AI Payment lifecycle

An AI-assisted analytics page for investigating acceptance and decline reasons across the payment lifecycle

Designed a payment routing admin panel that helps operators manage rules,
monitor traffic, and reduce manual errors
in high-volume payment flows.

Project overview

Checkout is a global payment platform that processes transactions across many providers and regions. When routing moved to the automation layer, operators lost visibility into why acceptance rate went up or down. The existing "Intelligent Acceptance" page showed how much revenue the platform recovered, but not what was happening inside the lifecycle when something failed.

Problem

The old Intelligent Acceptance page answered one question — how much revenue the platform recovered. It didn't help operators understand where payments were failing or why. To investigate a drop in acceptance, they had to export raw data, match BIN, bank, and country by hand, and look up decline codes in a separate document. Everything happened outside the page.

Goals

Give operators visibility into each lifecycle stage

Reduce dependency on the data team for routine checks

Grow AI assist adoption in daily analytics work

Cut time to find the cause of a drop by 40%.

Reduce ad-hoc data requests from ops by 30%

Reach 60% AI assist adoption among analysts.

Make analytics a regular tool, not an occasional one

Constraints

Design system

The internal library gave speed and consistency, but limited visual options for dense analytics like compact charts

AI trust boundary

The AI assistant answers questions in context, but cannot change routing or apply fixes. Any pattern that mixed those roles was avoided.

Domain fidelity

Lifecycle stages and decline codes had to match how the platform actually processes payments.

Key insights

Investigation

Operators check acceptance drops by looking at bank, BIN, and payment method together, not one at a time.

Operators could only expand rules one at a time, never see the full picture at once.

Decline codes need translation

There are 150+ codes, and most operators only remember 10–15. Raw codes without context are noise.

AI works in context.

Operators trust AI answers next to the chart. A separate chat was ignored in testing.

Key desicion

Three shifts moved the page from a summary to a working tool — a per-stage view of the lifecycle, an AI assistant tied to each chart, and drill-downs on BIN, bank, country, and payment method.

Operators can ask the AI assistant for a chart in natural language and get it built without opening the builder.

Operators can open any chart in the builder with its context pre-applied, dig into raw data, and export or save the result.

Operators can ask the AI assistant a question about the chart they're looking at and get an answer scoped to its data.

Final interface

The final page turns the payment lifecycle into a working analytics tool — each stage becomes a chart you can drill into, and an AI assistant answers questions next to the chart you're already looking at.

Impact

The page turned the lifecycle into a readable path — operators find causes by drilling from stage to segment to code.

Works