DollarDoggo

How Milo Feels
Buttons
Budgeting
Views

Inspiration

The idea for DollarDoggo started with a simple frustration: most financial literacy tools feel like punishment instead of help. Budgeting apps often reduce your life to red numbers, guilt-heavy charts, and notifications telling you what you did “wrong.” This is one of the reasons that around 85% of budgeting apps are abandoned within three months.

When we stepped back and thought about how habits actually form, one question kept coming up: What if checking your finances felt less like being judged, and more like taking care of something you care about?

That question led us to games like Tamagotchi. They work because they create emotional attachment. You don’t just stop checking on a virtual pet overnight, especially when its well-being depends on your actions. We realized that the same mechanic could apply to money.

DollarDoggo came out of that insight: combining emotional gamification with real financial behavior so building good money habits feels personal, empathetic, and even a little fun. Instead of shaming users for mistakes, the product encourages consistency and care.

What We Learned

Building DollarDoggo taught us lessons across product design, AI systems, and human psychology.

It taught us that motion beats information. Most people avoid finance because money creates stress, anxiety, and shame. Chatbots that forget everything between sessions break trust quickly. Memory and continuity are critical for long-term engagement.

We also learned that small wins build habits. A daily check-in that takes under a minute is more effective than asking users to build a “perfect” budget.

AI needs boundaries. Without structured reasoning and context checks, even advanced models can give advice that feels poorly timed or out of touch. The pet should not punish users for being broke. It should only react negatively to disengagement.

We also learned how research like MemGPT, LaMP, and Generative Agents can be applied as real product architecture rather than staying theoretical.

How We Built It

DollarDoggo is built as a stateful, agentic AI system rather than a single prompt-driven chatbot.

Core Architecture

At the center of the app is Milo, your financial dog, powered by a multi-agent pipeline where each agent has a focused responsibility.

RouterAgent classifies user intent and routes requests.

PetAgent handles conversation flow, empathy, memory access, and high-level reasoning.

BudgetAgent generates budgets, analyzes transactions, and produces recommendations.

InsightsAgent performs time-series analysis and detects spending and balance trends.

EducationAgent delivers financial education only when it is contextually relevant.

WishlistAgent evaluates affordability and guides users through larger purchases.

This separation keeps the system interpretable, debuggable, and easy to extend.

]In budgeting mode, Milo’s mood also reflects budget utilization, engagement streaks, and spending trends. The goal is for the pet to act as a live emotional mirror of financial health without overreacting to short-term noise.

Memory & Personalization

Inspired by MemGPT, we implemented a three-tier memory system. Working Memory stores recent conversational context. Summary Memory then stores compressed long-term facts, preferences, and goals. Finally, Recall Storage keeps full searchable conversation history.

This allows Milo to remember names, goals, spending patterns, and emotional context across sessions instead of resetting every time.

We also applied LaMP-style personalization to reduce friction during cold starts by prioritizing stable user facts early and deepening personalization gradually.