Génécia

An AI-powered platform that helps businesses build websites and PWAs, automate workflows, and manage customer services from one multilingual workspace.

Comment

Inspiration

Small businesses often manage their work through disconnected tools: customer information in one application, sales documents in another, website content somewhere else, and essential knowledge scattered across emails, messages, and files.

Génécia was created to bring these workflows together in one modular SaaS platform.

Our goal is not to place a generic chatbot beside existing business tools. We want to integrate controlled AI assistance directly into the workflows where people already work.

Our guiding principle is simple:

AI proposes. The user decides.

What it does

Génécia is a modular SaaS platform for small businesses, independent professionals, and organizations.

It brings together:

  • customer relationship management;
  • contacts and companies;
  • sales and administrative documents;
  • websites and progressive web applications;
  • page and content management;
  • blogs and communication campaigns;
  • media and document management;
  • analytics and operational automation.

Over the past two months, Codex has been our primary engineering partner in translating fifteen years of accumulated product knowledge and business rules into Génécia V2.

During OpenAI Build Week, we extended this foundation with a working GPT‑5.6 editorial assistant inside the Blog module.

The user can provide:

  • a topic;
  • an editorial instruction;
  • an existing draft when available.

GPT‑5.6 then generates:

  • a natural and precise title;
  • a concise introduction;
  • a complete HTML article divided into useful sections.

The result is inserted directly into the normal Génécia editor. It is never saved or published automatically.

The user can review, modify, save, publish, or discard the proposal. If generation fails, the content already entered in the form remains untouched.

GPT‑5.6 also supports the multilingual adaptation of Génécia V2 by helping us translate interface content while preserving consistent terminology across navigation, forms, instructions, and user actions.

The Blog assistant is the first fully deployed example of the reusable AI integration pattern being introduced across Génécia V2.

How we built it

Génécia is built primarily with PHP, MySQL, JavaScript, HTML, and CSS. It uses centralized routing and reusable components without depending on a large application framework.

Codex supported the V2 reconstruction by helping us:

  • inspect and understand the historical architecture;
  • identify active routes, shared components, and dependencies;
  • preserve existing business rules;
  • rebuild modules inside the V2 architecture;
  • create the new page-editing experience;
  • modernize interfaces and navigation;
  • maintain compatibility with existing customer data;
  • diagnose routing, database, encryption, and credential issues;
  • test complete workflows in the deployed application.

For the Build Week extension, Génécia uses the OpenAI API with the gpt-5.6-sol model.

Requests use a constrained JSON response format. The title, introduction, and article body are received as separate fields and validated before anything is inserted into the editor.

The Blog module does not store its own API credentials or duplicate the OpenAI integration. It uses Génécia’s centralized OpenAI configuration and AI activity history.

The central layer:

  • resolves the API key authorized for the current customer workspace;
  • uses the platform-level configuration only when permitted;
  • validates structured responses;
  • preserves the user’s existing content when an error occurs;
  • records the model, operation, request, response, and available usage data.

During the implementation, Codex connected GPT‑5.6 through this centralized layer, resolved compatibility issues between the historical and V2 architectures, adapted AI logging, and tested the complete workflow in the browser.

Challenges we ran into

The greatest challenge was modernizing a mature platform without losing its existing business knowledge, customer data boundaries, or historical functionality.

Génécia contains shared components, customer-specific configurations, and two generations of architecture. The new workflow had to remain inside the V2 boundary without disrupting existing routes or save behavior.

Every customer workspace must remain isolated. The AI integration therefore follows the same tenant-scoped rules as the rest of the platform.

We also had to bridge two generations of credential storage and database access. The OpenAI API key was correctly configured in Génécia V2, while part of the historical AI helper still relied on an older decryption and database path.

We updated the centralized credential resolver with backward-compatible fallbacks and adapted AI activity logging to work across both database connections.

Reliable structured output was another important challenge. Before inserting anything into the Blog form, the workflow validates the JSON object and every required editorial field. An incomplete or invalid response can never overwrite the user’s current work.

Accomplishments that we’re proud of

  • We accelerated the migration of a mature business platform with Codex as our primary engineering partner.
  • We preserved historical business rules and compatibility with existing customer data.
  • We created a modular foundation for Génécia V2.
  • We rebuilt the page-editing experience with Codex.
  • GPT‑5.6 supported the multilingual adaptation of the interface.
  • We deployed a working GPT‑5.6 editorial assistant during Build Week.
  • The assistant is integrated into the normal editorial workflow rather than a separate chatbot.
  • Existing tenant boundaries and application behavior are preserved.
  • OpenAI credentials and AI activity remain centralized and traceable.
  • Human review remains mandatory before saving or publishing.
  • The complete deployed workflow was successfully tested from instruction to generated article.

What we learned

Codex is particularly valuable when applied to an existing and complex application.

Its contribution was not limited to generating code. It helped us explore the architecture, understand dependencies, preserve conventions, identify compatibility problems, and make faster engineering decisions.

We also learned that useful AI does not need to replace existing tools or user decisions.

GPT‑5.6 becomes more valuable when it performs a clearly defined action inside a workflow the user already understands. The result is immediately usable, but control remains with the user.

Successful AI integration depends as much on safeguards as generation quality: validating responses, preserving existing work, isolating tenant data, centralizing credentials, and keeping the final decision in human hands.

What’s next for Génécia

The next step is to extend this controlled AI pattern across additional Génécia modules:

  • customer follow-up assistance;
  • email campaign preparation;
  • website page writing;
  • content adaptation for newsletters and social media;
  • sales and administrative document preparation;
  • operational summaries;
  • context-aware recommendations;
  • repetitive-task automation.

Each integration will follow the same principles:

  • explicit user intent;
  • authorized business context;
  • structured and validated results;
  • human approval;
  • tenant isolation;
  • traceable AI activity.

Our objective is not to create an autonomous system that acts instead of the business owner. It is to build an intelligent working environment that understands the task, prepares the work, and leaves the final decision where it belongs: with the user.

Built With

Share this project:

Updates