# Emilio Bogantes: full site text Source: https://ejbogantes.me. Updated 2026-11-05. Spanish versions of each page are at /es/.md. --- title: "Emilio Bogantes | Technology & AI Executive, Costa Rica" description: "Co-founder and former CTO of Pangea Holdings. 15+ years leading engineering across the US and Latin America, getting AI out of the demo and into the work." canonical: https://ejbogantes.me/ language: en updated: 2026-11-05 --- # I get AI out of the demo and into the work. Co-founder and former CTO of Pangea Holdings. I facilitate senior leaders on MIT Professional Education programs, with Global Alumni, and teach at the public university where my career began. ## In one paragraph Emilio Bogantes is a Costa Rican technology and AI executive with more than 15 years in software engineering and leadership. He co-founded Pangea Holdings and was its Chief Technology Officer from March 2023 to July 2026. Earlier he was Director of Engineering at Soom, a regulated SaaS company serving medical-device manufacturers. He is a Learning Facilitator on MIT Professional Education programs, delivered in collaboration with Global Alumni, and an ad honorem professor at Universidad Nacional. He is based in Heredia, Costa Rica. ## Four before-and-afters. Titles say where I sat. These say what changed. - **~80%** lower cloud costs. Soom, a regulated SaaS for medical-device manufacturers. A five-month rebuild of the live product, inside FDA and EU MDR constraints. [https://ejbogantes.me/projects/regulated-saas-transformation](https://ejbogantes.me/projects/regulated-saas-transformation) - **4–6 weeks → 1–3 days** to onboard a customer. The same platform. About 80% of the manual onboarding work now runs by itself. [https://ejbogantes.me/projects/regulated-saas-transformation](https://ejbogantes.me/projects/regulated-saas-transformation) - **2 weeks → minutes** for complex forensic reports. A legal team facing 13,000+ scanned files. Specialist AI agents do the reading. People verify the result. [https://ejbogantes.me/projects/ai-assisted-legal-research](https://ejbogantes.me/projects/ai-assisted-legal-research) - **15 sites in 3–4 days** moved off WordPress. A US insurer, about 35,000 daily users per site. Scripts and AI handled the volume. Humans validated it. [https://ejbogantes.me/projects/ai-content-migration](https://ejbogantes.me/projects/ai-content-migration) ## Why I do this > Latin America already supplies the engineers. I want it to supply the leaders too. I’ve spent my career inside the US and Latin America delivery model. It works, and it has built good careers, mine included. Too often, though, the decisions get made in one place and the work gets done in another. So I hire for ownership. I teach at Universidad Nacional without pay. I’ve run a community where engineers here learn from each other. The aim is plain: more people from this region in the rooms where technology decisions get made. ## How I think **People × Processes × Data × AI = Business Outcomes** Multiply by zero anywhere and you get zero, however good the model is. So I find the weakest factor first and fix that one. ## Three conversations I’m always glad to have. - **A senior role.** CTO, VP of Engineering, Head of AI. Technology decides the outcome, and the mandate is real. - **A hard assessment.** An outside read on a platform or an AI program, before you commit more money to it. - **Teaching and public work.** Universities, fellowships and civic programs working on technology and leadership. --- --- title: "About Emilio Bogantes | Technology & AI Executive" description: "From a part-time job at a public university in Costa Rica to CTO and MIT Professional Education facilitator. The story, the principles and the education." canonical: https://ejbogantes.me/about language: en updated: 2026-11-05 --- # My career began at a public university in Costa Rica. I still teach there. In February 2011 I took a part-time job as a junior systems engineer at Universidad Nacional in Heredia. I looked after Linux servers and Active Directory and built the department’s website. From there I went where the problems got harder. At Baum I built e-commerce and booking systems for hospitality and luxury brands. At Hangar Worldwide I was promoted from back-end developer to Technology Lead, working on digital platforms for Fortune 500 clients. In 2022 I became Director of Engineering at Soom, a regulated SaaS for medical-device manufacturers. I was the most senior technologist in the company, answering to the CEO and investors. I led four managers, hired the Costa Rica team, and we rebuilt the live product in five months. In 2023 I co-founded Pangea Holdings and ran its technology as CTO until July 2026: two managers, 15 engineers, and AI systems that had to work on Monday morning for real customers. Since 2025 I’ve been a Learning Facilitator on MIT Professional Education programs in applied generative and agentic AI, in collaboration with Global Alumni. It is a facilitation role, not employment by MIT. Senior leaders from around the world bring me the question I keep meeting in delivery: what is worth building, and who will own it? And in 2023 I went back to Universidad Nacional, this time as an ad honorem professor. ## Two rooms, one person accountable. Most transformations stall between the executive room and the delivery room. Investment, risk and priorities get decided in one. Architecture, cadence and talent get decided in the other. I stay accountable in both, because that gap is usually where the money goes missing. ### What I hold to - Technology does not transform organizations. Leaders do. - Predictable delivery is a leadership job. You can’t hand it to the team and hope. - An architecture decision should lower risk and raise throughput. If it does only one, ask why. - AI belongs where you can measure the value, govern the risk and actually ship the thing. - Say the uncomfortable thing early. I once challenged a rushed upgrade on sites with 30,000 to 40,000 daily visits, laid out the risks and a staged plan, and we shipped with no data loss. ## Where this is going: Costa Rica stays home base. Next, I want a senior technology or AI role with real authority over the outcome. Over the longer term, I want to do at a larger scale what I already do at Universidad Nacional and in MIT cohorts: help this region produce the people who lead this technology, as well as the people who build it. ## Education - MBA in Artificial Intelligence. University of Cumbria, through Robert Kennedy College. Dissertation pending. - Chief Technology Officer Program, Blended Professional Certificate. MIT Professional Education - Bachelor’s degree in Systems Engineering. Universidad Nacional, Costa Rica - Software Architecture: Principles and Practices. Software Engineering Institute, Carnegie Mellon University ## Languages Spanish (native). English (C2). ## Away from work Golf for patience. Horseback riding, because a horse can tell when you’re not calm. Travel, because it keeps showing me what I assumed. --- --- title: "Teaching & Community | Emilio Bogantes" description: "Ad honorem professor at Universidad Nacional, community lead and mentor. Why Emilio Bogantes works to bring more Latin American leaders into technology decisions." canonical: https://ejbogantes.me/impact language: en updated: 2026-11-05 --- # The work nobody pays me for. Costa Rica has no shortage of good engineers. What’s scarcer is access to people who’ve been in the room and will explain how it works. That’s the gap I try to close. ## The record - **Universidad Nacional, Costa Rica**, Professor, ad honorem, March 2023 to present. AWS, artificial intelligence and software engineering, taught without pay at the public university where I studied and first worked. - **Desafío Latam**, Technology Team Lead and Mentor, May to July 2024. Volunteer mentor to teams learning to ship real technology work. - **Docker Costa Rica**, Community Lead, October 2019 to March 2020. Ran local events, technical translation and community outreach. - **WordCamp San José**, Speaker, September 2019. Public workshop: “Docker Compose Essentials”. ## Decisions travel. So should the people who make them. The nearshore model brought good jobs to this region, and I’ve benefited from it. It also tends to leave strategy in one country and execution in another. As AI changes who does what, that split could harden. Teaching is my most direct answer. It also keeps me honest: students ask why, and “because the vendor said so” doesn’t survive the question. ## The same idea shows up in how I hire and teach for a living. - At Soom I hired and built the Costa Rica engineering team, inside a company answering to FDA and EU MDR requirements. - At Pangea Holdings I built contract-to-hire paths toward internal ownership, so the knowledge stays with the people doing the work. - On MIT Professional Education programs, delivered with Global Alumni, I facilitate senior leaders in English and Spanish, including responsible AI, data readiness and risk. --- --- title: "Facts & Bios | Emilio Bogantes" description: "Quick, quotable answers about Emilio Bogantes: roles, results, education, public work and contact, plus reusable bios and machine-readable versions." canonical: https://ejbogantes.me/facts language: en updated: 2026-11-05 --- # The quick reference. For search firms, selection committees, journalists and the agents working for them. Everything here is safe to quote. ## Who is Emilio Bogantes? Emilio Bogantes is a Costa Rican technology and AI executive with more than 15 years in software engineering and leadership. He co-founded Pangea Holdings and was its Chief Technology Officer from March 2023 to July 2026. Earlier he was Director of Engineering at Soom, a regulated SaaS company serving medical-device manufacturers. He is a Learning Facilitator on MIT Professional Education programs, delivered in collaboration with Global Alumni, and an ad honorem professor at Universidad Nacional. He is based in Heredia, Costa Rica. ## What does he do now? He is a Learning Facilitator in applied generative and agentic AI on MIT Professional Education programs, delivered in collaboration with Global Alumni, which is a facilitation engagement and not employment by MIT. He also does independent architecture and AI assessment work and teaches ad honorem at Universidad Nacional. He is open to selected senior technology and AI roles. ## What size of organization has he led? At Pangea Holdings, two managers and 15 software engineers. At Soom, four managers, each with a team of about five engineers, distributed across Costa Rica, the United States and the Netherlands. At Ignite Visibility he managed five engineers and supported AI-assisted engineering adoption across 30. ## What results can be attributed to his work? At Soom: cloud costs reduced by about 80% and customer onboarding cut from 4 to 6 weeks to 1 to 3 days, after a five-month rebuild of the live product. In a forensic legal engagement: complex reporting cut from about two weeks to minutes across more than 13,000 files, with human verification. For a US insurer: 15 websites migrated from WordPress to Contentful in about three to four days. ## Has he put AI into production? Yes. His production systems include customer-service and logistics agents for companies in Mexico and Costa Rica and a forensic document-intelligence platform, built with RAG, tracing, cost monitoring, guardrails and human review. He stays hands-on with the code. ## Does he have regulated-industry experience? Yes. At Soom he led engineering for a medical-device SaaS working within FDA, EU MDR, GDPR and CCPA requirements. He also led engineering on a Fortune 100 healthcare loyalty platform. ## What is his education? A bachelor’s degree in Systems Engineering from Universidad Nacional, Costa Rica. The Chief Technology Officer Program certificate from MIT Professional Education. An MBA in Artificial Intelligence from the University of Cumbria through Robert Kennedy College, with the dissertation pending. ## What public or community work does he do? He has taught AWS, artificial intelligence and software engineering as an ad honorem professor at Universidad Nacional since March 2023. He was Community Lead of Docker Costa Rica, a volunteer mentor with Desafío Latam and a speaker at WordCamp San José. ## What is his leadership framework? People × Processes × Data × AI = Business Outcomes. Because it is a product, a factor near zero takes the whole result with it, so he finds and fixes the weakest factor first. ## Where is he based, and what languages does he work in? Heredia, Costa Rica, working with organizations in the United States, Latin America and Europe. He works in Spanish (native) and English (C2). ## How can he be contacted? By email at emilio@ejbogantes.me, through the contact form at ejbogantes.me/contact, or on LinkedIn at linkedin.com/in/ejbogantes. ## Bios, free to reuse **One line.** Emilio Bogantes is a Costa Rican technology and AI executive, co-founder and former CTO of Pangea Holdings, and a Learning Facilitator on MIT Professional Education programs. **Short.** Emilio Bogantes is a Costa Rican technology and AI executive with more than 15 years in engineering leadership. He co-founded Pangea Holdings and served as its CTO, led the rebuild of a regulated medical-device SaaS at Soom, and now facilitates senior leaders on MIT Professional Education programs delivered with Global Alumni. He teaches ad honorem at Universidad Nacional. --- --- title: "Work with Emilio Bogantes | Roles, Assessments, Teaching" description: "Three ways to work together: a senior technology or AI role, an independent platform or AI assessment, and teaching or speaking in English and Spanish." canonical: https://ejbogantes.me/services language: en updated: 2026-11-05 --- # Three ways to work together. I’m based in Costa Rica and work remotely, on site when the work needs it. ## A senior technology or AI role For CEOs, owners and boards with a real mandate. CTO, VP of Engineering, Director of Engineering, Head of AI. The fit is strongest where technology decides the outcome: a platform to rebuild without stopping the business, an AI program that has to reach production, or a distributed team that needs an operating model. ## A hard assessment For leaders about to commit more money. An outside read on a platform or an AI program: security, architecture, data exposure, cost and who actually owns what. You get findings ranked by business impact and a plan your team can carry out. ## Teaching and speaking For universities, executive programs and leadership groups. Applied generative and agentic AI, responsible AI and leading technology teams across borders. In English or Spanish, from practice rather than slides. --- --- title: "Career of Emilio Bogantes | Engineer to CTO" description: "Fifteen years with precise titles and dates: Universidad Nacional, Hangar Worldwide, Soom, Pangea Holdings, Ignite Visibility and MIT Professional Education." canonical: https://ejbogantes.me/career language: en updated: 2026-11-05 --- # From a university’s servers to the CTO’s chair. Fifteen years, in order, with dates. First I built systems, then teams, then the operating models that make both deliver. ## MIT Professional Education ### Learning Facilitator, Applied Generative AI and Agentic AI Jun 2025 - Present · Remote - In collaboration with Global Alumni. - Program focus: Applied Generative AI for Digital Transformation. - Mentorship and facilitation for senior leaders and global executives. - AI use-case selection, value measurement, and delivery governance. - Responsible AI, data readiness, security, risk, and organizational change. - Bilingual facilitation in English and Spanish. ## Independent engagements ### Architecture and AI Assessment Oct 2026 - Present · Remote - Contract assessment of application code and architecture for security, performance, maintainability, and AI integration risk. - Findings ranked by severity and business impact, with actionable hardening recommendations. ## Ignite Visibility ### Solutions Architect III Mar 2026 - Jul 2026 · Remote - Promoted based on sustained outcomes and stronger cross-account collaboration. - Senior architecture reference across multiple engagements. - Drove technical consistency, cross-team alignment, and client trust. ### Solutions Architect II Jan 2024 - Mar 2026 · Remote - Cloud-native modernization and enterprise platform delivery. - Architecture aligned with scalability, reliability, and measurable outcomes. - Roadmaps, dependency management, and delivery milestones. - CI/CD, operational readiness, and performance standards. - Executive guidance on sequencing, risk, and technical trade-offs. ## Pangea Holdings ### Co-Founder and Chief Technology Officer Mar 2023 - Jul 2026 · Costa Rica, Remote - Led technology strategy and execution across distributed US and LATAM teams. - Designed blended workforce models and vendor governance under MSA, SOW, and T&M. - Directed enterprise integrations across Oracle, SAP, Salesforce, HubSpot, and Zoho. - Defined architecture guardrails, cloud cost management, and operational automation. - Drove delivery accountability and contract-to-hire paths toward internal ownership. ## Shrine Development Client engagement context: Fortune 100 healthcare loyalty platform. ### Senior Engineering Lead and Consultant Aug 2023 - Jan 2024 · Remote - Delivery leadership in a regulated, high-reliability environment. - Coordinated execution across multiple pods and integration workstreams. - Led release readiness, architecture decisions, and sequencing. - Owned SOW execution, scope, milestones, and acceptance criteria. - Applied secure, risk-aware engineering practices. ## Soom Inc. ### Director of Engineering Mar 2022 - Nov 2023 · Greater Boston, Remote - Owned technology for a regulated SaaS platform with medical-device and eIFU constraints. - Rebuilt the platform from concept to production across architecture, team, and operating model. - Reduced delivery cycle from four weeks to one. - Reduced cloud infrastructure cost by 80 percent. - Automated 80 percent of manual operations. - Scaled the engineering organization by 50 percent while maintaining delivery predictability. - Managed vendor governance under MSA, SOW, and T&M. ## Edify Software Consulting ### Software Architect Technical Lead Jun 2021 - Mar 2022 · Costa Rica, Remote - Architectural leadership across multiple EdTech engagements. - Set architecture direction and delivery guardrails. - Mentored tech leads and engineers. - Led scalable backend and integration design. - Combined hands-on engineering with consulting and team leadership. ## Hangar Worldwide ### Technology Lead May 2019 - Jun 2021 · Remote - Enterprise digital delivery for Fortune 500 client environments. - Drove platform strategy and architecture across AWS, Azure, and GCP. - Defined CI/CD quality controls and operational readiness standards. - Led collaboration platforms and distributed vendor coordination. - Supported the Nissan customer-experience platform for dealer approvals. - Worked under MSA and SOW governance. ### Senior Back End Developer May 2018 - May 2019 · Remote - Built high-performance backend systems and scalable service architectures. - Led technical estimation, code reviews, and performance optimization. - Strengthened documentation and knowledge transfer practices. ### Intermediate Back End Developer Jun 2017 - May 2018 · Remote - Developed REST APIs using Laravel and Node.js. - Built third-party integrations and CI/CD flows on Linux with Jenkins and Docker. - Collaborated across cross-functional product and engineering teams. ## Baum Digital Marketing ### Web Developer Jul 2014 - Jun 2017 · Costa Rica, On-site - Built full-stack web and product solutions across e-commerce, hospitality, and luxury sectors. - Implemented booking platforms, third-party integrations, and custom CMS solutions. - Worked across PHP, JavaScript, payment integrations, and data-enabled experiences. - Partnered with design and business stakeholders to ship usable products. ## Independent Consulting Parallel engagements while holding full-time leadership and engineering roles. ### Freelance Software Engineer and Technology Consultant Dec 2011 - Mar 2022 · Remote - Delivered architecture and full-stack execution for startups, growth companies, and enterprise clients. - Advised on technology strategy, platform direction, and modernization paths. - Worked across backend, frontend, and data modernization in multiple industries. ## SEPUNA, Universidad Nacional de Costa Rica ### Jr. Information Systems Engineer Feb 2011 - Dec 2011 · Heredia, On-site, Part-time - Started in systems administration, Linux servers, and Active Directory support. - Contributed as webmaster and full-stack developer using PHP, Joomla, HTML, CSS, and JavaScript. - Supported technical operations and infrastructure reliability. ## Alongside the day job - **Professor, Universidad Nacional (ad honorem)**, Mar 2023 - Present. Community education and mentorship in technology-related programs. - **Technology Team Lead and Mentor, Desafío Latam**, May 2024 - Jul 2024. Mentored teams and supported practical technology execution pathways. - **Community Lead, Docker Costa Rica**, Oct 2019 - Mar 2020. Led local community initiatives and technical learning events. - **Speaker, WordCamp San José**, Sep 2019. Delivered public talk for the regional technology community. --- --- title: "Case Studies | Applied AI & Regulated Platforms | Bogantes" description: "Six case studies with before-and-after results: production AI in logistics and legal work, a 15-site migration and a regulated medical-device SaaS rebuild." canonical: https://ejbogantes.me/projects language: en updated: 2026-11-05 --- # What changed, and how. Six engagements from applied AI, enterprise delivery and regulated platforms. Some clients stay anonymous under NDA. Figures marked as approximate are approximate. ## [Applied AI for construction materials and logistics](https://ejbogantes.me/projects/ai-construction-logistics) An AI operation that captured after-hours demand, extended customer-service capacity, and optimized delivery routing. - Avoided hiring approximately three to four additional people - Sales increased approximately 20% - Route costs decreased approximately 30% ## [AI-enabled logistics operations](https://ejbogantes.me/projects/ai-logistics-operations) An AI layer over logistics operations that recovered missed leads and increased qualified opportunities at scale. - Sales increased approximately 15% ## [Internal AI operations at Pangea Holdings](https://ejbogantes.me/projects/pangea-internal-ai-operations) Internal AI operations that reduced manual workload and lifted sales across the organization. - Sales increased approximately 10% ## [AI-assisted legal research](https://ejbogantes.me/projects/ai-assisted-legal-research) A forensic research tool that interprets and searches a large body of legal protocols. - Complex forensic reporting reduced from about two weeks to minutes, with human verification - Minimum viable product in two weeks, production in six - Supported legal work connected to a claim exceeding USD 1 million ## [AI-assisted enterprise content migration](https://ejbogantes.me/projects/ai-content-migration) AI-assisted migration of enterprise content from WordPress to Contentful across 15 websites. - Migrated 15 sites in approximately three to four days, including scripting and AI plus human validation ## [Regulated SaaS transformation](https://ejbogantes.me/projects/regulated-saas-transformation) A full platform reconstruction for a regulated SaaS serving medical-device manufacturers, with distributed delivery and disciplined vendor governance. - Delivery cycle reduced from four weeks to one - Cloud infrastructure cost reduced by 80% - Approximately 80% of manual operations automated --- --- title: "Writing on AI & Technology Leadership | Emilio Bogantes" description: "Eight essays from practice: why AI pilots stall, cost per completed task, governing agentic AI, responsible AI and running teams across the US and Latin America." canonical: https://ejbogantes.me/media language: en updated: 2026-11-05 --- # What I’ve learned, written down. Essays on operating models, AI in production and governance that helps delivery. They come from the work and cite it. - [People × Processes × Data × AI: An Operating Model for Measurable Transformation](https://ejbogantes.me/media/operating-model-measurable-transformation): Transformation results come from designing four factors as one system. When any factor is near zero, the outcome collapses no matter how good the technology is. - [Why Enterprise AI Pilots Fail Before Production](https://ejbogantes.me/media/why-enterprise-ai-pilots-fail): Most AI pilots do not fail because the model is wrong. They fail because they were never designed to reach production in the first place. - [How to Govern Agentic AI Without Slowing Delivery](https://ejbogantes.me/media/govern-agentic-ai-without-slowing-delivery): Agentic systems act, not just answer. Governance has to move to where the action happens, or it becomes a brake instead of a control. - [Cost per Completed Task: The Metric That Changes AI Economics](https://ejbogantes.me/media/cost-per-completed-task): Measuring AI by tokens or seats hides whether it is actually cheaper. Cost per completed task tells you the truth and changes which use cases are worth funding. - [Scaling Engineering Organizations Across the US and Latin America](https://ejbogantes.me/media/scaling-engineering-us-latam): Distributed US and LATAM teams do not succeed because of cost arbitrage. They succeed when hiring, operating cadence, and vendor governance are designed on purpose. - [Responsible AI Is an Engineering Discipline, Not a Policy Deck](https://ejbogantes.me/media/responsible-ai-engineering-discipline): Principles on a slide do not change what an AI system does in production. Controls in the system do. - [From AI Use Case to Measurable Business Value](https://ejbogantes.me/media/ai-use-case-to-business-value): A practical path from picking an AI use case to proving it moved a number leadership cares about. - [Vendor Governance for Enterprise Technology Delivery](https://ejbogantes.me/media/vendor-governance-enterprise-delivery): MSA, SOW, and T&M are not paperwork. Used well, they are how external delivery stays accountable to outcomes instead of hours. ## Elsewhere - [What Is Headless CMS: A Guide to These Key Platforms](https://ignitevisibility.com/headless-cms-platforms/). Ignite Visibility, 2026, Author - [The Game-Changing Benefits of SaaS for Modern Businesses](https://ejbogantes.medium.com/the-game-changing-benefits-of-saas-for-modern-businesses-say-goodbye-to-old-school-tech-struggles-4b4f7edc70f7). Medium, Author - [Creating Your First AWS Lambda](https://ejbogantes.medium.com/creating-your-first-aws-lambda-821d8105e2db). Medium, Author - [Hackeo en CCSS enciende alarmas en seguridad de datos personales](https://www.elfinancierocr.com/tecnologia/hackeo-en-ccss-enciende-alarmas-en-seguridad-de-datos-personales/KVYT2VYBONGTNPMCQHMGHU3L5E/story/). El Financiero, 2015, Quoted as an expert --- --- title: "Contact Emilio Bogantes" description: "Write to Emilio Bogantes about a senior technology or AI role, an independent assessment, or teaching and public work. He reads every message himself." canonical: https://ejbogantes.me/contact language: en updated: 2026-11-05 --- # Tell me what’s stuck. What’s happening, what it costs if nothing changes, and when you need to decide. I read every message myself and reply when I can genuinely help. ## Useful to include - What’s stuck, in a sentence or two - What it costs if it stays stuck - Who decides, and by when ## Or directly - Email: [emilio@ejbogantes.me](mailto:emilio@ejbogantes.me) - LinkedIn: [https://www.linkedin.com/in/ejbogantes](https://www.linkedin.com/in/ejbogantes) - Form: https://ejbogantes.me/contact --- --- title: "Applied AI for construction materials and logistics" description: "An AI operation that captured after-hours demand, extended customer-service capacity, and optimized delivery routing." canonical: https://ejbogantes.me/projects/ai-construction-logistics language: en updated: 2026-11-05 --- # Applied AI for construction materials and logistics Construction materials and logistics. Querétaro, Mexico. Client described anonymously under NDA. An AI operation that captured after-hours demand, extended customer-service capacity, and optimized delivery routing. ## Challenge Leads were missed outside business hours, customer-service capacity was insufficient, and delivery routes were inefficient. ## Role Architecture, team leadership, implementation oversight, and hands-on contribution throughout delivery. ## Architecture - Zoho - Dify - Langfuse - PostgreSQL - Google services - WhatsApp - LLM providers ## Scale - 10 internal users - ~2,500–5,000 external users ## Outcomes - Avoided hiring approximately three to four additional people - Sales increased approximately 20% - Route costs decreased approximately 30% ## Lesson After-hours demand is often the cheapest growth available. Capturing it needed integration and routing discipline, not additional headcount. --- --- title: "AI-enabled logistics operations" description: "An AI layer over logistics operations that recovered missed leads and increased qualified opportunities at scale." canonical: https://ejbogantes.me/projects/ai-logistics-operations language: en updated: 2026-11-05 --- # AI-enabled logistics operations Logistics operations. Heredia, Costa Rica. Client described anonymously under NDA. An AI layer over logistics operations that recovered missed leads and increased qualified opportunities at scale. ## Challenge Missed leads and operational workload limited the volume of qualified opportunities the team could pursue. ## Role Architecture, team leadership, delivery oversight, and hands-on implementation. ## Architecture - Odoo - Dify - Langfuse - PostgreSQL - Google services - WhatsApp ## Scale - 15 internal users - ~30,000–35,000 external users ## Outcomes - Sales increased approximately 15% ## Lesson An AI layer only compounds when it sits inside the real operational workflow. Recovered leads become revenue when the process, not just the model, is in place. --- --- title: "Internal AI operations at Pangea Holdings" description: "Internal AI operations that reduced manual workload and lifted sales across the organization." canonical: https://ejbogantes.me/projects/pangea-internal-ai-operations language: en updated: 2026-11-05 --- # Internal AI operations at Pangea Holdings Technology holdings. Heredia, Costa Rica. Internal AI operations that reduced manual workload and lifted sales across the organization. ## Challenge Internal operations carried manual workload that constrained responsiveness and commercial output. ## Role Architecture, team leadership, implementation oversight, and hands-on contribution. ## Architecture - Jira - Dify - Langfuse - PostgreSQL - Google services - WhatsApp ## Scale - 5 internal users - ~2,000–2,500 external audience ## Outcomes - Sales increased approximately 10% ## Lesson Internal AI operations earn trust by proving value on real work first. Measured gains make the next investment an easy decision. --- --- title: "AI-assisted legal research" description: "A forensic research tool that interprets and searches a large body of legal protocols." canonical: https://ejbogantes.me/projects/ai-assisted-legal-research language: en updated: 2026-11-05 --- # AI-assisted legal research Legal. San José, Costa Rica. Client described anonymously under NDA. A forensic research tool that interprets and searches a large body of legal protocols. ## Challenge The engagement required a forensic research tool capable of interpreting and searching a large body of legal protocols. ## Role Full architecture and implementation ownership. ## Architecture - Parallel AI agents (Claude and OpenAI) - Tesseract OCR - Next.js - Retrieval-augmented generation (RAG) ## Scale - 5 internal users - 13,000+ scanned TIFF files ## Outcomes - Complex forensic reporting reduced from about two weeks to minutes, with human verification - Minimum viable product in two weeks, production in six - Supported legal work connected to a claim exceeding USD 1 million ## Lesson In high-stakes research, observability and human validation are not overhead. They are what make an AI system responsible enough to rely on. --- --- title: "AI-assisted enterprise content migration" description: "AI-assisted migration of enterprise content from WordPress to Contentful across 15 websites." canonical: https://ejbogantes.me/projects/ai-content-migration language: en updated: 2026-11-05 --- # AI-assisted enterprise content migration United States insurance. United States. Client described anonymously under NDA. AI-assisted migration of enterprise content from WordPress to Contentful across 15 websites. ## Challenge Content had to be migrated from WordPress to Contentful across 15 production websites without disrupting large daily audiences. ## Role Full architecture and implementation ownership. ## Architecture - Parallel Claude and OpenAI workflows - Node.js and CLI scripts - AI-assisted planning, content structures, and validation ## Scale - 15 websites - ~25 internal users - ~35,000 daily users per website ## Outcomes - Migrated 15 sites in approximately three to four days, including scripting and AI plus human validation ## Lesson Large migrations compress when planning, structure, and validation are treated as engineering with AI in the loop, not as manual labor. --- --- title: "Regulated SaaS transformation" description: "A full platform reconstruction for a regulated SaaS serving medical-device manufacturers, with distributed delivery and disciplined vendor governance." canonical: https://ejbogantes.me/projects/regulated-saas-transformation language: en updated: 2026-11-05 --- # Regulated SaaS transformation Regulated SaaS (medical-device manufacturers). United States, remote-first. A full platform reconstruction for a regulated SaaS serving medical-device manufacturers, with distributed delivery and disciplined vendor governance. ## Challenge A regulated SaaS platform required full reconstruction while improving delivery speed, cost, and operational reliability. ## Role Director of Engineering: architecture, team, and operating model from concept to production. ## Architecture - Cloud-native platform reconstruction - Delivery automation and operational tooling - MSA, SOW, and T&M vendor governance ## Scale - Distributed, remote-first engineering organization - Engineering organization grew by 50% ## Outcomes - Delivery cycle reduced from four weeks to one - Cloud infrastructure cost reduced by 80% - Approximately 80% of manual operations automated ## Lesson Cost, speed, and reliability improve together only when the operating model is redesigned as one system, not optimized in isolated parts. --- --- title: "People × Processes × Data × AI: An Operating Model for Measurable Transformation" description: "Transformation results come from designing four factors as one system. When any factor is near zero, the outcome collapses no matter how good the technology is." canonical: https://ejbogantes.me/media/operating-model-measurable-transformation language: en updated: 2026-11-05 --- # People × Processes × Data × AI: An Operating Model for Measurable Transformation *Transformation results come from designing four factors as one system. When any factor is near zero, the outcome collapses no matter how good the technology is.* By Emilio Bogantes, 2026-08-18 Technology does not transform organizations. Leaders do. After fifteen years leading engineering and transformation programs, the pattern I trust most is simple to write and hard to execute: People × Processes × Data × AI = Business Outcomes. It is a product, not a sum. You cannot compensate for a weak factor by over-investing in another. A brilliant model on top of broken processes still produces broken results. Strong engineers without clean data spend their time firefighting. This is why isolated technology decisions rarely move the number that leadership actually cares about. ## Why it multiplies instead of adds Treat each factor as a coefficient between zero and one. If people are aligned and capable you are near one; if processes are ad hoc you might be at 0.3; if data is unreliable you might be at 0.4. Multiply those together and the ceiling on your AI investment is already set before the first model runs. The discipline is to find the lowest factor and raise it, not to add more technology on top of the constraint. ## How I apply it in practice - People: clarify ownership and decision rights before tooling. Accountability is a design choice, not a personality trait. - Processes: make the delivery cadence explicit and measurable, so improvement is observable rather than anecdotal. - Data: treat data readiness as a prerequisite for AI, not a side effect. Fix the inputs before trusting the outputs. - AI: select use cases by business value and feasibility, then instrument them so value can be measured and governed. ## A concrete example In a regulated SaaS rebuild serving medical-device manufacturers, the gains did not come from a single technology. They came from redesigning the operating model as one system: clearer accountability, an explicit delivery cadence, and automation applied where the process was already understood. The result was a delivery cycle reduced from four weeks to one, roughly 80% of manual operations automated, and an 80% reduction in cloud infrastructure cost, while the engineering organization grew by 50% and delivery stayed predictable. None of those numbers is a technology story on its own. Each is the output of the four factors moving together. That is the point of the model: it forces you to invest where the constraint actually is. ## What to do first Start by scoring your four factors honestly for one initiative. The lowest score is your real roadmap. Raise it before you add anything else, and measure the outcome in business terms, not activity. Transformation becomes measurable when the system is designed on purpose. --- --- title: "Why Enterprise AI Pilots Fail Before Production" description: "Most AI pilots do not fail because the model is wrong. They fail because they were never designed to reach production in the first place." canonical: https://ejbogantes.me/media/why-enterprise-ai-pilots-fail language: en updated: 2026-11-05 --- # Why Enterprise AI Pilots Fail Before Production *Most AI pilots do not fail because the model is wrong. They fail because they were never designed to reach production in the first place.* By Emilio Bogantes, 2026-08-18 A demo is not a system. I have watched capable teams build an impressive AI prototype in weeks and then spend a year unable to ship it. The failure is rarely the model. It is that the pilot optimized for a convincing demonstration instead of a path to production. ## The four gaps that stall pilots - No owner of the business outcome. A pilot with no one accountable for the number it should move becomes a science project. - Data that works in the demo but not in reality. Curated sample data hides the messy inputs production will actually send. - No integration surface. If the pilot cannot reach the systems where work happens, it cannot change how work is done. - No measurement or governance. Without instrumentation you cannot prove value, and without controls you cannot operate it safely. ## Design for production from day one The fix is to invert the sequence. Before building, define the single business metric the use case must move, the integration points it must touch, and the guardrails it must respect. Then build the smallest version that runs against real data through those real integrations. It is less impressive in a demo and far more likely to survive contact with production. ## What this looks like when it works In an applied-AI operation for a construction-materials and logistics business, the goal was set in business terms first: capture demand outside business hours, extend service capacity, and improve routing. The system was built against the real stack the team already used, with observability in place. It captured after-hours demand, helped avoid three to four additional hires, and moved the numbers that mattered: sales up by roughly 20% and route costs down by roughly 30%. In a separate logistics operation, the same discipline of tying an AI layer to a real operational workflow and instrumenting it produced an approximately 15% increase in sales. The common factor was never a smarter model. It was designing for production, integration, and measurement from the start. ## A short checklist before you fund a pilot - Name the business metric and the person accountable for it. - Confirm the pilot runs on realistic, not curated, data. - Identify the integration surface into real systems of work. - Decide how value and risk will be measured before you build. --- --- title: "How to Govern Agentic AI Without Slowing Delivery" description: "Agentic systems act, not just answer. Governance has to move to where the action happens, or it becomes a brake instead of a control." canonical: https://ejbogantes.me/media/govern-agentic-ai-without-slowing-delivery language: en updated: 2026-11-05 --- # How to Govern Agentic AI Without Slowing Delivery *Agentic systems act, not just answer. Governance has to move to where the action happens, or it becomes a brake instead of a control.* By Emilio Bogantes, 2026-08-18 The moment an AI system can take actions rather than only produce text, governance stops being a document and becomes an engineering concern. Agentic AI raises the stakes because a mistake is no longer a wrong answer a human reviews; it can be an action that already happened. The instinct is to slow everything down with approvals. That is the wrong lever. ## Put the controls in the system, not in the meeting Governance that lives in review meetings scales poorly and frustrates good teams. Governance that lives in the system scales with the work. The goal is to make the safe path the default path, so delivery speed and control move in the same direction. - Scope the agent's authority explicitly: what it may read, what it may write, and where a human must confirm. - Instrument every action with tracing so behavior is observable and auditable after the fact. - Bound autonomy by reversibility: let agents act freely on reversible steps, and require confirmation for irreversible ones. - Evaluate continuously against real tasks, not one-time acceptance, because agent behavior drifts as inputs change. ## Reversibility is the key dial Most delivery friction comes from treating every agent action as equally risky. It is not. Reading and drafting are cheap to undo; sending, deleting, and committing are not. When you tie the level of human confirmation to how reversible an action is, you remove approvals from the low-risk majority of work and concentrate human judgment where it actually matters. ## Grounded in real delivery In an AI-assisted legal-research tool, parallel agents interpreted and searched a large body of legal protocols using OCR and retrieval-augmented generation. Observability and human validation were part of the design, which is what let the output responsibly support legal work connected to a claim exceeding one million dollars. In an enterprise content migration across fifteen websites, the same pattern of AI plus human validation and traceable steps let the work compress into days rather than weeks without losing control. Responsible AI is an engineering discipline. When the controls are built into how the system runs, governance stops competing with delivery and starts enabling it. --- --- title: "Cost per Completed Task: The Metric That Changes AI Economics" description: "Measuring AI by tokens or seats hides whether it is actually cheaper. Cost per completed task tells you the truth and changes which use cases are worth funding." canonical: https://ejbogantes.me/media/cost-per-completed-task language: en updated: 2026-11-05 --- # Cost per Completed Task: The Metric That Changes AI Economics *Measuring AI by tokens or seats hides whether it is actually cheaper. Cost per completed task tells you the truth and changes which use cases are worth funding.* By Emilio Bogantes, 2026-08-18 Most AI budgets are measured in the wrong units. Tokens and per-seat licenses tell you what you spent, not what you got. The question a business actually needs answered is simpler: what did it cost to complete one real unit of work, end to end, at an acceptable quality? ## Define the task, then price the whole thing A completed task is a full unit of work the business already recognizes: one migrated page, one researched question, one qualified lead handled. The honest cost includes the model calls, the human review, and the rework when the first attempt is wrong. Leave out review and rework and you are measuring a demo, not an operation. - Direct model and infrastructure cost per task. - Human-in-the-loop review and correction time. - Rework rate: how often a task has to be redone. - The human baseline: what the same completed task cost before. ## Why this changes decisions When you price the whole task, some exciting use cases turn out to cost more than the humans they were meant to replace, because review and rework quietly dominate. Others look modest in a demo but collapse the cost per task by an order of magnitude. In an enterprise content migration across fifteen websites, the economics worked because AI plus human validation compressed work that would take weeks into days. The point was never tokens; it was the cost to reliably complete the migration. Track cost per completed task from the first pilot and you fund the use cases that are genuinely cheaper, not the ones that merely look impressive. --- --- title: "Scaling Engineering Organizations Across the US and Latin America" description: "Distributed US and LATAM teams do not succeed because of cost arbitrage. They succeed when hiring, operating cadence, and vendor governance are designed on purpose." canonical: https://ejbogantes.me/media/scaling-engineering-us-latam language: en updated: 2026-11-05 --- # Scaling Engineering Organizations Across the US and Latin America *Distributed US and LATAM teams do not succeed because of cost arbitrage. They succeed when hiring, operating cadence, and vendor governance are designed on purpose.* By Emilio Bogantes, 2026-08-18 The cheapest reason to build in Latin America is the wrong reason. Teams that treat nearshore as pure cost arbitrage get exactly what they pay for. The organizations that win treat US and LATAM as one distributed system, designed for real-time-zone overlap and clear ownership. ## Three things you actually design - Hiring and time zones: staff for genuine overlap with US working hours so collaboration is synchronous where it matters and asynchronous where it does not. - Operating cadence: make the delivery rhythm explicit across locations, so status is observable rather than negotiated in every meeting. - Vendor governance: structure external delivery under MSA, SOW, and T&M so accountability is contractual, not informal. ## Toward internal ownership Blended workforces work best when they are a path, not a permanent dependency. Contract-to-hire routes move the strongest external contributors toward internal ownership, which is where durable institutional knowledge lives. As a CTO leading distributed US and LATAM teams, I designed exactly this: blended models with vendor governance that trended toward internal ownership over time. ## Scale without losing predictability Growth is the moment delivery predictability usually breaks. In a regulated SaaS rebuild, the engineering organization grew by 50% while delivery stayed predictable, because the operating model scaled with the headcount instead of after it. Distributed scale is an organizational design problem first and a hiring problem second. --- --- title: "Responsible AI Is an Engineering Discipline, Not a Policy Deck" description: "Principles on a slide do not change what an AI system does in production. Controls in the system do." canonical: https://ejbogantes.me/media/responsible-ai-engineering-discipline language: en updated: 2026-11-05 --- # Responsible AI Is an Engineering Discipline, Not a Policy Deck *Principles on a slide do not change what an AI system does in production. Controls in the system do.* By Emilio Bogantes, 2026-08-18 Most responsible-AI programs stop at a set of principles everyone agrees with and no system enforces. Fairness, transparency, and safety on a slide are necessary and completely insufficient. What changes behavior in production is whether those principles are expressed as controls that ship with the product. ## From principles to controls - Scope: encode what a system may access and act on, rather than trusting it to behave. - Observability: trace inputs, decisions, and actions so behavior is auditable after the fact. - Evaluation: test against real tasks continuously, because model behavior drifts as inputs change. - Human validation: place a person where the stakes and irreversibility justify the cost, and nowhere it does not. ## What it looks like in high-stakes work In an AI-assisted legal-research tool, responsibility was not a document; it was design. Observability and human validation were built into how the system ran, which is precisely what let its output responsibly support legal work connected to a claim exceeding one million dollars. Remove those controls and the same model becomes a liability rather than an asset. Treat responsible AI as an engineering discipline and it stops being a compliance conversation held after the fact. It becomes part of how the system is built, which is the only place it can actually take effect. --- --- title: "From AI Use Case to Measurable Business Value" description: "A practical path from picking an AI use case to proving it moved a number leadership cares about." canonical: https://ejbogantes.me/media/ai-use-case-to-business-value language: en updated: 2026-11-05 --- # From AI Use Case to Measurable Business Value *A practical path from picking an AI use case to proving it moved a number leadership cares about.* By Emilio Bogantes, 2026-08-18 Most AI programs die in the gap between an interesting use case and a reported business result. Closing that gap is not luck. It is a short, repeatable path from selection to proof. ## The path - Select by value and feasibility: prefer use cases where a clear business metric is within reach, not the most technically impressive one. - Name the metric and the owner: one number to move, one person accountable for it. - Instrument before you scale: if you cannot measure the metric before and after, you cannot prove value. - Report in business terms: present the delta in revenue, cost, or cycle time, not model activity. ## Proof, not anecdotes In an applied-AI operation for a construction-materials and logistics business, the metrics were chosen first and measured throughout: sales rose by roughly 20% and route costs fell by roughly 30%, while the system helped avoid three to four additional hires. In a separate logistics operation, the same discipline produced an approximately 15% increase in sales. None of these are model metrics. They are business outcomes, which is the only kind of result that survives a budget review. Pick for value, measure honestly, and report the number that matters. That is how a use case becomes a decision leadership can stand behind. --- --- title: "Vendor Governance for Enterprise Technology Delivery" description: "MSA, SOW, and T&M are not paperwork. Used well, they are how external delivery stays accountable to outcomes instead of hours." canonical: https://ejbogantes.me/media/vendor-governance-enterprise-delivery language: en updated: 2026-11-05 --- # Vendor Governance for Enterprise Technology Delivery *MSA, SOW, and T&M are not paperwork. Used well, they are how external delivery stays accountable to outcomes instead of hours.* By Emilio Bogantes, 2026-08-18 External delivery fails quietly when the contract rewards effort instead of results. Vendor governance is not procurement overhead; it is the mechanism that keeps outside teams accountable to the same outcomes as your own. ## Match the contract to the risk - MSA sets the durable terms: security, IP, liability, and the rules of engagement across every statement of work. - SOW is where accountability lives: scope, milestones, and explicit acceptance criteria tied to outcomes. - T&M fits genuinely uncertain work, but only with visibility and guardrails so time does not drift away from value. ## Acceptance criteria are the real control The single most useful clause is a clear definition of done. When acceptance criteria are explicit, a vendor is accountable to a result you can verify, not to a plausible-sounding status update. I have run delivery for a Fortune 100 client environment where SOW execution, scope, and acceptance criteria were the difference between a controlled release and an open-ended engagement. ## Govern toward ownership The best vendor relationships have an exit built in. In regulated SaaS delivery, I managed vendors under MSA, SOW, and T&M while steering the strongest contributors toward internal ownership through contract-to-hire. Governance is not about controlling vendors forever; it is about keeping delivery accountable while you build the capability you ultimately want to own.