Powered and governed by ARPIA.
Most companies have one of them: a chat that does not know their data, or data with no AI that acts on it. Put together, and governed, they are what makes AI worth giving to everyone.
A governed reflection of your ERP, CRM, data platforms and documents, organized by your own ontology, plus Cerebro, the memory of what your organization already decided.
Governed. Your IP stays in your workarea, with zero data retention at the model providers.
See Cerebro →Anthropic, OpenAI, Gemini, open models or your own on your own GPUs, each job on the model that does it best.
Governed. Change the model and the work stays the same. No vendor lock-in.
See model choice →The tools that put AI to work: dashboards and documents, apps your team opens, governed actions in your systems, Experts and Oria Machine.
Governed. Spend capped by area and by model, and every action recorded.
See the tools →Nothing here needs a project plan. You sign in, point Oria at something you already care about, and it works from what your organization knows rather than from what the internet knows.
No prompt technique to learn. Ask about a customer, a month, a process. Oria decides which of your systems to read and shows you the path it took to the answer.
Most assistants stop at a summary. Oria runs the governed actions your business already defined, under the same permissions your people have, and asks before the ones that matter.
Correct it once and the correction becomes institutional. The rule you explain today is the rule it applies next month, with the reason and the author attached.
Every turn is priced against a budget you set. There is no month where the bill is a surprise and no way for one heavy user to spend the whole pool quietly.
These are not four products. They are four things the same session can do, on the same governed data and the same permissions.
Ask, and it reasons over your governed data and your institutional memory. Dashboards, maps, relationship graphs and timelines come out of the conversation instead of a ticket.
It writes and runs code against your own model of the business, inside your security policy, with a review gate before anything ships.
Author the ontology itself: the nodes, the relations and the tables behind them, with permission gates on propose, apply and sample.
A governed execution environment. Code runs in its own pod with your scope and permissions, its own storage, and human approval where it counts.
A question that ends in a spreadsheet is a question you get to ask again next month. Oria builds the thing itself, connected to the data it was built from, versioned and kept with the session. Every tool below is listed, configurable and removable, so what the AI can reach stays a decision somebody made.
KPI tiles and charts laid out together from the data the session is already reasoning over.
Real PDF, Word, PowerPoint and Excel files, styled with your document design system and handed over as a download.
Kanban with per stage counts, WIP limits and aging. It writes the status back where a governed action allows it.
Timelines, Gantt schedules and calendar heatmaps, with span, gap and duration analysis.
How customers, suppliers, owners and referrals connect. Hubs, clusters and orphans read as figures.
Interactive maps on the basemaps your workarea already has, with layers, controls and popups.
What if sliders that recompute the output live from the formula behind it.
A connected app built for the job: lookups, reconciliation views, form driven query builders. Oria writes the components, ARPIA compiles them.
Visuals generated and edited in the session, dropping straight into the documents and decks it produces.
Storage that outlives the session, for the analysis and the small apps that are not ready for the organization ontology.
An Expert holds its own ontology nodes and its own tools, so the person chasing collections and the person writing a proposal do not both start from the same blank session. Set the kit once and everyone in that role opens with it.
Cerebro is where your organization writes down what it has already settled: the rules it learned the hard way, the reasons behind the calls that worked, the procedures that actually get followed. Oria answers from it and cites what it used.
Cerebro comes with the account from the entry tier. It is not an add-on.
Just over a minute: a session that already knows what the organization decided, and where the answer came from.
Almost none of them are about the model. Data privacy and security is the barrier enterprises name first, at 71%, ahead of accuracy and ahead of cost (Deloitte, 2026), and 73% of organizations spent more on AI than they had planned (FinOps Foundation, 2026). Oria was built as the answer to these four, which is why the governance is the product and the chat is the interface.
73% of organizations went over their AI budget, some by 2.4 times (FinOps Foundation, 2026). Here the spend is prepaid and enforced: the AI Wallet is a pool you load, every model carries a published rate per million tokens, and global and per-model ceilings warn before the limit and refuse the call at it. One heavy week cannot drain the pool on everyone else's behalf.
Model leadership has changed hands twice in two years, and a stack wired to one vendor has to be rebuilt each time. Oria connects Anthropic, OpenAI, Gemini, open models and your own models on your own GPUs through one governed connection. You choose which model runs which work, by role and by profile, and changing it does not change the work.
Half of technical leaders name integration with legacy systems as a blocker (CodeNinety survey of 142 CTOs, 2026), and an assistant that cannot see the ERP is a demo. Oria reads a governed reflection of your ERP, CRM, POS, data platforms and documents through your ontology, with lineage on every figure. It never writes to the live systems, and what it may reach is a decision somebody made.
Zero retention has become a gating term in enterprise AI contracts, ahead of price. ARPIA runs its own infrastructure and GPUs, isolated from the model providers, and its contracts with them are zero data retention, so what you send is not kept on their side and your data does not train models. Each workarea holds its own Cerebro, apart from every other tenant. For a higher standard, the platform deploys on-premise inside your own datacenter.
ARPIA holds SOC 2 Type II and ISO 42001, the international management standard written for AI systems, both audited by A-LIGN across the platform Oria runs on. The reports are available on request.
Everything above is infrastructure: the ontology, the budgets, the security profiles, the model routing, the isolation and the audit trail. Oria is the interface over all of it. You open a session and ask, and the governance is already applied, so nobody has to learn where any of it lives.
Budgets are enforced at the point the request is made, not reconciled at the end of the month. When a limit is reached the call is refused and the person is told why.
ARPIA is SOC 2 compliant. Live service status is published at status.arpia.ai.
Billed per user per month and prorated daily, so a seat added on the 20th costs eleven days. The AI itself is paid separately out of a wallet you size, which is why the seat price is not carrying a guess about how much anyone will use it.
For everyone who needs to ask, analyse and produce.
For the people whose tools other people end up needing.
Oria Unlimited. Give Oria to your whole organization and stop counting seats. One platform commitment covers every user, and you scale on AI Wallet consumption instead: billed on the work that actually happened, not on how many people have a login.
Contact us →Past those allowances, storage and compute are billed at the normal published rate, the same way the AI Wallet bills tokens.
Every new workarea starts with $100 in it, so the first sessions run before anybody loads a cent.
Only the seat recurs. The wallet is a prepaid pool the organization loads and draws from, spent on work that actually happened: a question answered, a document rendered, an app built. Nothing is charged for capacity nobody used, and what you load carries over. It is also why a governed seat lists at fifteen dollars instead of sixty, because the seat is not carrying a guess about anyone's usage.
You load it, you watch it, you top it up when it runs low. A quiet month does not produce a bill for a busy one.
Every model carries a rate per million tokens, input and output, and each call is costed against it. You can see which model spent what.
Global and per model budgets alert before the ceiling and refuse the call at it. One heavy week cannot drain the pool on everyone else's behalf.
Connect your own provider accounts and the tokens stay on your bill. ARPIA charges a flat $1 per million tokens for the governance that traffic passes through.
Sizing it, as a starting point: about $5 of wallet a month per active Basic seat, and about $50 per Pro seat, because Pro is the one doing the building. The calculator on arpia.ai/oria works it out for a real headcount.
What Oria builds starts as an artifact inside the conversation: a simulator, a board, a reconciliation view, a small connected app. On Basic it lives in your session and answers your question, which is often all you wanted.
On Pro you publish it. It becomes a Personal App you open without the chat, and then a Team App your colleagues open too, with a version history behind it so the change that broke something can be rolled back. That is the whole line between the two seats: Basic produces the work, Pro turns the work into a tool other people use.
The colleagues who open it do not need a Pro seat. A Basic seat runs what a Pro seat published, which is why most teams buy a few Pro seats and Basic for everyone else.
Authoring the ontology itself, writing governed actions and deploying what the whole company runs is the ARPIA Builder seat. That is a platform conversation rather than a signup, and it lives on arpia.ai/oria.
One person, one real question, one session. The organization can come later, and when it does, everything you taught Oria is already there.