Agentic AI · Retail operations · Concept project

Flo - A Retail Floor App Powered by an Agentic AI

Flo is a retail floor app designed around the realities of a store associate’s shift. Its AI assistant, Flora, helps associates understand priorities, continue interrupted work, report issues naturally, and stay confident without adding more screen time to an already busy day.

People in the journey

The experience was shaped around three roles

Flo is not simply a task-list product. It connects the person doing the work, the person supervising the shift, and the AI agent supporting both.

Retail associate

Alex

Alex knows the store deeply, but works with full hands, constant interruptions, limited attention, and little patience for tools that slow him down.

AI assistant

Flora

Flora understands task urgency, listens to natural speech, classifies issues, routes reports, and helps Alex continue his work without needing to manage the system.

Supervisor

Tiara

Tiara assigns work, monitors progress, responds to issues, and needs confidence that the floor is being handled without following up on every small task.

The morning that started this

It is 9:02 AM. Alex has just clocked in.

The store is already busy. Alex may be carrying a scanner, handling a trolley, answering a customer, or walking between the stockroom and the floor. His phone is usually in his pocket or apron-not held carefully in both hands.

He does not dislike technology. He dislikes technology that assumes he has time, attention, and two free hands. He needs something that works within his shift rather than asking the shift to adapt around the app.

Illustrated sequence showing Alex during a busy retail shift — Alex moves through physical work, customer interruptions and shifting priorities throughout the day.

The actual problem

This is Alex’s problem-not a product problem looking for a solution

Alex is not bad at his job. The problem is that most tools assume a version of his day that does not exist: someone sitting down with full attention available for a screen.

His hands and attention are often occupied

Anything that requires two hands, prolonged focus or several taps is likely to be skipped, delayed or handled outside the system.

The task list does not explain priority

Alex may know what needs to be done, but not what should happen first or why one task matters more than another at that moment.

Interruptions erase working context

After helping a customer or handling an unexpected issue, Alex can return to a task and genuinely lose his place-not because he is careless, but because the interruption displaced the context.

Three illustrated retail floor problem scenarios

Why Jobs to Be Done

Alex is not trying to use an app

He is trying to complete a shift feeling capable and in control. A feature-led approach would produce a better checklist. A job-led approach asks what Alex is actually trying to achieve through that checklist.

Functional job

Help me know what to do next

When I start my shift, I want to know exactly what to do and in what order, without repeatedly asking Tiara or rereading a long list.

Emotional job

Help me recover without feeling behind

When something unexpected happens, I want to handle it and continue without feeling that I have failed or made the shift harder.

Social job

Help others trust that I handled it

When my shift ends, I want Tiara to see that the work was completed properly without needing to follow up with me repeatedly.

Jobs to Be Done visual for the retail assistant experience

Core experience

From task priority to issue resolution

The primary flow keeps Alex within one continuous work context. Flora handles the transition from recognising an issue to logging and routing it while the task screen remains visible behind the interaction.

Complete Flo retail task and reporting flow — Home, active task, reporting, AI interpretation, confirmation and return-to-task states.

Agent behaviour

Flora acts independently-but not invisibly

What Flora can do alone

Flora can classify reports, create logs, identify the right team and notify the relevant person when those actions only affect the system.

When Flora asks first

Flora asks for confirmation when an action affects Alex’s task, schedule or time-for example, sending a report, pausing a task or rescheduling work.

What Flora never does

She does not ask Alex to categorise his own issue, repeat confirmed information or answer open-ended system questions during a busy task.

Language and tone

Calm, multilingual and sensitive to pressure

Alex naturally mixes Hindi and English. Flora is designed to understand the language and tone he uses rather than requiring formal commands or perfect sentences.

Context

Flora’s response

Alex is on time

Flora gives enough context to support the task and explains where to go.

Alex is running behind

Flora removes unnecessary detail, becomes more direct and makes urgency visually clear.

Alex sounds frustrated

Flora remains brief and steady instead of matching his stress or asking more questions.

Error handling

Nothing should fail silently

Error recovery was designed to preserve Alex’s momentum. He should not need to restart a report or wonder whether something disappeared.

Flo could not hear voice input and opens typing — If the microphone fails, Flo explains what happened and opens typing automatically.

Flo report network failure with retry and save for later — If sending fails, Alex can retry or save the report for later. The system never silently loses his input.

Conclusion

From task management to a supportive floor companion

Flo shifts retail task management away from a long checklist and toward a context-aware assistant that helps associates stay clear, confident and in flow.

Flora does not replace Alex’s knowledge of the store. She supports it by reducing uncertainty, preserving context, handling administrative work and helping the right information reach the right person.