see

Two lanes in 3D, each made of coloured blocks for understanding, waiting, building and testing, and reviewing. Block length is duration on a logarithmic time scale, so 30 minutes stay visible next to several months. Four scenarios are available: bookkeeping, quotations, a feature and a core system. You can also set the level of process understanding and the tool’s complexity, from a small tool to a department solution to a certified core. A clock shows how much faster the agent lane is and how well the result fits.

Try it

Start the race with high process understanding: both lanes deliver a fitting tool, the agent lane just much sooner. Then lower process understanding. The understanding block gets shorter, but fit drops on both lanes – building the wrong thing faster gains nothing. Finally pick “certified core”: here the agent lane stays long too, because review and certification take time. Notice that the understanding block is always the same length on both lanes.

Frequently asked questions

How do AI agents change software development in companies?
Agents take over much of the coding and testing, so internal tools can be built in hours instead of weeks. Tendering and waiting often disappear for small tools. What remains is understanding the process, describing the requirements and reviewing the result.
Why can’t process understanding be shortcut?
To build a fitting tool, someone has to know how the work really happens, which exceptions occur and what the goal is. That knowledge comes from conversations, observation and experience. An agent builds exactly what it is told – without understanding, it quickly builds the wrong thing.
When do AI agents save little time?
For systems that must be reviewed or certified, such as core systems under regulatory requirements. There, review, acceptance and certification dominate the total duration. Faster building only shortens a small part of the path.
Why does the app use a logarithmic time axis?
The phases range from minutes to months. On a linear axis the short blocks would vanish. On a logarithmic scale each power of ten takes the same width, so both lanes stay comparable.

Subject: AI transformation in business

More from AI transformation in business

A VisuApp by heyprof: interactive, free in your browser, no sign-up. What is a VisuApp? · All apps