Digital Experience × Agentic AI Miami 🇺🇸 · 🇨🇴 EN / ES

Digital experience and agentic AI are colliding — and I've spent a decade standing right where they meet.

I'm Jonathan Montes. For ten years I helped Fortune 500s understand how people actually behave on their digital products. Now I build the agents that act on that understanding — sitting, by luck and a lot of curiosity, right where the two come together.

// a chatbot that actually knows my work — go ahead, ask it
10+
years in digital
experience & CX
$10M+
enterprise revenue
surfaced
4
AI products shipped,
most in production
// what I'm doing now

Leading the room, and building the thing.

Right now I lead Solutions Engineering across the Americas at Conviva — building a pre-sales motion for an AI-native platform, and sitting across from CIOs and CMOs while we work out what to build next.

But the throughline of my whole career is digital experience: understanding how real people move through products, where they get stuck, what the data is actually saying. Agentic AI is the most powerful tool that's ever existed for acting on that understanding — so I build with it, constantly. Most of what's below is running in production. I'd rather show you a working thing than a slide about it.

Jonathan Montes

Solutions Engineering Lead · Conviva
Miami, FL · Americas + LATAM Bilingual EN / ES Claude · OpenAI · Python · TS linkedin.com/in/jonathaneljach
// the intersection

Where two worlds collide.

Digital experience spent twenty years answering one question. Agentic AI showed up asking another. The interesting work is where they meet.

Digital Experience

How people actually behave

Journeys, friction, intent — the story the data tells. Two decades of getting software to understand the human in front of it.

×
the seam
Agentic AI

Software that acts

Reasoning, tools, autonomy — systems that don't just answer, they do. A couple of years of the ground shifting under everyone's feet.

Most teams are staffed for one side or the other. The rare, valuable work lives in the middle — agents grounded in real experience data, and experiences that adapt because something is reasoning underneath them.

I've spent a decade on the experience side and the last stretch deep in the agentic one. I won't pretend I planned that — I just happened to be standing in the right place when the two collided, with the tools and the curiosity to start building at the seam.

// things I've built

What building at the seam looks like.

Experience-led products with agents doing the work underneath. Code, not slideware — and a few you can open in a new tab right now.

Dechiva

Live
dechiva.com ↗

A World Cup travel assistant for groups. Drop in your reservation screenshots and it reads them, untangles who's booked what, flags the friend with no hotel two days out, and plans the free days around the matches. Built in Spanish, for the way friends actually travel.

Claude VisionRAGMulti-modelPython · JS
Open it →

Bartie

Live
bartolomeo.me ↗

The orchestration brain that runs my own life — agents for work, home, school, and family, proactive and bilingual, living on an Intel NUC under my desk and reachable over WhatsApp. It's where I prototype the agent patterns before they ever touch a customer, or a product like Dechiva. Full deep dive just below.

Multi-agentMCP serversSupabase · pgvectorClaude SonnetSelf-hosted
See it running →

Las Pelotas de Eva

Launching '26
laspelotasdeva.com ↗

Football, fandom, and group-trip logistics for Latin audiences — powered by Eva, an AI that talks like she's already in the group chat. Built for the 2026 World Cup across Colombia, Mexico, and US Latinos. Eva te lleva.

AgentWhatsApp-firstSupabaseapi-football
Take a look →

The Demo Engine

In the field
pre-sales · production GenAI

The GenAI demo system I bring into enterprise deals — vertical-specific agents for retail and fintech prospects, built on Anthropic and OpenAI with LangChain and LlamaIndex, prompt-engineered per use case and demoed under live evaluation. Where building and selling stop being separate jobs.

LangChainLlamaIndexVertical agentsPrompt eng.
// flagship build — the deep dive

Bartie: a personal AI chief of staff, running in production.

The card above is the short version. This is the system underneath — the most honest demo I have of what I believe about agents, because I run my actual life on it.

Bartie is one conversational interface over seven specialized agents — email, calendar, finances, smart home, language learning, football, notes-to-self. It's proactive (it messages first when something needs attention), bilingual, and reachable from WhatsApp, an Android app, or the web. The public site tells the product story; what it can't show you is the engineering underneath:

01It owns its infrastructure. I migrated the whole stack — orchestrator, dashboard, Postgres, vector embeddings — off Cloud Run and Vertex onto an Intel NUC behind a Cloudflare tunnel. Same public URLs, ~$3/month to run, cloud kept frozen as the rollback path.
02It helps build itself. A nightly dev-loop mines Bartie's own audit log for capability gaps, writes the code, runs the test gate, and deploys — after asking me for approval over WhatsApp.
03It's built like a product, not a script. Multi-tenant with row-level security and per-user credential vaults, first-party analytics, a watchdog with a live health map, and a 300-test suite gating every deploy.
7
specialized agents behind one interface
40
tools · 187 distinct actions
35
scheduled jobs, every single day
100%
self-hosted — no cloud runtime

Runs 24/7 on a NUC under Jon's desk.

Channels
WhatsAppAndroid appWeb dashboardOffice mic → Whisper
The brain
Express orchestrator7 agents40 toolsClaude Opus · Sonnet · Haiku
Foundation
Supabase Postgres + pgvectorLocal embeddingsContext-graph memory
all self-hosted · Intel NUC · Cloudflare tunnel · systemd full interactive architecture ↗
// work for customers

Proof, in the field.

Signature engagements — the enterprise wins behind the résumé. Experience data turned into outcomes executives could act on.

Automotive · Enterprise SaaS

Winning a $1M+ automotive deal with live behavioral intelligence

Contentsquare · 2022–2025

A major automotive OEM knew traffic was coming in but couldn't translate it into intent signals. We designed a POC demonstrating impression tracking at scale — capturing exactly what users viewed on vehicle listing pages and correlating that with CRM & inventory data in real time. When the most skeptical stakeholders saw predictive recommendations triggered by on-site behavior appearing live, the deal expanded on the spot.

POC DesignBehavioral AnalyticsCRM IntegrationExecutive Storytelling
$1M+
Initial contract value
30%
Reduction in eval cycle length
100%
Skeptic conversion (live demo)
Real-time
CRM + behavioral data merge
02
Fintech · Fraud Prevention

Building a cross-device fraud fingerprinting engine

Glassbox · 2018–2022

Fraudsters were gaming credit applications by switching devices, locations, and slightly altering names. We built a correlation engine that recognized the same individual across fragmented signals — device, location, browser fingerprint, typing behavior.

+5%lift in legitimate
application approvals
Identity ResolutionFraud AnalyticsRegulatory Compliance
03
Telecom · CX Operations

Saving $3M/year by giving call agents perfect context in <1 second

TracFone Wireless · 2015–2018

Before AI assistants existed, I unified 5+ data sources — digital journey, app logs, CRM, call history, telephony — into a real-time agent context panel. Average handle time dropped 30 seconds. At millions of calls per year, that compounded fast.

$3Msaved in year 1
— company-wide win
Real-Time PipelinesAPI ArchitectureVOC Operations
04
Financial Services · Data Strategy

Unifying Wells Fargo's digital data layer for executive-speed reporting

Wells Fargo · 2022

The Digital SVP needed a unified view across fragmented analytics sources to drive faster strategic decisions. I consolidated the data layer, standardized KPIs, and built the instrumentation framework that became the backbone of flagship reporting.

40%faster executive
reporting cycle
Data UnificationKPI FrameworksExecutive Stakeholders
// how I think about the work
01

Build the thing first. Talk about it second.

02

The best demo is a working one.

03

Translate, don't dazzle — executives buy clarity.

// the path here

A decade learning how people and products actually behave.

Every role here was digital experience and CX — the half of the equation most AI folks skip. It's the reason the agents I build hold up in the real world.

Dec 2025 — PresentConviva

Solutions Engineering Lead

Agentic AI · streaming intelligence platform

Leading pre-sales across the Americas: hiring and coaching the team, building an AI-native demo and evaluation methodology, and translating model behavior — capabilities, limits, and safety — into outcomes executive buyers can act on.

Aug 2022 — Dec 2025Contentsquare

Expert Solutions Consultant

Digital experience analytics · Series F

Owned technical wins on $1M–$5M enterprise deals across financial services, retail, and media. Designed AI-assisted POCs that cut evaluation cycles 30%, and built the ROI and governance frameworks adopted at CIO and CMO level.

120%+ quota · 3 years running
Jan 2022 — Aug 2022Wells Fargo

Strategic Advisor to SVP of Digital

Financial services · regulated environment

Consolidated fragmented analytics into a single data layer and standardized KPIs across teams, accelerating executive reporting by 40% — the instrumentation backbone for flagship reporting.

40% faster executive reporting
Jun 2021 — Jan 2022Tapestry

Sr. Manager, Global Digital Experience

Coach · Kate Spade · Stuart Weitzman

Built predictive customer-journey programs across three luxury brands, lifting checkout conversion and average order value, and turning behavioral data into operational strategy with C-suite stakeholders.

Nov 2018 — Jun 2021Glassbox

Manager, Global Business Insights & Pre-sales

Digital experience intelligence · NA · LATAM · EMEA

Owned pre-sales technical strategy across three regions and surfaced $10M in incremental revenue through behavioral analytics and session-replay POCs. Built a cross-device fraud-fingerprinting engine that lifted legitimate application approvals by 5%.

$10M revenue surfaced
Sep 2015 — Nov 2018TracFone Wireless

Operational Research & Digital CX Manager

Telecom · 25M+ customers · 40-person org

Architected a real-time agent-context layer — five data sources merged into a sub-second panel that gave call agents perfect context the instant a call connected. Handle time dropped 30 seconds; at millions of calls a year, that compounded to $3M+ saved annually. It was agent assist — a decade before the term existed.

$3M+/yr · agent assist, v0
2010 — 2015Early career

Research Analyst & PMO

CX analytics · BI · delivery

CX analytics, business intelligence, and SDLC delivery for large consumer businesses. Where the "find the signal, make it useful" instinct started — long before there was a model to hand it to.

// let's talk

Curious how it all fits together? Let's talk.

Whether it's the work above, where digital experience and AI are headed, or just a coffee and a good conversation — I'm easy to reach.

Currently leading Solutions Engineering at Conviva · always open to a good conversation
Hablo tu idioma. Escríbeme cuando quieras.