Ulpian Enterprise
Ulpian in your environment,with your architecture and your rules.
Enterprise is the offer for organisations that need a dedicated environment, private AI, integrations with their systems or their own development. It applies above 150 users, and earlier when professional secrecy, demanding clients or your IT department require it. Every project is priced on quotation, based on your actual scope.
What you will see in Ulpian
Fictitious matter, sample data.
The path of a request
What happens when someone at the firm asks the AI a question. Switch between a local model and an external provider to see what changes.
03 Your keys, your provider and your processing region, or a local model that stays in your infrastructure.
With an external provider, identifiers are pseudonymised before anything leaves, to the provider and region you chose.
Claude, ChatGPT or your legal AI tool, connected to your matters with your rights
Ulpian supports MCP (Model Context Protocol), the open standard that lets an AI assistant connect to other software. Claude, ChatGPT or any compatible legal AI tool connects to the firm through a personal token: it sees what you see, and its proposals arrive in the same decision inbox. Choose an assistant.
Claude proposes 40 minutes on the Acme matter.
The token can be revoked at any time from the account settings.
What slows large legal projects down
Large organisations do not need one more tool. They need legal work to fit into an existing architecture, under the control of their IT department.
- 01
Data spread across many systems
ERP, document management, billing, data warehouses and business tools each hold part of the matter, with no link between them.
- 02
AI that IT cannot approve
Models hosted outside your environment, with no log and no control over the processing region, which your security team rightly refuses.
- 03
Rules specific to each entity
Each subsidiary, department and jurisdiction has its own processes, permissions and templates, which the platform must follow.
- 04
Projects that get stuck
A specification that is too broad, no first deliverable scope, and months without visible results for the teams.
What an Enterprise project includes
The quotation separates one-off services, annual licences and maintenance of the agreed use cases. Each element has its scope, deliverables and conditions.
Dedicated environment
A dedicated deployment in your own cloud account (Azure, AWS, Google Cloud) or in your data centre, or a dedicated instance in the European Union operated by Ulpian. Infrastructure requirements are quoted separately.
Setup and configuration
Entities, organisation, roles, permissions and ethical walls, processes, templates, time categories and fee scales, configured with your teams.
Dedicated modules and custom development
Features specific to your organisation, designed with you and delivered against agreed acceptance criteria.
Integrations and custom backend
ERP, accounting, banking, registers, document management, directory and data warehouses such as Snowflake, Databricks, BigQuery or the one you already use, through an API or a dedicated connector.
Private AI, under your rules
The model of your choice, dedicated or installed on your premises, with network isolation, retention rules and evaluations on your matters, with no training required. As an option, adaptation to your practice when a precise, measured objective justifies it.
Licences and maintenance
Annual licences per user, maintenance of the agreed use cases, a named contact and a product review every quarter.
How an Enterprise project runs
- Step 0101
Scoping
A workshop to set the scope, the IT constraints and the first use cases.
- Step 0202
Architecture
Hosting, identities, data flows and choice of AI model, designed by our team, our partners or together, and validated with your IT department.
- Step 0303
Configuration and integrations
The platform is configured and connected to your systems for a first scope, by our team, our partners or together, depending on the project.
- Step 0404
Pilot
One team works in production against criteria agreed in advance, and gaps are corrected.
- Step 0505
Rollout and maintenance
The rest of the organisation follows in stages, with maintenance of the use cases and a review every quarter.
The guarantees your IT team will ask for
Isolation and ethical walls
Each organisation has its own boundary, and a walled-off matter stays invisible to people and agents without access, including in AI answers.
Exportable audit log
Sign-ins, changes, AI runs and approvals, exportable for authorities, clients and the data protection officer.
AI under your governance
Your own keys or a local model, an enforced processing region, identifiers pseudonymised before any external call and every call logged.
Certifications and single sign-on
The path towards ISO 27001 and SOC 2 Type II is set out in the security brief. SAML or OIDC single sign-on is planned for Enterprise, with a timetable confirmed in writing.
Private AI: the model of your choice, under your rules
Private AI can be deployed without any training: a dedicated or local model, your data, your rules and evaluations on your own matters. Adapting the model is an option on top, when it answers a precise, measured objective.
When we adapt a model, we start from an existing base model. We do not present any model as built from scratch.
The model of your choice
No training required: frontier models from the major providers, or open models run in a dedicated environment or on your infrastructure. You choose the model per task and can switch without changing platform.
Your data stays yours
Your matters are never used to train third-party models. Network isolation, retention rules and a record of every call, according to your policy.
Evaluations drawn from your practice
With your teams, we build evaluation sets from your matters and your lawyers' corrections. They measure what the model gets right and what it misses in your real work.
Optional: tuning and reinforcement
When a precise objective justifies it, we start from an open model, tune it on your examples, then train it by reinforcement on tasks drawn from your work, scored against criteria you approve. Each version is compared with the base model on your evaluations before any rollout.
A choice to measure
Tuning and reinforcement learning are not useful in every situation. An adapted model can be overtaken by the next general model: we only propose it for a precise objective, measured on your evaluations, and plan for the adaptation to be redone on a newer base.
An adapted model that belongs to you
If a model is adapted, it is yours, within the terms of the base model's licence, and is not shared with any other organisation. We show you the name of the base model, the data used and the evaluation results, before and after.
Services on quotation
Beyond the standard start, we can adapt Ulpian to your processes, your templates and your IT constraints.
Each service is priced separately.
Firm configuration
Organisation, roles, workflows, templates, time categories and fee scales.
Agent configuration
Agent missions, limits and approval steps by a person at the firm.
Private AI
A dedicated or local model, network isolation and retention rules adapted to the firm.
Integration with your systems
Accounting, banking and official registers, through an API or a dedicated connector.
Maintenance of custom use cases
Your automations evolve when regulations or your organisation change.
Training
Hands-on workshops on the platform, matter management and the use of AI in the firm.
Advisory engagement
A few weeks to set your digital priorities and a roadmap.
Want to go further with AI?
Cyberesia SA, the company that develops Ulpian, also builds Cyclones.ai: agents that integrate with your tools, under human supervision and with respect for confidentiality. Cyclones.ai automates tasks and processes with AI agents, at the scale that suits you, without locking you into closed technology.
Cyclones tools
Multi-Agent Intelligence Hub
Orchestrate several AI agents with intelligent task distribution and context-aware collaboration.
Operations Hub
A single platform for intelligent meeting management, automated scheduling and process coordination.
Process-Clone Factory
Turn repetitive manual tasks into automated workflows. Clone and automate your processes at scale.
AI Call Center
An AI voice agent available around the clock that handles inbound and outbound calls in natural conversation.
Productivity Suite
A complete set of AI productivity tools: intelligent text editing, summarisation and translation.
Integration Fabric
Connectors for ERP, CRM and enterprise systems, with real-time data synchronisation and workflow automation.
Sovereign LLM Vault
On-premises LLM hosting with no external dependencies, for secure AI processing and data sovereignty.
Cyber Sentinel
Real-time threat monitoring, security compliance tracking and automated incident response.
Enterprise AI services
Custom agentic AI systems
Tailored solutions: autonomous agents adapted to your infrastructure and your priorities.
Private AI training and deployment
Custom AI models, data sovereignty and suitable infrastructure, in the cloud or on premises.
AI and software development
Tailored solutions from design to deployment, software and pipelines included.
AI outsourcing
Dedicated AI experts working as an extension of your team.
Third-party AI integration
Connecting AI solutions from external providers to your tools, data and processes.
AI consulting
Strategy and organisational consulting: roadmap, governance and implementation.
AI cybersecurity
End-to-end protection for your AI infrastructure and data.
AI observability
Advanced monitoring and diagnostics for your AI systems in production.
Initial AI audit
Identify opportunities and build a clear roadmap before you scale up.
AI workshop
Prototyping and acceleration, from zero to a working use case (half a day to several days).
Partner network
For specialised AI infrastructure needs, we work with a network of trusted partners to deliver complete enterprise-grade solutions. Information systems architecture, digital transformation, integrations and enterprise AI services are delivered by our team, by our partners or together, depending on the project. Hardware and data-centre capacity come from infrastructure partners.
AI-related cybersecurity
End-to-end protection for your AI infrastructure and data.
AI observability
Monitoring and diagnostics for your AI systems.
Infrastructure services
Robust foundations for your AI deployments.
Hardware solutions
Specialised computing resources for AI workloads.
Data protection
Secure and compliant data practices for AI systems.
Who it is for
- 01
Large firms
Several departments, several offices and institutional clients who require a dedicated environment.
- 02
Group legal departments
Entities in several countries, outside firms to manage and enterprise systems to connect.
- 03
Public sector and justice
Administrations, courts and prosecution services, with an installation on their own infrastructure.
- 04
Firms with specific requirements
A small team, but professional secrecy or clients that require private AI and an isolated environment.
Questions about Enterprise
See pricingHow much does Enterprise cost?
Enterprise is priced on quotation. The quotation separates setup, any development and integrations, annual licences per user, maintenance of the agreed use cases and, where applicable, private infrastructure.
Do we need at least 150 users?
No. Enterprise applies above 150 users, but a smaller organisation can choose it as soon as it needs a dedicated environment, private AI, integrations or its own development.
Can you connect Ulpian to our systems?
Yes, through an API or a dedicated connector: ERP, accounting, banking, registers, document management, directory or data warehouse. Each integration is quoted and delivered with its acceptance criteria.
Who designs the architecture and runs the project?
Depending on the project, our team, our partners or both together. This applies to information systems architecture, digital transformation, integrations and enterprise AI services: we can carry them out ourselves and already work with partners on these subjects. Hardware and data-centre capacity are provided by infrastructure partners.
Where is our environment hosted?
With Enterprise: in your own cloud account, on your own infrastructure, or in a dedicated instance in the European Union. The standard platform is hosted in Switzerland by a Swiss provider, out of reach of the US CLOUD Act.
Is our data used to train models?
No. Your data is never used to train a third-party provider's models. It is used to adapt a model only if you ask for it: that private model remains yours, stays in your environment and is not shared with any other organisation.
Do you have your own AI model?
We do not claim to have built a model from scratch. When an adaptation is justified, we start from an existing base model, which we name, and show you the evaluation results before and after. We advise you to ask for the same transparency from any vendor that claims to have its own model.
A first conversation about your scope
Describe your organisation, your systems and your constraints. We come back to you with a proposed architecture and a first scope with its price.