For Norwegian and European SMBs and their IT partners

Connect every system. Put any agent on top.

Agents are only as useful as the systems they can reach, and connecting those systems has been the bottleneck. foundAItion lifts it: new connectors are built in days, not projects, and every system becomes one ontology of objects, links and actions that any agent can work on, with your approval before every write. It is built to run inside the company’s own Azure, on your choice of AI model.

See the demo How your data flows
How it connectsBuilt to run in your Azure · EU
  1. 01 · Your systems

    Where the business already lives

    Nothing is migrated away.

    • CRM
    • Email
    • HR
    • Files
    • Time
    • Accounting
  2. 02 · Connectors

    MCP, or built by the API builder

    A system with an MCP server is connected as it is. For one with a REST API, the API builder writes the connector.

    • MCP
    • REST
    • API builder
  3. 03 · Ontology

    One model of the business

    Every connected system becomes objects, links and actions.

    • Objects
    • Links
    • Actions
  4. 04 · Agents

    Ours, or ones you build

    They work across every connected system, on your choice of AI model.

    • Our agents
    • Your agents
  5. 05 · You approve

    Every write waits for a yes

    An agent proposes a change. A person approves it before anything is written.

    ApproveReject

  6. 06 · Dashboards and actions

    Work gets done

    Dashboards you build from widgets, and approved actions written back to your systems.

    • Dashboards
    • Actions
Stored
Built to run in your own Azure tenant, in an EU region.
Leaves
Pseudonymised prompts only, to a model your policy allows.
Writes
A proposal first. A person approves every write to your systems.

The agentic future

Integrations were the bottleneck. We lift it.

Agents can already read, reason and propose. What holds them back is reach: the systems a business runs on.

  1. 01

    Agents need reach

    An agent is only as useful as the systems it can reach. Real work spans the CRM, the inbox, HR, files, time sheets and the books at once.

  2. 02

    Integrations were the bottleneck

    Every new system used to mean an integration project: built by hand, then maintained. So most agents stayed inside one tool.

  3. 03

    The bottleneck is lifted

    A system with an MCP server is connected as it is. For one with a REST API, the API builder writes the connector. New connectors are built in days, not projects.

  4. 04

    Agents on everything, safely

    Every connected system lands in the same ontology, so every agent and dashboard can use it at once, and every write still waits for a person to approve it.

How foundAItion is built to handle your data

Built so your data stays in your tenant, and names never reach a model.

This is how foundAItion is built to handle your data. It is built to be installed in your own Azure tenant, in an EU/EEA region you choose, with EU-hosted, open-weight models by default and the anonymiser in front of any model. Names, national ids, emails and phone numbers become pseudonyms before a model sees a request, and a call still carrying a real one is refused. A model outside the EU data zone is refused unless you allow its region. Your records, the model of your business, the pseudonym map and the audit log stay in your tenant. Nothing is written back to your systems until a person says yes.

Public demoThe public demo is not the product: it runs on a US-region host with fictional data about a made-up company, and its Ask bar may use a US model. Please do not type personal data there.

How foundAItion is built to handle your data, in ten steps across three zones: your systems, foundAItion inside your own tenant, and an optional model outside the EU. Select any step for its data, retention, legal basis and who can act.

Your systems

They stay where they are, and stay the source of truth.

  • Accounting
  • CRM
  • HR
  • Email
  • Files
  • Time
From step 01 to step 02, read: carries personal data and business records.

foundAItion, inside your own tenant

Built to run in your Azure tenant, in an EU/EEA region you choose.

From step 02 to step 03, map: carries personal data and business records.

From step 03 to step 04, role-filtered: carries personal data and business records.

From step 04 to step 05, ask: carries personal data and business records, and the answer comes back.

Asking AIThe anonymiser sits in front of any model. A call carrying a real name is refused.

On by default · set per model

From step 05 to step 06, one call: carries pseudonymised, and the answer comes back.

From step 04 to step 08, propose: carries personal data and business records.

From step 08 to step 09, review: carries personal data and business records.

Four-eyes for sensitive actions

From step 09 to step 10, commit: carries personal data and business records.

From step 10 to step 01, write back: carries personal data and business records.

From step 05 to step 07, only if allowed: carries pseudonymised, and the answer comes back. Only if you allow it.

Optional: outside model

Refused unless your policy allows its region.

Off unless you allow it

Key

  • Personal data. Names, national ids, emails, phone numbers.
  • Business records. Invoices, orders, time entries, documents.
  • Pseudonymised. Names swapped for codes like P-qbvfhz.
  • Aggregated. Totals and trends on dashboards and reports.
  • Solid arrow. Stays inside your tenant.
  • Dashed arrow. Optional; leaves the EU data zone only if you allow it.
  • Striped top. The gate in front of every model.

For your data-protection officer: our DPIA template, the data processing agreement and the sub-processor list are on the Trust page, next to the full data-flow description.

Connectors

Your system is not on the list? We build the connector.

Built so far

  • Microsoft 365Email, calendar, peopleMCP

    People, calendar events and mail metadata. Read only.

  • TripletexAccountingREST

    Accounts, customers, suppliers and vouchers, plus one write that waits for approval.

A new connector

  • Small

    The system already has an MCP server.

    One-time build
    Price on request
    Monthly connector fee
    Price on request
  • Medium

    The system has a documented REST API.

    One-time build
    Price on request
    Monthly connector fee
    Price on request
  • Large

    The system's API is undocumented or complex.

    One-time build
    Price on request
    Monthly connector fee
    Price on request

PaybackThe client who commissions a connector earns 20% of what later clients pay for it, as credit on their own bill, for 3 years or until it reaches 2x their build fee.

Try the live demo

Demo

Try it on an example company.

The live demo runs on Acme Inc, a made-up company with made-up people. Nothing you do there touches a real system.

About us

We learned this where the rules are strict.

Before foundAItion, we built FinControl, a system for an accountancy bureau. An accountancy bureau is bound by the bookkeeping law, the GDPR, the need for a full audit trail and rules on where client data may be kept, and FinControl was built to follow them.

That work proved we can connect real systems safely, under strict rules.

foundAItion takes the same method to any business and any system: CRM, email, HR, files, time tracking, and accounting as one of many.

Trust

Built for the GDPR from day one.

The documents your DPO or lawyer will ask for are published on our trust page.