Runner-up
PLENO APP
Latin American entrepreneurs juggle 5 apps and still lose track. PLENO is the AI-native app that unifies finances, habits and projects, with agents that actually run the work for them.
Core trick
A transaction agent parses natural-language entries into amounts, currencies and installment plans, then categorizes them.
Built with
ai-agentsandroidcapacitorcloud-functionsfintecfirebasefirebase-authfirebase-hostingfirestoregeminigoogle-analyticsgoogle-play-billinggoogleantigravity2.0javascriptlemon-squeezymercado-pagomulti-agentpwareactresendweb-speech-api
Description
The problem nobody was solving
Latin America has ~99 million freelancers and independent
entrepreneurs. They run their entire professional lives across
five disconnected apps: a spreadsheet for money, one app for
tasks, another for notes, a calendar somewhere else, and
WhatsApp messages to themselves for everything that doesn't fit.
But the real problem isn't fragmentation — it's that
every
tool available was built for a different economy
. They charge
in USD in countries where inflation eats income monthly. They
assume one currency, when a freelancer here gets paid in dollars
and spends in pesos. They don't understand installment payments
("cuotas"), which is how most purchases actually work in
Argentina, Brazil and Mexico.
We researched hundreds of real Play Store reviews from Spanish-
speaking users of Notion, Evernote, Todoist, Trello, Asana and
Monday. The pattern was consistent: apps that paralyze users
with blank-page syndrome, mobile versions that break, aggressive
paywalls that remove features people already had, and pricing
models that require minimum seat counts that no solo entrepreneur
can justify.
What we built
PLENO
unifies finances, habits, projects, tasks, focus
sessions and notes into one app — but the unification isn't the
product. The AI is.
PLENO runs on a multi-agent architecture powered by Gemini,
where each agent handles a distinct operational domain:
TransactionAgent
— parses and categorizes financial
entries from natural language
InsightAgent
— proactively surfaces spending patterns and
behavioral alerts before the user asks
ProjectionAgent
— forecasts cash flow, calculates the
user's "free zone" (what they can spend guilt-free), and
simulates purchase decisions
NaturalLanguageAgent
— executes app actions from voice
commands
NotesAgent
— semantic search and synthesis across the
user's entire note history
ProfileAgent
— generates a monthly financial personality
profile from real behavioral data
The centerpiece is the
PLENO Sphere
: a single button where
tapping opens conversational AI, and holding it activates voice
commands that execute real actions — "I spent 2800 at the
supermarket" creates the transaction, categorized, in the right
account, without opening a single form.
From prototype to real users
PLENO went through a closed testing phase on Google Play with
real users — recruited from our own network, using the app
daily for two full weeks. That process surfaced real bugs (an
Android permission issue, incorrect date handling across
finance and task records) that we fixed and shipped as new
versions before requesting production access. PLENO is now
live on Google Play, moving through Google's review for public
release.
What makes this AI-native, not AI-decorated
The distinction matters. Most apps bolt a chatbot onto existing
features. In PLENO, the agents
are
the operations:
Financial categorization happens automatically, not manually
The app tells you when you're heading toward a negative
balance before it happens — you don't have to check
"Can I afford this?" is a question you ask the app, and it
analyzes your real cash flow, commitments and goals to answer
with conditions
Voice replaces forms entirely for daily logging
Building for a real currency reality
The multi-currency system isn't a display feature — it's
infrastructure. Exchange rates are frozen per transaction (a
purchase from two weeks ago keeps the value it had that day,
not today's rate), users can override official rates with
parallel-market rates (a daily reality in Argentina), and net
worth is calculated across accounts in different currencies.
How we built it
Frontend:
React (via Babel standalone, no bundler),
custom CSS design system, Capacitor 6 for Android
Backend:
Firebase (Auth, Firestore, Cloud Functions,
Hosting)
AI:
Gemini API for all agent operations
Payments:
Google Play Billing (Android), Mercado Pago
(Argentina), Lemon Squeezy (international)
Web:
Deployed as PWA at app.plenoapp.com.ar with real-time
sync to the mobile app through the same Firestore layer
Challenges we faced
Building a complex product with AI tools, not a dev team.
PLENO was built end-to-end using AI coding assistants, one
person directing every decision — architecture, UX, business
logic. That meant learning to give AI the right context to
build reliably, and personally verifying every piece before
trusting it in production, since there was no second engineer
to catch what the first one missed.
The scope of "all-in-one" is a constant fight against
sprawl.
Finance, habits, projects, tasks, focus and notes
could each be their own startup. Keeping them coherent — one
design system, one AI layer, one data model — instead of six
disconnected features bolted together, was the hardest and
most ongoing design challenge of the whole build.
There is no playbook for launching a fintech-adjacent app
from Argentina.
Every "standard" path — payment processors,
banking, app store monetization — assumes an economy this
market doesn't have. We had to discover, one blocker at a
time, which global tools simply don't work here, and build
around each one without a roadmap to follow.
Real collaboration and real payments only reveal their
cracks under real use.
Multi-user projects that looked
correct in isolation broke under actual concurrent use. Payment
webhooks that worked in testing sent duplicate confirmations
in production. Shipping to real testers, not just testing
alone, was what actually exposed what needed to be rebuilt.
Racing a hard deadline while the product kept teaching us
what it needed to be.
Every week of user testing surfaced
something that needed rethinking — pricing, onboarding, which
features actually mattered. Balancing "ship what's real now"
against "this insight changes the product" under a fixed
deadline was a constant tension.
What we learned
Building for an underserved market means the constraints aren't
technical — they're structural. Every "standard" solution
(Stripe, seat-based pricing, USD-first billing) assumes an
economy that doesn't exist here. The features that matter most
to our users are the ones that no global tool considers worth
building.
What's next
Bank and Mercado Pago transaction sync, a Notion-grade note
editor with embedded diagrams, and expansion beyond Argentina
into Mexico and Colombia.
The linked repo returns 404, so it was deleted or made private.