Every AI on your teams, working from the current decision.
Policies, procedures, price lists and quality standards live across dozens of systems and versions. OIDA keeps track of which is current, for which office or function, and serves it to every AI assistant the organisation approves. With roles, review and audit.
Five questions your teams’ assistants cannot settle today.
These are not search questions. They are questions of state: which version, for which office or function, since when and whether anyone has decided. Here is how OIDA answers.
Illustrative examples. Choose a question and compare the answer and sources.
Which remote-work policy applies to the Milan office?
resolve_decision_statescope Italian offices
Reading the sources in the registryHR policy v6, §2 · Management announcement, Feb 2026
State resolvedIn force
Italian offices work on site three days a week.
In force · scope Italian offices · from 1 Mar 2026 · supersedes Two-day rule, HR policy v5
Three days on site per week from 1 March 2026, HR policy v6, §2, for all Italian offices. The two-day rule in policy v5 is superseded.
Sources
HRHR policy v6, §2From 1 March 2026, staff in Italian offices work on site at least three days a week.Primary
CDManagement announcement, Feb 2026Evidence
resolve_decision_statescope Lisbon branch
Reading the sources in the registryProcurement procedure v3, appendix 2 · Board resolution, Dec 2025
State resolvedIn force
For the Lisbon branch, the threshold is €2,000 until 30 June 2027. Other offices use €5,000.
In force · Scoped exception · scope Lisbon branch · from 1 Jan 2026 · until 30 Jun 2027
No. For Lisbon, the threshold is €2,000 under procurement procedure v3, appendix 2, until 30 June 2027. Other offices use €5,000.
Sources
PPProcurement procedure v3, appendix 2For the Lisbon branch, the threshold for purchases without comparing quotes is €2,000 until 30 June 2027.Primary
BRBoard resolution, Dec 2025Evidence
resolve_decision_statescope public sector
Reading the sources in the registryCommercial resolution 2026-02 · Commercial note, Mar 2026
State resolvedSuperseded
The 2025 price list has been superseded since 1 April 2026. Proposals use the 2026 price list, including for public-sector clients.
Superseded · scope public sector · superseded on 1 Apr 2026 · superseded by 2026 price list
No. Since 1 April 2026, the 2026 price list applies under commercial resolution 2026-02, including to public-sector proposals. The 2025 price list stays in history.
Sources
CRCommercial resolution 2026-02From 1 April 2026, all proposals, including those to the public sector, use the 2026 price list.Primary
NCCommercial note, Mar 2026Evidence
resolve_decision_statescope North-East logistics
Reading the sources in the registry2026 supplier register, rev. 2 · Procurement email, 28 Aug 2026
State resolvedContested
The 2026 supplier register qualifies Brenta Logistica for the North-East. A procurement email dated 28 August announces its suspension. The conflict is open.
Contested · scope North-East logistics · from 15 Jan 2026
According to the 2026 supplier register, rev. 2, yes. A procurement email dated 28 August announces its suspension. Open conflict, awaiting review: until then, the register applies.
Sources
AF2026 supplier register, rev. 2Brenta Logistica is qualified for transport in the North-East area.Primary
PEProcurement email, 28 Aug 2026Conflicting
resolve_decision_state
Looking for an approved decisionno source in the registry
State resolvedUnknown
No approved decision on this topic. Nothing has been inferred.
Unknown
The workspace registry has no approved decision on this point, and I have not inferred one. If a decision exists, an owner or admin can record it.
Ask something
The solution
A shared memory that grows as your company grows, and builds your own intelligence, for full sovereignty over your data.
Intranet, email and chat disagree. OIDA identifies the rule in force.
A person asks the organisation’s approved assistant. The answer includes the state, the office or function it applies to, and the source.
assistant · connected to OIDA
Your AI assistant · AI modelShare
How many days on site are required in the Milan office?
oida · resolve_decision_state
Resolving the remote-work topic, scope: Italian offices
Contested
Italian offices work on site three days a week.
HR policy v6, §2 · scope: Italian offices · from 1 Mar 2026 · contested while a conflict remains open
Three days a week, HR policy v6, §2, from 1 March 2026. Customer support has a four-day remote-work exception. An office manager message announces a return to two days: it is recorded as an open conflict, not approved.
Ask something
oida · leadershipOpen in OIDA
Decision registrytopic lavoro-da-remoto
In force
Italian offices work on site three days a week.HR policy v6, §2 · scope: Italian offices · from 1 Mar 2026 · contested while the conflict below remains open
Superseded
Staff work on site two days a week.HR policy v5, §2 · superseded by HR policy v6 · kept as history, never returned as current
Scoped exception
Customer support works remotely four days a week.Office agreement, 20 Mar 2026 · scope: customer support · until 31 Dec 2026 · returned only for that scope
Open conflict
“From September, we are going back to two days on site.”Office manager message, 1 Sep 2026 · non-authoritative · recorded, not applied · awaiting review
Unknown
Which rule applies to external contractors?No approved decision · nothing inferred
Every communication checked against the policies it depends on.
Before writing, the assistant asks OIDA which policies apply to that office or function. Exceptions stay within their scope. Open conflicts are flagged, never applied.
assistant · connected to OIDA
Your AI assistant · AI modelShare
Write the offer letter for a customer support employee in Milan, including the remote-work section.
Checking one policy, one exception and one open conflict
Draft for review
For customer support, the exception applies: four remote days until 31 December 2026. The general rule is three days on site. The message about returning to two days is flagged as an open conflict.
HR policy v6 applied · office agreement in scope · 1 conflict flagged for review
Letter written with four remote days, citing the office agreement, and a reference to HR policy v6 from 1 January 2027, when the exception expires. The message about returning to two days is not an approved decision: I have noted it for the reviewer.
Ask something
oida · leadershipOpen in OIDA
Decision registrytopic lavoro-da-remoto
In force
Italian offices work on site three days a week.HR policy v6, §2 · scope: Italian offices · from 1 Mar 2026 · contested while the conflict below remains open
Superseded
Staff work on site two days a week.HR policy v5, §2 · superseded by HR policy v6 · kept as history, never returned as current
Scoped exception
Customer support works remotely four days a week.Office agreement, 20 Mar 2026 · scope: customer support · until 31 Dec 2026 · returned only for that scope
Open conflict
“From September, we are going back to two days on site.”Office manager message, 1 Sep 2026 · non-authoritative · recorded, not applied · awaiting review
Unknown
Which rule applies to external contractors?No approved decision · nothing inferred
A management committee decision arrives as a candidate for review.
People record decisions from their client. OIDA stores them as proposals, with a source and excerpt, until an owner or admin decides.
assistant · connected to OIDA
Your AI assistant · AI modelShare
Record this from today’s management committee: the new Bologna office starts with two days on site for six months.
oida · ingest
Storing a candidate from the management committee, 7 Sep 2026
Proposed
The new Bologna office starts with two days on site for six months.
topic: remote work · scope: Bologna office · source: management committee, 7 Sep 2026 · awaiting an owner or admin
Recorded as a proposed decision, not as current. Nothing in the registry changes until an owner or admin approves it with a reason. It concerns the same topic as the open conflict, so the reviewer will see both.
Ask something
oida · leadershipOpen in OIDA
Decision registrytopic lavoro-da-remoto
In force
Italian offices work on site three days a week.HR policy v6, §2 · scope: Italian offices · from 1 Mar 2026 · contested while the conflict below remains open
Superseded
Staff work on site two days a week.HR policy v5, §2 · superseded by HR policy v6 · kept as history, never returned as current
Scoped exception
Customer support works remotely four days a week.Office agreement, 20 Mar 2026 · scope: customer support · until 31 Dec 2026 · returned only for that scope
Open conflict
“From September, we are going back to two days on site.”Office manager message, 1 Sep 2026 · non-authoritative · recorded, not applied · awaiting review
Unknown
Which rule applies to external contractors?No approved decision · nothing inferred
Proposed
The new Bologna office starts with two days on site for six months.Management committee, 7 Sep 2026 · candidate scoped exception · awaiting an owner or admin
Owners and admins approve, reject or supersede, with a recorded reason.
The review queue is available from the client and the web app. No candidate becomes policy on its own, and each decision stays in history with who made it and why.
assistant · connected to OIDA
Your AI assistant · AI modelShare
What is waiting for HR review on remote work?
oida · list_pending_decisions
Two pending items in oida · leadership
2 pending
Proposal: the new Bologna office starts with two days on site for six months. Open conflict: “From September, we are going back to two days on site” versus HR policy v6.
Approve, reject or supersede · reason required · owner or admin
Two items. I can open them and record the decision with its reason, but I cannot approve on behalf of the person with that role.
ApproveRejectReason required · owner or admin
Ask something
oida · leadershipOpen in OIDA
Decision registrytopic lavoro-da-remoto
In force
Italian offices work on site three days a week.HR policy v6, §2 · scope: Italian offices · from 1 Mar 2026 · contested while the conflict below remains open
Superseded
Staff work on site two days a week.HR policy v5, §2 · superseded by HR policy v6 · kept as history, never returned as current
Scoped exception
Customer support works remotely four days a week.Office agreement, 20 Mar 2026 · scope: customer support · until 31 Dec 2026 · returned only for that scope
Open conflict
“From September, we are going back to two days on site.”Office manager message, 1 Sep 2026 · non-authoritative · recorded, not applied · awaiting review
Unknown
Which rule applies to external contractors?No approved decision · nothing inferred
Proposed
The new Bologna office starts with two days on site for six months.Management committee, 7 Sep 2026 · candidate scoped exception · awaiting an owner or admin
Three moments when the registry speaks your language.
01
A new office
The person opening the office asks which policies, procedures and price lists apply there from day one. The answer includes current rules for that office, valid exceptions and their expiry dates, with sources.
02
An audit
Reviewers read the same decision as teams and their assistants: who approved it, when, why and what it replaced. History accumulates: every change adds a row. None removes one.
03
A change of supplier
A supplier’s qualification changes in a resolution, an email and a chat. OIDA serves the approved one and records the others as open conflicts until someone with the right role decides. Procurement and logistics assistants read the same state.
Who can decide, who can read, and the history of every change.
OIDA treats the decision registry as a governance record: explicit roles, reasoned writes, accumulating history and logged access.
01
Three roles
Owner, admin and member. In reviewed mode, owners and admins approve; in peer mode, members can also approve, revoke and resolve conflicts. Anyone can propose ontology changes, but only owners and admins can approve them.
02
Writes with a reason
Every approval, rejection or supersession carries the person’s name and a required reason. No candidate becomes current on its own.
03
Append-only history
Superseded decisions stay in the registry with the date they stopped applying and what replaced them. Nothing is deleted to make room for a new version.
04
Every call logged
Every assistant request to OIDA tools is logged for audit: who, when and which topic. Clients authenticate with OAuth 2.1 and PKCE. OIDA does not issue long-lived tokens.
05
The workspace is the boundary
Each organisation has its own workspace, the tenancy boundary: decisions, sources and roles do not cross it. An assistant connected to a workspace reads only that workspace.
06
Row-level security
The database enforces workspace separation at row level: every query carries the workspace and reads only its rows. Separation lives in the database as well as the application.
The same registry, in your industry’s examples.
Three pages with questions, records and moments in the language of the people doing the work.
01
Manufacturing
Drawing revisions, approved suppliers, process parameters and line-specific deviations, current for each plant.
The memory says what is in force now. An adapter changes something else: the models’ tendency to answer the way you answer. It is your own history of decisions, compressed into a file that stays yours and that you can take with you.
No data-collection project is needed: it is the registry you keep every day. Volume is not the value, signal is, context, scope, outcome and the correction when it came.
A ruling is a fact and stays in the memory. The criterion, what you weigh before deciding, what makes you review a rule, who has to approve, is the part a model can be taught.
A few matrices are trained instead of every parameter: it costs little, it is versioned and it is replaced. One adapter per function or per site, if the criteria differ.
The weights are a file of your own: you run it where you like and you take it with you. The base model stays general, the specialisation is yours and does not live inside a vendor’s service.
It is not cloning answers: it is raising the chance that the same criteria hold on similar problems, even when the person answering arrived last week. A candidate is compared with a baseline before it replaces the one in use.
Illustrative example of the same question, before and after the adapter.
Can we approve a full-time remote request for the Milan office?
Base model only
It depends on your internal policy and on the role. Usually the job, seniority and impact on the team are weighed: it is worth checking with human resources before answering.
Defers to a document it does not cite
No scope and no date
Does not close, and does not say who decides
Base model + adapter
Rule in force
the memory brings it, with its scope and effective date.
Scope
which office and which role it applies to, and whether a recorded exception exists.
Answer
yes or no, and on what conditions: the request is settled, not deferred.
Who approves
who has to sign the exception, and how long it lasts.
The same structure every time
Scope and validity always stated
Closes, and says who approves
The memory brings the rule. The adapter brings the shape of the answer. Both are needed: current state plus learned behaviour.
Today OIDA Cloud stores and does not train. The OIDA Intelligence pipeline is implemented and tested in a separate codebase, is not connected to Cloud and is off by default. If your organisation used it, training would start only from a dataset an owner had admitted under recorded rules, and a candidate would be compared with a baseline before an owner decided on promotion.
SupersededADR 001 · Historic defaultMySQL is the default database.
Your AI assistant
QuestionCan we use DynamoDB for the events service?
Toolfind_conflicts
OIDAwaiting for review
Open conflict
Two sources disagree. I have applied neither.
ADR 024 · Choice of databaseNew services use PostgreSQL.authoritative
Team chat · 14 Mar 2026We are moving new services to DynamoDB.not authoritative
ApproveRejectreason required · owner or admin
Your AI assistant
QuestionWhat have we already decided about the database?
Toolget_decision_history
OIDAnew session · same registry
Three decisions found, database topic
01PostgreSQL for new servicesIn forceADR 024 · since 1 Feb 2026
02MongoDB for legacy servicesScoped exceptionADR 011 · scope legacy
03MySQL as the defaultSupersededADR 001 · superseded by ADR 024
Architecture and sovereignty
Your decisions stay with you, even when you change AI.
Decisions stay in a registry separate from the AI model. In Cloud, data is in the database we run and each organisation has its own workspace. For enterprise integration we talk directly to your IT department and keep your data in your own infrastructure, with your own policy.
Your AI assistantscompatible, via MCP
…
Your peopleweb app
…
Who asksyour people and your tools
the decision layer
explicit rules for decision state
traceable sources and evidence
recorded authority and scope
No copy of your documentsOIDA stores the source text you submit, decisions and evidence. MCP answers return excerpts and references.
No trainingOIDA trains no model on the data in the registry. The dataset stays yours: any use starts from a decision of yours.
Your databasewhere the decisions stay
Cloudone workspace per organisation, data kept separate, Frankfurt
Enterprise integrationyour infrastructure and your policy, agreed with your IT department
Your data stays where you decideOriginals stay in their systems; OIDA retains submitted sources.
The model is the tenant. The structure is yours.
Data sovereignty starts with where data is stored and who can access it: we define those boundaries together for the pilot. AI sovereignty starts with a memory separate from the model, queryable through MCP, the protocol that connects assistants to external tools. You can change compatible assistants while keeping the same registry.
Boundaries. Where the registry lives and who may read it is your call, settled before the pilot starts.
Models. Authorised assistants query the same registry, kept separate from models.
Permissions. Access by workspace, actions by role. Applicability scope is not a reading permission.
Efficiency. Answers carry excerpts, state and references. Use the pilot to assess the context your agent needs.
Accelerated in Turin, with a scientific advisor from MIT.
OIDA is accelerated within Kakashi Venture Accelerator, an AI-native venture studio based in Turin, with senior scientific advisor Pierfrancesco Beneventano, researcher in machine learning theory at Massachusetts Institute of Technology. The method is described in a publication with public evaluation corpora.
Connect a compatible client and authorise workspace access. Your agent can query decisions, sources and states through MCP. Another authorised AI client can consult the same registry.
How to start, who approves, who can read, and what works today.
Does data stay in Europe? What does sovereignty mean here?
The Cloud registry is in Frankfurt, in the database we run, with data separation by organisation. OIDA does not train models on recorded data. Connected assistants receive excerpts and references under their respective providers’ terms. For enterprise integration we talk to your IT department: the registry stays in your infrastructure, with your own policy.
How do roles and audit work?
Three roles per workspace: owner, admin, member. Members read and propose. Admins and owners approve, reject and supersede, with a required reason kept in history. History is append-only: superseded decisions remain with their date and successor. Every assistant call to OIDA tools is logged, and the review queue is available in the web app.
Can we keep data in our own database?
For a database managed by your company, we agree on requirements and integration first. This is not a self-service Cloud configuration feature. OIDA Local covers one operator on their own machine. A multi-user on-premises deployment is not currently available.
Is OIDA an assistant or a chat?
OIDA is a decision registry your AI assistant can query. Keep working in your existing compatible client. You record the relevant sources: OIDA does not automatically listen to calls or conversations.
Who approves a decision?
By default, owners and admins approve. Member contributions stay proposed. An owner can enable peer governance, allowing members to approve decisions too. Inferences from prose stay proposed. Ontology changes always require separate approval by an owner or admin.
Does OIDA train models on our data?
OIDA Cloud stores decisions, evidence and history without training models. OIDA Intelligence is a separate pipeline, not connected to Cloud and off by default. Using it requires an owner-admitted dataset, recorded rules and evaluation before a candidate model is approved.
Which clients work today?
Claude Code, Cursor and Codex have completed a real end-to-end run. ChatGPT in developer mode, the Claude apps and GitHub Copilot in VS Code follow the vendor’s published setup path. Any client supporting Streamable HTTP and OAuth 2.1 can connect. Clients that send only a static header cannot.
How much does it cost?
OIDA Cloud is in pilot. Request a demo to agree on the use case, access and pilot terms.
Start with one decision. Check what your AI answers.
Choose a decision that changed recently: an offer, a procedure or a policy. Bring the earlier source and the approved one.
01
Create your workspace
Sign up with your email and organisation name. Confirm your email when requested and sign in to your workspace.
02
Connect your assistant
Follow the installation guide: copy OIDA’s address into a compatible assistant’s settings and authorise access.
03
Record, approve, check
Ask your assistant to propose a map of topics and scopes, then approve it. Record the decision with its source, approve it and ask what applies now.
OIDA uses the technical cookies that sign you in and keep your session open. With your consent it also records how these pages are used, to learn which content and which campaigns bring people here: pages viewed, clicks and a replay of the navigation with every field you fill in masked, through PostHog, on servers in the European Union. No advertising, no profiling.
Decision registry
