Andrés Briñez · Quito, Ecuador
Enterprise solutions architect · Odoo ERP and applied AI
And once it does, it can sell for less without earning less. That is where I come in: I put the ERP, the data and the AI inside the processes the company already has, until the work done by hand stops being done at all. I have done it with Odoo, in the cloud and on-premises, for manufacturing, retail, floriculture and services across Ecuador, Colombia and Australia.
A mid-sized company can run the way the big ones do.
Underneath, that takes structure. This is what there is today.
Queried on 2026-09-25
Make the sales report and the ledger agree on the same number.
3,442 accounting entries · 748 bank statement lines · 352 supplier invoices · 2 companies, two countries
My own books, in production. I try it here before I propose it.
A map of your systems that the board understands and the engineers can build from.
brinez_campana, brinez_conta_co, brinez_conta_ec, brinez_diagnostico, brinez_endurecer, brinez_escritorio, brinez_libros, brinez_seo, brinez_suscripciones, brinez_ventas, theme_brinez, web_company_color.
brinez_campana ← mail, crm, sale_management; brinez_conta_co ← account, l10n_co, l10n_co_withholding, report_xlsx; brinez_conta_ec ← account, analytic, l10n_ec_base; brinez_diagnostico ← website, crm, sale_management, portal, payment, account_payment; brinez_endurecer ← portal, website, website_crm, website_payment, sign_oca, helpdesk_mgmt, mail_tracking; brinez_escritorio ← web, website; brinez_libros ← website_blog, theme_brinez; brinez_seo ← website; brinez_suscripciones ← subscription_oca, brinez_ventas, sale_crm, website_crm, l10n_ec_sale, base_automation; brinez_ventas ← sale_management, sale_project, account_payment, payment_custom, payment_paypal, l10n_ec_sale, l10n_ec_account_edi, base_automation, crm, website_crm, brinez_campana, brinez_diagnostico; theme_brinez ← website, website_crm, brinez_seo; web_company_color ← web, base_sparse_field.
12 modules in this repository · 51 declared dependencies
The drawing comes from the depends of each manifest; nobody placed it by hand.
One system: the public site, the client portal and the back office on the same database.
203 modules installed today · 133 from Odoo · 49 from the OCA · 17 written here · 4 third parties
Counted in the production database, by the author each module declares.
Every invoice that arrives is filed by the first layer that knows.
By supplier rule → By the tax office activity code → By AI
355 invoices filed · 77 supplier rules
The three layers come from the code that applies them, in the order it tries them.
The answer comes back with the document it came from and the exact line, so it can be checked.
Retrieval, both ways at once
Answers, with their citation
Measurement, four numbers and a threshold
The page that explains all of it is published: /en/ai-over-your-own-data.
An agent that does the repetitive work, writes down why it did it, and stops where the decision is yours
131 books written up, one by one · 49 of those, about AI
Every book read leaves a published note: what it says and what changed because of it.
What I propose to you is already running here.
70 published pages · 1,071 automated tests
Spanish and English, with not one page left without its twin.
Four components that tend to show up together in the same project: architecture decides, the ERP executes, AI accelerates and the platform holds it all up.
An as-is survey, a gap analysis, and on that basis the target model: systems integration and architecture decisions written down. Modelled with C4, BPMN 2.0 and ArchiMate; documentation as code, versioned, that outlives the project.
Context map from USD 700
Requirements, configuration of the core modules, functional testing, go live and support. Community and Enterprise, versions 14 to 20, with electronic invoicing implemented in Ecuador and Odoo's fiscal localisations for the rest of the region.
Process mapping from USD 1,100
Language models and agents inside real processes: document automation, transcription and minutes, assistants over operational flows. Business cases with measurable return and governance from the design stage, with ISO/IEC 42001 and the NIST AI RMF as reference.
Quoted on the process
Cloud and on-premises deployments with Docker and Terraform, observability, backups and secure access with Cloudflare Zero Trust. Executive dashboards over ERP data and automation of the flows between systems.
Quoted on the scope
Herbert Simon called it satisficing: set the level a solution has to meet, then look for the alternative that meets it. There is no need to maximise effort, time and cost when the solution has to fit the resources and the deadline the client actually has.
How I work, step by step03 /
Architecture for a transformation programme, an ERP implementation that has to go well, or a conversation about bringing AI into a process without breaking it.
contacto@andresbrinez.com