LLMs in the enterprise: how to use them safely and efficiently. The ProxyPT case
Generative AI is on everyone’s lips: it promises efficiency, creativity, automation.
But as soon as it moves from experimentation to the real world of companies, some fundamental questions come up:
- How do you ensure security and regulatory compliance when using generative models?
- How do you prevent sensitive data from leaking through prompts?
- How do you manage the costs of commercial and open source models?
- How do you integrate AI with existing processes and systems?
To answer these challenges, at Interlogica we developed ProxyPT: an LLM proxy for the business use of generative AI, designed to provide scalability, control and operational intelligence.
What ProxyPT is and why it matters for companies
ProxyPT is a platform that acts as a centralised gateway for access to Large Language Models (LLMs), allowing company teams to use generative AI securely and compliantly, without depending on external solutions.
Thanks to its modular architecture based on open source technologies, ProxyPT:
- connects to different LLMs;
- centralises access management (roles, SSO, permissions);
- logs every interaction, ensuring auditability;
- collects metrics to understand how AI is used in the company.
It is a lean solution, designed to be installed on-premise or in a private cloud, adaptable to different contexts and IT policies.
Security, compliance and cost control
Why should a company use a proxy like ProxyPT instead of relying on a consumer service?
1. Cost optimisation
Choosing the most suitable model for each use case, while tracking AI usage analytically, avoids waste and maximises the return on investment.
2. Data protection
ProxyPT manages access and identities (SSO), keeping control over roles and permissions. The traceability of interactions ensures compliance with regulations such as the GDPR.
3. Simplified access
Through a single entry point connected to the company system, employees can access LLMs without individual accounts. Subscriptions and configurations are managed centrally.
4. Flexibility and governance
Models can be added or replaced without any impact on end users, and custom usage rules can be defined by role, activity or authorisation level.
A guided, scalable approach, ready to use
One of ProxyPT’s strengths is its ability to adapt quickly to the existing infrastructure, avoiding long consulting cycles or complex integrations.
The adoption path is simple: initial analysis, installation (cloud or on-premise), custom configuration and targeted training. In a few weeks you can move from experimentation to operational use.
ProxyPT does not only protect and optimise: it lets you really understand how AI is used in the company. A dedicated dashboard monitors users, purposes, costs and results. This supports strategic decisions based on concrete data, avoiding improvised approaches.
The platform is designed to evolve: future features such as querying company documents or building an internal knowledge base will extend its operational potential.
Benefits for management and IT strategy
Those who lead digital transformation need tools to measure, assess and decide.
ProxyPT thus becomes a true strategic radar:
- it identifies the teams most active in using AI;
- it highlights established use cases and those to be reviewed;
- it flags new automation opportunities;
- it helps build an AI culture based on data, not hype.
A solution created by people who innovate every day
ProxyPT comes from a cross-functional Interlogica team: software engineers, cloud architects, IT security experts and digital transformation consultants.
It was designed to answer our clients’ real questions:
“How do I avoid losing control?”, “How do we protect our data?”, “How do I explain this to my CISO?”.
Today ProxyPT is also a strategic asset for Interlogica: a concrete example of an aware, modular and secure approach to AI.