I turned a narrow online top-up brief into Fluxy: an autonomous commuting agent that protects the passenger’s goal, prepares decisions and intervenes only when circumstances change.
The mismatch
Commuters still have to remember their balance, compare routes, monitor disruption and recover when the network changes.
The product question was not how to make one task faster, but which decisions software could responsibly take off the passenger’s plate.
The business bet
For an operator, the value is operational: prevent low-balance failures, reduce avoidable support contacts and protect journey completion when the network changes.
From route planner to agent
Fluxy works in the background until the passenger’s objective is threatened. It observes context, evaluates options and prepares a recommendation.
The interface appears at the decision boundary — not before.
Arrive at work by 08:45
Balance, disruption and timing change
Fluxy evaluates and prepares options
Passenger confirms consequential actions
Interactive model
Start with a passenger goal, introduce a change in context and see how Fluxy prepares a recommendation while keeping the final decision explicit.
Designing responsible autonomy
An agent can only reduce effort if people understand what it is doing, why it intervenes and where their control begins.
Do not ask for attention when the goal remains achievable.
Every intervention reveals what changed.
Consequential choices require consent.
Optimise for the passenger’s objective.
Product leadership & AI process
I translated fare policy, accessibility, operations and passenger trust into explicit autonomy rules, then used AI to simulate edge cases and generate auditable decision traces.
The model accelerated scenario coverage; it did not define the boundaries.
Proving the model
Daily brief
A single morning view communicates the plan, confidence and anything worth knowing — without turning commuting into another dashboard.
One glance replaces four checks — balance, weather, timing and route confidence, before the passenger leaves home.
Smart recharge
Fluxy monitors balance and recommends a recharge only when necessary. It prepares the action; the passenger authorises the transaction.
Fluxy stages the top-up and shows why — but the payment button stays with the passenger. This is the autonomy boundary made visible.
Goal
When disruption threatens arrival, Fluxy explains what changed and recommends a new route in terms of the passenger’s priority.
The route can change without changing the passenger’s goal — the recommendation is explained in terms of what still matters.
Good design isn’t about adding more interactions. It’s about removing the unnecessary ones.