The problem
A “journey” in aiRA is a canvas — a drag-and-drop flow a customer walks through: an entry point, waits, multi-channel engagement over WhatsApp and push, offers, milestone and task checks, conditional points, and decision splits that branch and later merge. It’s the most involved thing you can configure on the platform, and it was entirely manual.
Building one by hand took two to three hours, and a team could produce maybe fifteen to twenty a month. For an enterprise that wants hundreds of personalized journeys, that math simply doesn’t work.
The constraints
- The builder is genuinely complex. Many block types, real branching, multi-channel side effects. Automating it isn’t filling in a template — it’s orchestrating a stateful flow.
- A launch is unforgiving. It happens on the customer’s calendar, on their site, with their revenue attached to the outcome.
- It had to outlive the launch. Anything that only I could operate would just relocate the bottleneck from the customer to me.
What I built
I automated the entire canvas flow from scratch and orchestrated it end to end — turning a described journey into a built one across all the block types, channels, and branching. Two interns worked with me on the block-authoring capability while I owned the orchestration.
Then the real test. I embedded on-site with the launch customer — one of India’s largest wires, cables and FMEG manufacturers — for four to five days, and we built 150+ personalized journeys in about three days: multi-store, multi-product, region-based variations, four-stage flows, personalized WhatsApp and push. Against a manual baseline of fifteen to twenty a month, that’s a difference in kind, not degree.
The decision that mattered
The launch was the visible win. The decision I actually care about is what came after: I productionized the module so the customer self-serves.
It would have been easier — and honestly more flattering — to remain the person who runs the journeys. Instead I built the capabilities that made me unnecessary: in-product AI image generation and image upload, custom personalized HTML templates with mobile and web preview, customized messages, external integrations inside journeys, and dynamic audiences (upload, DSL, or SQL). The point of the engagement was never to ship 150 journeys. It was to ship the ability to ship journeys.
The outcome
The launch journeys drove roughly a 40% revenue lift for the customer. And the module went from something the team operated to something the customer operates themselves — the manual configuration bottleneck didn’t get widened, it got removed.
What I’d change
Being embedded with a customer through a launch taught me more than any amount of internal building. You feel every rough edge in real time, with someone’s business on the line. If I did it again I’d instrument the self-serve paths more heavily from day one, so the signal about what customers actually reach for shapes the roadmap faster.
The principle I took away: the most valuable thing you can hand a customer isn’t the output — it’s the capability. And you should measure your work partly by how quickly you can leave.