Kotusev found that successful enterprise architecture practices in real organizations have nothing to do with the most widely known frameworks, such as Zachman, TOGAF, or FEAF, which he describes as management fads promoted by consulting companies [1]. ILEA™ comes from practice: these are the author’s own products, not client engagements, whose architecture was designed with the model.
Retana
Retail information for the whole organization: a single application where stores, suppliers, finance, sales, and supply chain teams check their KPIs.
How ILEA Was Applied
- Business: following a lean approach, it eliminates manual processes such as building sales and performance reports, which used to take analysts 3 to 4 days and delivered out-of-date information. The targets set by leadership reach every store manager daily, and each one checks sales and performance against those targets with up-to-date data: the whole organization works toward the same goals.
- Data: canonical retail data models, in the style of Informatica’s, as a common business language for the organization and for AI.
- Integration: ingestion and orchestration with Microsoft Fabric and Azure Data Factory.
- Platform: managed services on Azure instead of self-managed servers. Clients in Mexico are served from the Querétaro region (MXC) and clients in the United States from South Central US (SCUS): every user works close to their data, with low latency.
- Security: per-user roles and access, an audit log, and a choice of data residency (Querétaro or South Central US). Each client operates on its own subdomain, secured with a dedicated SSL/TLS certificate, which reinforces isolation between tenants.
- Applications: a single multi-tenant application for every area, built with open-source languages and components (C#, Python, .NET 10, and Radzen) and running on Azure App Service on Linux.
- Agents and AI: a conversational agent built with MCP (Model Context Protocol), in pilot, that answers questions about the KPIs in natural language.
Result: half of all queries are answered in under 0.2 s over more than 1.2 million sales rows.
EonDoc
Cloud safekeeping for a company’s documents that require a retention policy and must remain available for audits and tax verifications.
How ILEA Was Applied
- Business: gathering the documents requested for a tax verification or an audit used to take days or weeks; now they are located and delivered in minutes. The retention policy is also met by design, reducing the risk of being unable to produce a document when the authorities request it.
- Applications: a single multi-tenant application that serves every client company.
- Security: AES-256 encryption with a unique key per document; access with a corporate Microsoft account plus a virtual token, and download links that expire after 2 minutes.
- Platform: the application runs on Azure Web Apps in Querétaro (MXC), and documents are stored on AWS with redundancy across at least three availability zones.
- Integration: an agent on the client’s server syncs the selected folders.
- Data: retention aligned with the 5-year tax retention period.
See EonDoc on eondoc.com (in Spanish)
Veria Notarial
A robot that does a notary office’s data entry on government portals: the IRC Nuevo León alerts that used to be filled in by hand.
How ILEA Was Applied
- Business: the alert process was redesigned with a lean approach before automating it. The legal team no longer types 25 fields and 7 checkboxes per alert and spends that time reviewing and advising, with fewer errors because the data is validated before it is entered: at 200 alerts a month, the notary office recovers about 33 hours of work.
- Agents: the robot captures the data and a lawyer reviews and submits it (human-in-the-loop).
- Integration: it takes the DBA Notarial reports from a folder and fills in the forms on the MiPortal IRC-NL portal.
- Data: it validates CURP, RFC, email, and phone number, and detects duplicates before capturing.
- Security: the robot runs on-premises at the notary office: it reads the confidential data locally, enters it into the portal over a TLS-encrypted connection (HTTPS), and does not retain it, protecting personally identifiable information (PII). It also keeps a log of every process, with no personal data.
- Artificial Intelligence: AI reads and interprets the government capture portals (their forms, fields, and flows) and designs the execution plan the robot follows step by step. The automation is planned intelligently, not hand-coded.
Result: time per capture drops from about 10 minutes to 7 seconds.
See Veria Notarial on notariaweb.com (in Spanish)
References
- Kotusev, S. (2016). Enterprise Architecture Frameworks: The Fad of the Century. British Computer Society (BCS). https://www.bcs.org/articles-opinion-and-research/enterprise-architecture-frameworks-the-fad-of-the-century/