AI that knows your institution because it lives inside it. Witty Boxes by EDF deploys a proprietary ecosystem on your own servers: the right context for every decision, under your institution's full control.
These are the most common friction points when evaluating AI for educational institutions. This is how Witty Boxes resolves them.
External AI exposes institutional data. You lose control over who accesses it and where it is stored.
100% local processing. No external clouds or third parties with access to your sensitive information.
Implementing AI requires months, specialised teams and a complexity most institutions cannot handle internally.
We handle all the technical complexity. Your team uses the system from day one.
AI makes decisions without an auditable trail. Impossible to know what it responded, which sources it used or who accessed what.
Every response is traceable and auditable. Immutable record of sources, decisions and access.
Without real metrics, you can't know whether AI adds value or whether the technology budget is well spent.
Real-time Grafana dashboards: usage, impact and where to optimise, all from your own infrastructure.
Generic models don't know your institutional context and generate inaccurate responses or hallucinations.
The AI learns from your institution's official knowledge. Precise, always verifiable responses.
A proprietary AI ecosystem that adapts to every area of your institution, without jargon or complexity.
An AI assistant that answers students and faculty using your institution's official knowledge, available on web, WhatsApp and Telegram.
Agents that handle admin processes, admissions and support without manual work, freeing up your team's time.
Clear dashboards showing how AI is used and the real impact it has on your institution.
Connected to your LMS and current systems from day one, with no forced migrations.
Beyond our standard catalogue, we create and adapt specialised, custom agents for your institution's specific needs.
View agent catalogueEach module of the Witty Boxes operates in isolation and integrated simultaneously, always within your institutional security perimeter.
Every student query leaves your institution. Generic AI models process academic data on servers you cannot audit, control or shut down.
AI that truly protects an institution lives inside it — not in a cloud managed by someone else.
Academic data confined to your infrastructure. Native GDPR compliance. No variable costs or external dependencies.
ChatGPT doesn't know your academic regulations or what your committee decided last week. Without institutional context, AI confidently makes things up.
The right context transforms a generic model into an expert on your institution.
Answers anchored in your official documents. Verifiable, traceable and updated on your schedule.
LMS, admissions, support, assessments — four systems, four profiles of the same student, no complete picture.
A single student profile turns isolated data into actionable AI for the whole institution.
Full student context in every interaction. Retention decisions based on reality, not assumptions.
A single AI cannot handle regulations, management data and performance metrics simultaneously. Real complexity demands specialisation.
Real AI is not a model — it's a system of specialised agents collaborating with shared context.
Every query reaches the right agent. Validated, traceable flow with no context loss between modules.
Dropout starts weeks before the student disappears. By the time traditional methods detect it, it's already too late.
Digital academic behaviour predicts dropout weeks in advance. Time is the advantage.
Automatic alerts before the point of no return. Intervention driven by real data, not intuition.
Salesforce, HubSpot or Pipedrive: every admissions lead lives on servers owned by companies whose interests are not yours.
The admissions pipeline is as sensitive as academic records. It deserves exactly the same protection.
Complete admissions funnel, private and audited, from the first contact to enrolment.
Witty Boxes connects the three critical environments of any modern educational institution under a single On-Premise infrastructure.
Tools students already know, powered by proprietary AI that learns from their academic behaviour without exposing their data.
No public clouds, no external dependencies. Your IT team installs it once and the entire campus benefits securely and privately.
LMS, Flashcards and assessment tools capture pedagogical behaviour. Every interaction generates high-value learning data processed locally, never leaving the institutional perimeter.
Autonomous agents and a private educational CRM automate administrative processes and the enrolment funnel without exposing any data to external services.
All campus activity feeds a local central infrastructure. Raw data is transformed into Business Indicators and Education Indicators in real time, building your institution's Knowledge Heart.
Witty Boxes
When is the Operating Systems exam?
Witty Boxes AI doesn't just live on your servers. It learns from your institutional knowledge and operates under the trust standards your institution deserves.
A breach in a cloud you don't control can trigger multi-million GDPR fines. And your students' trust, once broken, doesn't come back.
Data sovereignty is not an add-on. It is the starting point for any responsible AI deployment.
No data leaves the perimeter. Native GDPR compliance and full institutional autonomy.
OpenAI, Anthropic or Google can triple prices, restrict usage or shut down. Your institution cannot depend on decisions made by others.
AI running on your servers has no usage bills and no external dependencies that can fail.
Predictable budget. No billing surprises. Full freedom to upgrade or migrate whenever you choose.
Every document you send to an external AI can be used to train models sold to other institutions. Your knowledge funds the competition.
Your documents, regulations and procedures represent years of accumulated knowledge. They must not leave the perimeter.
AI learns from your documents locally. Precise answers without any content leaving your infrastructure.
In multi-tenant infrastructure, a misconfiguration can expose your institution's data to another organisation. Resource sharing is not harmless.
Every institution deserves a completely independent environment with no data mixing or cross-contamination.
Dedicated infrastructure per institution. Zero risk of data leakage between organisations.
Without granular access control, a student can view another's grades or an administrator can access management reports. Unauthorised access is not a theoretical risk.
Trust is built by defining exactly who can access what, when and from where.
Enterprise-grade roles and permissions. Each profile accesses only what they are entitled to.
Without traceability, AI is a black box: you don't know what it replied, which sources it used, or who accessed the system.
Institutional trust in AI requires every decision to be observable, documentable and reversible.
Immutable log of every response, source and access. Complete audit of all system activity.
Three phases to transform your institution without disruption or risk
We install the entire platform on-premise on your servers. No external internet access, no data in the cloud. Your IT team validates every step.
We integrate Moodle, Canvas or Blackboard and load institutional knowledge into the RAG engine. Teachers upload materials and the system learns from your context.
From day one, students and teachers access the assistant. Leadership sees real metrics on Grafana. The institution operates with sovereign AI.
Three plans differentiated by your institution's size and complexity. Request a tailored quote based on your infrastructure and needs.
Vocational centres, schools and academies
Universities and university centres
Large universities and educational consortia
* Personalised quote based on infrastructure and needs. Implementation and migration included. Discounts available for public consortia.
We answer the most common questions from CIOs, rectors and IT managers before evaluating Witty Boxes.
On-Premise means the entire AI infrastructure —models, data, embeddings and agents— is installed and runs on your institution's own servers. No data ever leaves the university perimeter. This is critical for GDPR compliance, protecting intellectual property and avoiding dependency on external vendors who can change their prices or policies at any time.
The process has three phases: an audit of your current infrastructure, configuration and integration with your LMS, and training and go-live with your team. We adapt to your current infrastructure —bare-metal servers, private cloud (VMware, Proxmox) or dedicated hardware— and size the solution to your institution's real needs.
The RAG pipeline has three phases: (1) Docling processes your documents (PDF, DOCX, HTML, PPTX) extracting text, tables and images with advanced OCR. (2) Chunks are indexed with hybrid BM25, combining lexical and vector semantic search in Qdrant. (3) On each query, the orchestrator retrieves the most relevant chunks and sends them to the local LLM as context. Everything happens inside your server.
No. Foundation models (LLMs) are installed locally in air-gapped mode: no internet access, no telemetry, no calls to external APIs. Model updates are done manually via signed and verified packages before installation. A fully network-isolated environment can be configured if the institution's security policy requires it.
The design is GDPR-by-default: data never leaves the institutional server, there are no external sub-processors, access logs are immutable and auditable, and the system includes consent management and right-to-erasure tools. We provide technical documentation ready for DPOs. Grafana Compliance dashboards generate automatic compliance reports.
Yes. We have native connectors for Moodle (REST API + webhooks), Canvas (LTI 1.3 + API), Blackboard and SCORM. The two-way integration allows the system to access courses, grades, events and student profiles in real time, and for agents to return relevant information directly within the LMS environment.
Yes. We offer a technical pilot at no cost and with no commitment, in a scoped environment (one faculty or department). The pilot includes RAG deployment with up to 500 documents, 1 specialised agent and full access to Grafana dashboards. At the end, we deliver a feasibility report with recommendations for full deployment.
All data —documents, embeddings, conversations, configurations— resides exclusively on your servers. At contract end, we provide a clean uninstall script and a data export guide in standard formats. We retain no copy and have no remote access once the business relationship concludes.
Obtén acceso a toda la información de Witty Boxes en un documento completo y presentable.
Complete the form and we will audit the On-Premise feasibility for your current infrastructure. Guaranteed response within 24 hours.