Cloud communication platform for businesses

The incumbents were built for developers. I designed for the person running the business

I audited Twilio, Plivo and MessageBird. I mapped each one's user journey and ran a heuristic evaluation against it. Twilio's and Plivo's builders are powerful, but they're made for engineers reading API docs, not for someone trying to route a support line. Both had the same three gaps.

• ChannelsLimited flexibility for different channels.
• BuilderA visual builder that's hard to use.
• NavigationComplex navigation to find basic features.

So every later decision got one test. Could a business owner do this alone?

One agreed map, with the builder as a tool every channel shares

Airtel IQ has three channels, two use cases and a set of tools they all share. Agreeing one structure for all of it took many rounds with Airtel's team.

The call-flow builder is the clearest example. I argued it should be a shared tool with its own place, not a page inside one channel, because a flow can run across voice, SMS and WhatsApp. Airtel's team took some convincing. Feedback from testers settled it.

I kept the trial to one simple path, to find out if anyone wanted this

The trial is self-serve. Register, and Airtel IQ gives you a working test number, an "intelliphone", before you commit to anything. That part wasn't my decision. Keeping it to one linear path was. The trial had a question to answer before the builder got built. Did the market want call-flow templates and a no-code builder? One simple path was the quickest way to find out.

So nobody starts from a blank canvas. You activate the number by calling it, pick a ready-made template with a live preview, and edit the flow: reroute a branch, swap the audio, add or remove a step. Testing gives you a real number to dial, so you hear your own configuration as an incoming call. Then the call shows up in your logs, with the same metrics a paid account gets.

More than 1,000 businesses signed up. That started the next phase, the full platform.

Two card layouts lost to a tree that arranges itself

Two directions came first. A few rounds of client and user feedback ruled both out.

Sectioned cards laid the call out in named parts: greeting, menu, other inputs. That worked for a plain IVR and nothing more. Adding an agent, sending an SMS or jumping between flows didn't fit. Neither did a flow that goes back and forth between menus.

Big cards held a whole step each, with a connector leaving the card for every outcome. They got messy to build and maintain, worst of all for IVR menus with nine or ten options.

The shipped builder is a node tree because it handles both. A simple flow stays simple and a complex one still fits. Nodes arrange themselves, the canvas follows the node you click, and zooming out is optional.