Products

Ask Aforo

All

123

Live

Active

Draft

Product Name

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Microsoft OneDrive

P-SKU-367238

Manual

Apigee

Kong

Apigee

Kong

API

06 Jan, 2025 08:58 IST

Draft

Staged

Priced

Live

Priced

Live

06 Jan, 2025 08:58 IST

Mira

Mira

Mira

Mira

Vishal

Vishal

Vishal

Vishal

Vishal

Create a custom plan for Acme: $5,000 annually, 10M API calls included, then usage-based pricing. Keep our existing plans unchanged.

What you said

$5,000 annually

10,000,000 API calls included

then usage-based pricing

for Acme

What it becomes

Base fee $5,000 on a rate plan billed Annual

Allowance on your existing API Call unit, pooled

Overage $0.002/call, same unit, in arears

Private offering, Acme only — off your storefront

Before I build it, checking what else this touches.

Ask anything

Send

Madhu

I thought I was designing a product. Turns out, I was designing an AI agent’s Reasoning System.

PRODUCT DESIGN

Role

Product Designer

Team

Design, Frontend, Backend

Design, Development

Duration

2 months

Status

Ongoing

Context

Aforo is a digital product monetization platform that turns complex Digtial product usage (APIs, Ai agents, LLMs) usage into custom pricing and billing.

AI made it possible to build the product faster than ever. But as the codebase expanded to 50+ modules, understanding how everything fit together became the bigger challenge. I was brought in to fix the flows that were already moving and found that the product had no structure underneath it for anyone to reason from.

Designing at the Systems Layer

I mapped 75+ scattered modules into 6–7 categories,

Built meaningful relationships between product modules,

Did rapid prototyping with claude,

And used that structure as the reasoning substrate for Aforo's AI agent. So the agent could ask one sharp question instead of guessing, and nothing it does on someone's behalf happens where they can't see it.

3 Core Units:
From Information Architecture to Agentic Interaction

The Decision

I started simplifying the product instead of patching screens

Mapped explicit dependencies between modules

Mapped explicit dependencies between

Established impact tracing before agent execution

What it Changed

50+ scattered modules → 6 categories with clean sub-categories

Agent decomposes prompts into ordered tasks instead of guessing

Agent surfaces downstream consequences instead of acting blindly

Skills Explored

Systems thinking

Agentic substrate design

Agentic UX - In progress

High level Outcome

Clear mental model for team and users

Legible to the machine

Machine actions legible to the user

When Development Velocity Changed the Design Workflow

The product moved to an AI-assisted development workflow, dramatically increasing implementation velocity. Flows that previously took days could now be explored and built within hours.

That changed the design problem itself. I wasn’t introducing AI into the team, I was designing within a workflow where AI-assisted development was already accelerating how quickly the product could evolve.

The challenge became maintaining product coherence, establishing reusable systems, and giving AI enough product context to make useful design and implementation decisions.

01 I stopped patching screens and simplified the product, outside the scope I'd been given

Claude had Arrived & Development suddenly became fast

To maximize shipping speed, leadership migrated the frontend to a rapid, AI-assisted development workflow. But during that velocity jump, our design foundations got left behind. A token library already existed in Figma and the code, it just had never been structured as working context for Claude.

Leadership

We finally have momentum. I want to make sure the product keeps
moving without design becoming a bottleneck.

We finally have momentum. I want to make sure the product keeps moving without design becoming a bottleneck.

Got it. I’ll go through the product and look for where design can help.

So I opened the product. And immediately… it felt off.
Two problems stood out fast.

No Design system Applied

The product didn’t follow any design system, a real token library existed in Figma and the old frontend, it just hadn’t carried over into the Claude-generated codebase.

No Product Design Strategy

Without a real product architecture, unrelated modules looked identical and the same UI pattern was copy-pasted across three unrelated problems.

No Product Design Strategy

Without a real product architecture, unrelated modules looked identical and the same UI pattern was copy-pasted across three unrelated problems.

We tackled the surface first, reconnecting existing Figma components, tokens, and styles to the build. But consistent buttons didn't solve the deeper issue: what were we actually building?

Leadership

Give us design inputs in the product creating screen.

Leadership

Big design changes mean more development. That slows us down.

Keep it simple, nothing that breaks what already works.

A fair call. Changing structure means changing code, and velocity was the only thing we had. But every fix I made uncovered two more questions underneath it.

A fair call.

Changing structure means changing code, and velocity was the only thing we had. But every fix I made uncovered two more questions underneath it.

…Maybe micro design inputs isn’t the answer. If I keep solving one screen at a time, we’ll just keep creating more decisions to solve.

What if I step back and understand the system first?

People were building fast, while barely understanding how

the pieces connected. I was guessing. So was everyone else.

Off the table

Redesigning the backend

Starting over

If I let this continue, it would be someone’s first

impression of Aforo. Trust would break on contact,

before we’d earned any. And once that happened, no

amount of marketing was going to save it.

If I let this continue, it would be someone’s first impression of Aforo. Trust would break on contact,

before we’d earned any. And once that happened, no amount of marketing was going to save it.

I checked which modules were actually related, and grouped the ones that belonged together.

The constraint

50+

scattered modules

Home

Getting started

Products

Revenue metrics

Pricing studio

Offerings

Susbscriptions

Invoices

Dashboard

Analytics

Storefront analytics

Reports

Real time ingestion

Data Pipelines

Manual Ingestion

Customers

Partners

ERP Sync

Plugs sources

Customise storefront

Domain and routing

Authentication

Visibility

Customer community

Developer hub

T & C

Aforo discussions

Gateway orchestration

Integrations

Webhooks

Ai chatbox

Notifications

Settings

Support tickets

Trial Conversions

Alerts

Admin panel

Usage

Visit Public site

Help

Redeem invite code

Developer Hub

Billing

Embedded Studio

Embedded Studio Keys

Headless API

COGs & Margins

Platform Health

Telemetry

real categories, each with

clean sub‑categories

Define it

Products

Pricing studio

Offerings

Sell it

Selling Channels

Manage Customers

Bill it

Subscriptions

Invoices

Analyse it

Analytics

Reports

Profile, settings

Ai Help

02 I checked the dependencies between modules and sub-modules so that product has its story and AI could understand them.

I WAS Asking:

How do I fix this user journey?

I Started Asking:

Why does this journey exist?

What is this flow actually representing?

What are these things connected to?

Why are we making the same decision in three different places?

Things, this time.

Relationships.

Dependencies.

Rules.

I could finally see the product as a system.

And for the first time in weeks, I could breathe.

me, to myself

Phew.

And suddenly everyone started

connecting their own problems to the map.

Guys… look at this.

Leadership

This feels different.

Yeah. There’s finally a structure behind everything.

Leadership

The team seems much more confident now.

Because they don’t have to keep figuring out what connects to

what.

Because they don’t have to keep figuring out what connects to what.

Leadership

And this is actually something we can talk about externally.

You mean the system?

Interesting.

Leadership

Even Ask Aforo can work with this much better. The product is

clearer, so the agent has much cleaner information to reason from.

So the system isn’t just helping people understand Aforo.

Leadership

Exactly. It helps the AI understand it too.

Leadership

Good job. This is real systems design work. This approach looks

clean.

Haha, thank you.

The product is now structured so an agent could reason over it

03 From simplifying the system to making it agent-ready

This is still Work-in-progress.

Following is the interactive prototype built with Claude.

Learnings

What I learned building this

Structure is what an agent needs, not what it replaces — the design-system consolidation was the agentic-UX work, a year before you knew that's what it was for.

Structure is what an Asking everything upfront is worse than asking exactly when it matters — what building the Nova flow itself taught you, tied to the corrected sequencing we just worked through. needs, not what it replaces — the design-system consolidation was the agentic-UX work, a year before you knew that's what it was for.

A convincing demo path isn't a resolved interaction model — the honest gap between "the prototype runs great" and "the diagram exposed open questions" (mid-run stop, post-edit review) from our earlier structure conversation.