Companies in Zaragoza face the same paradox: they have more data than ever, but the relevant information remains trapped in silos. A corporate intranet with AI is not just an internal news portal; it is an intelligent layer that connects people, processes, and data. To make it work, it is necessary to combine custom software applications with language models, automation, and a clear data strategy. This is the approach Q2BSTUDIO, a software and technology development company, applied in a recent case in Zaragoza, and it serves as a reference for any executive evaluating a similar transformation.
The company in this case belonged to the industrial services sector and had 90 employees, spread across offices, warehouse, and commercial teams. Years of organic growth and small acquisitions had left a heterogeneous infrastructure. Operations worked with an old ERP, sales used a different CRM, and key documents were distributed across shared folders, SharePoint, and email. Management was aware of the problem, but did not have an objective figure to measure the cost of disorganization.
The first analysis confirmed that critical knowledge depended on specific people. If someone was absent, answers took hours or days. A typical incident cycle included at least three emails, two meetings, and a manual search in a file system with more than 200,000 documents. In addition, lack of traceability made it impossible to know which document version was current. This situation is common in many SMEs in Zaragoza and is not solved with more storage alone.
A traditional corporate intranet merely posts links and news. In contrast, an AI-powered intranet is designed to understand questions and return contextual answers. This requires a semantic data model, a hybrid search engine, and a portal that adapts content to each user's role. The success case described in this article used exactly that architecture, avoiding unnecessary technologies and focusing on use cases with clear return.
The starting point was a two-week diagnostic. Q2BSTUDIO did not start coding but measuring. Key indicators were defined: average search time, number of duplicated documents, percentage of incidents resolved in less than 24 hours, and employee satisfaction with internal tools. This baseline allowed setting realistic goals and limiting scope to two critical processes: incident management and employee onboarding.
The solution included a single web portal and an enterprise AI solution based on RAG architecture. The system indexed documents from SharePoint, the ERP, and network folders, preserving original permissions. When a user asks, for example, what is the return procedure?, the AI searches relevant fragments, combines them with current policies, and generates a clear answer that cites the original source. This model reduces hallucinations and lets a new employee understand the business without relying on asking a colleague.
In addition, AI agents were configured for repetitive tasks. The classification agent read each new incident, categorized it, assigned priority, and suggested a response based on previous cases. Another agent extracted data from invoices and delivery notes to enter it directly into the ERP, reducing transcription errors. This is where the intranet stops being a simple search tool and becomes an operational platform.
A decisive element was integration. It was not necessary to replace the ERP or eliminate SharePoint. Q2BSTUDIO built an integration layer using APIs and n8n to connect the intranet with the CRM, corporate email, and Microsoft Teams. Approvals were processed from the portal but remained visible in the tools teams already used daily. This strategy reduces resistance to change and accelerates adoption.
The environment was hosted in AWS/Azure cloud, with redundancy and encryption at rest and in transit. From a cybersecurity perspective, role-based access policies, multi-factor authentication, and audit logs were applied. Sensitive data did not go out to public models; private endpoints and VPN tunnels were used so AI services could access on-premises systems securely. This is essential in projects handling personal data or contractual information.
For management, the intranet also became a source of business intelligence. A dashboard developed with BI/Power BI showed the evolution of resolution time, workload by department, adoption level by office, and estimated cost of time saved. With this information, management could make decisions about staffing, schedules, and training without waiting for middle managers to prepare manual reports.
Project results were measured at the end of the third month. Time spent searching for information fell by 36%. Incident management went from an average of 4 hours to 2.5 hours. Manual data re-entry fell by 41%. Operational information accuracy rose from 79% to 94%. Employee onboarding went from ten days to two. Management estimated return on investment in less than seven months, considering only recovered productive hours.
The most complex aspect was not technology but governance. A data quality committee was created with managers from each area. The AI was trained with approved versions of procedures, not obsolete files. Change control and human review were also configured for critical decisions, such as dissemination of contractual information or payment approval. Transparency about what the system does increased staff confidence.
The implementation plan was executed in phases to avoid disrupting operations. The first phase consisted of the corporate portal with semantic search and about fifty master documents. The second added classification and extraction agents. The third extended integration with the ERP and Microsoft Teams. The fourth activated dashboards and automated weekly reports. Each phase ended with validation by a pilot group of fifteen users, who provided improvements before general deployment.
Training was designed around real cases. Instead of generic sessions about how AI works, employees practiced with questions and tasks from their own department. A short guide with example queries was created, and an internal copilot was appointed in each area to resolve doubts and propose improvements. This role was key to keeping adoption above 90% after sixty days.
The main lesson is that data is as valuable as the algorithm. Without prior cleaning, the best AI offers fast but incorrect answers. The second lesson is that cybersecurity must be present from design, not at the end. The third lesson is that an AI intranet grows with the organization: early use cases are simple, but the platform must allow adding new sources and agents without rewriting code.
In short, a corporate intranet with AI in Zaragoza is viable when approached with technical rigor and business vision. Q2BSTUDIO's experience shows that the key lies in combining custom applications, language models, automation, and secure cloud infrastructure. Technology is no longer the bottleneck; the ability to integrate it into real processes is. Any company that wants to move in this direction can rely on a software development team that understands the local context and current business standards.
