PRODUCT

The governed record of what an organization decides with AI

What INTENTIA+ does today, what is in progress and what it doesn't do yet · September 2026

AI has made producing fast. The hard part now is knowing later what was decided, why, with what information and who approved it. Conversations with an assistant get lost, and decisions end up spread across chats, documents and people's memory.

INTENTIA+ keeps and governs that work: what the AI was asked, what it read, which rules applied, what it proposed and what a person decided. With whichever AI each organization chooses, and with its own data.

It is not a chatbot, an assistant with memory or a document search tool. It is the layer where work done with AI stops being a conversation and becomes a record someone can answer for.

The problem

• The why gets lost. A decision made with AI's help lives in a chat nobody will open again. Six months later it has to be rebuilt from memory, or argued all over again.
• The AI doesn't know what was already decided. In long conversations the context degrades, the assistant fills gaps with invented details, and it recommends again what was ruled out weeks ago.
• Rules live outside the work. They sit in a policy document the assistant has never seen, and nobody checks whether they are followed while the work happens.
• Nobody can show how AI was used. What it was asked, which documents it had in front of it, what it answered and what was accepted. Anyone accountable to a client, an auditor or a regulator will eventually have to answer that.

How it works

Work happens in a workspace: an isolated space per client, per project or per area. Every request to the AI is made with a declared intent, and the intent determines what gets recorded.

Nothing enters the organization's memory without that human step. It is slower than automatic memory, and that is on purpose: automatic memory piles up; signed memory can answer for itself.

The five pieces

Cognitive repository

Each workspace's persistent, governed memory: facts, decisions, documents and their versions, with who added what and when.

AICA, the cognitive agent

Greets you on your home page and tells you what is waiting for you, what is about to expire and where you left off. It proposes and prepares. It never decides or acts on its own.

Multi-source

You connect the folders where the work already lives, such as Google Drive or a server over SSH, and their content reaches the assistant. Every answer says which documents it read and why.

Multi-purpose

The same product serves a law firm, a consultancy or a software team. Each workspace has its own purpose, documents and rules, and can start from a sector starting point.

Context rules

The organization's rules are product objects, with a scope, a version and a declared consequence. They can apply everywhere, to one account or to a single workspace. They reach the assistant with every request, and the record shows which ones applied.

What sets it apart

Plenty of products give agents memory, search documents or manage fleets of agents. INTENTIA+ does not compete there. Its differences are by design.

• A person signs the memory. What the organization knows is only what someone approved it should know.
• The decision is the core object, with its authority. Not the document, not the agent. Each decision keeps what was asked, what was decided, the reasoning and its references. Its events carry a cryptographic fingerprint, so they cannot be altered without it showing.
• Automation never decides. An external tool can come in, read and prepare work. Approving, rejecting or amending is done only by a person, whatever role the key carries.
• Your AI, your bill, your data. Each organization uses its own OpenAI, Anthropic or Google account. INTENTIA+ does not resell AI or keep a fallback key: if yours fails, the turn stops and you are told.
• Every answer explains itself. It says which documents it read and why, which rules governed it and what it cost, in cents.
• Decisions expire. What nobody decides within its deadline stops waiting, and you get an email warning beforehand.

Who it is for

INTENTIA+ does not depend on the AI model, the sector or the size of the organization. The question it answers, who decided what, under which rules and with what authority, is the same for a four-person firm and for a four-hundred-person program.

• Professional firms. Law firms, accounting practices and consultancies that use AI in client work and put their signature on it. One workspace per client, with its own rules.
• Software and product teams. AI speeds up the work, and architecture and scope decisions get lost across conversations.
• Complex programs and projects. Many people, handovers, and decisions that get reopened because their reasoning was lost.
• Organizations starting to deploy agents. They need to write down what an agent may propose and what a person must approve, and keep a record of it.

There will be two ways in: on your own, with a folder and an AI key, or with our help, when the organization needs to integrate its systems and corporate identity.

Security and data

• Isolation per workspace. Each workspace is separate, and permissions are checked on every request.
• Standard identity. Sign-in with OpenID Connect. INTENTIA+ never receives or stores passwords.
• Encrypted AI keys. They are kept in a secrets vault, and the product stores only a reference.
• Your text goes only to your provider. It travels to your AI provider, under your account and your terms.
• A trail of what matters. Who created, approved, archived or changed what, and when.

INTENTIA+ does not certify compliance with any regulation. What it does is produce, as the work happens, the evidence that whoever has to prove compliance will need.

Where the product stands

Done and running

The five pieces run in a test environment reachable over the internet: user sign-up, workspaces, chat with eight intents, decisions and reviews, context rules at all three scopes, sector starting points, connected sources, usage costs, AICA, external tool access with a service key and a read-only MCP server. In Spanish and English, with more than 5,500 automated tests.

In progress

Export, usage metering, backups and a security audit.

Not yet

Corporate identity (SSO), connectors beyond folders, semantic search and billing. The product is not open to the public yet.

Where it is going

Your assistant as the way in

Today you work inside INTENTIA+. The next step is not having to: you keep using the assistant you already use, whether ChatGPT, Claude or Copilot, and INTENTIA+ sits behind it. The assistant converses and chooses; INTENTIA+ governs and keeps the record. Part of the technical piece already exists: MCP access.

Governed autonomy, for physical agents too

The question asked today about an assistant will be asked about any agent, robots included. A warehouse robot may be able to enter a zone and still not have the authority to do so today. INTENTIA+ does not move robots or replace their safety systems. Its place is the layer above: under which mandate, which rules and whose approval an agent acts, and what evidence remains. This line is at first-prototype stage and is not part of the current product. The design already allows for it: the record of who acted does not assume it was a person.

Models generate. Agents act. INTENTIA+ governs why, when and on whose authority, and remembers it.

If you want to know when it opens, write to us →
Read the manifesto →