tuAliado

• 

• 

• 

• 

• 

• 

• 

• 

Inspiration

Tuali already helps small shop owners buy products, access promotions, earn loyalty points, and use suggested orders. But for many store owners, especially those with low digital confidence, the hard part is not only having these tools available. The hard part is knowing what to do next.

We designed tuAliado for Raul, a shop owner who depends heavily on a promoter and avoids complex digital workflows. If the experience works for Raul, it can work for more digitally confident users too.

Our guiding idea was simple:

Tuali helps you buy. tuAliado helps you grow.

What it does

tuAliado is a mobile-first growth companion that lives inside the Tuali experience. It helps a shop owner choose one clear business goal, understand their current situation, receive concrete recommendations, act on one recommendation, and track progress over time.

The flow is:

Diagnosis -> Goal -> Recommendations -> Action -> Follow-up

The demo focuses on one understandable journey: Raul wants to sell more. tuAliado uses his current profile to show his ticket average, ordering channel, loyalty points, and opportunities. Then it recommends up to three actions connected to that goal, such as using an active promotion, ordering through the app, or activating a loyalty challenge.

There is also a floating chat with text and voice support. The chat helps Raul understand the recommendations in simple language, but it does not make the business decisions. The recommendations come from a deterministic engine, and the LLM only explains them.

How we built it

We built tuAliado as a functional mobile prototype with Next.js, TypeScript, and Tailwind CSS.

The product is intentionally structured around data consistency:

• A deterministic recommendation engine calculates the diagnosis and recommendations.
• Mock data is typed and labeled by origin: TUALI, CLIENTE, or ESTIMACION.
• Gemini is used only as an explanation layer, translating already-calculated recommendations into natural language.
• If the Gemini API fails, the core flow still works because the business logic does not depend on generated text.
• The voice mode uses the Web Speech API for speech-to-text and text-to-speech support.

The current demo uses mock TypeScript data based on the Tuali challenge context and protopersonas. It does not use real customer data.

Challenges we ran into

The main challenge was avoiding the trap of building a generic chatbot. A chatbot can answer questions, but Raul needs guidance, not another blank input box.

We also had to keep the experience simple enough for a low-digital-literacy user:

• one main idea per screen,
• large mobile touch targets,
• short visual choices instead of long forms,
• icons, numbers, and progress indicators instead of dense text,
• voice support for users who prefer speaking over typing.

Another challenge was data coherence. Tuali was explicit that incoherent data would hurt the solution. Because of that, we separated business decisions from LLM explanation and avoided inventing metrics that the mock data could not support.

Accomplishments that we're proud of

We are proud that tuAliado is not just a concept screen. It is a working end-to-end prototype:

• onboarding with clear business goals,
• diagnosis from the user's Tuali behavior,
• goal-based recommendations,
• daily check-in,
• progress tracking,
• chat and voice assistance,
• deterministic logic that still works if the LLM is unavailable.

We are also proud of the product direction: tuAliado helps Tuali increase ticket average and customer autonomy while giving shop owners a more accessible way to grow their business.

What we learned

We learned that AI is more useful here as a translation layer than as the decision-maker. For this use case, the most important thing is trust: the user and Tuali need recommendations that are grounded in consistent data.

We also learned that accessibility for low-digital-literacy users is not only about visual polish. It changes the product architecture: the flow must guide the user, the data must be explainable, and the interface must reduce decisions instead of adding more.

What's next for tuAliado

The next step is connecting tuAliado to real Tuali data and piloting it with a small group of shop owners.

Future improvements include:

• real-time personalization from production behavior,
• stronger integration with Tuali promotions and loyalty programs,
• adaptive recommendation ranking based on what the user actually completes,
• seasonal and inventory-aware suggestions,
• analytics for Tuali and Arca Continental to understand which actions help customers grow.

Built With

• gemini
• next.js
• react
• tailwind
• typescript
• vercel
• webspeechapi