ONTONYM

Give your AI a memory.

One source of truth from everything your business knows—no re-explaining, no guessing, no hallucinating.

You ask“When did we promise the Acme order, and at what price?”

Across your email, docs & chat

“…June 30 delivery in the original quote…”

“…July 8 + 5% discount agreed in a later email thread…”

conflicting · different points in time

Ontonym

July 8 · 5% discount · current

1 superseded quote · 1 open follow-up

owner, due date, and source email attached

operational state

The offer

You already explained it—in an email, a quote, a meeting. Ontonym remembers, so you never have to explain it to AI again.

Every chat with an AI tool starts from zero: paste the context, correct the guesses, repeat next week. Ontonym continuously reconciles your email, docs, chat, and calendar into live business state—one shared memory your whole team and every AI tool starts from. Correct it once, and it's current for everyone.

current statecommitmentschange impactpermissions
Why RAG isn't enough

A search index answers questions. A memory holds what's true.

RAG re-reads your documents from scratch on every question. It can't be corrected, can't show each person only what they may see, and can't tell you what changed—because an index is not a record.

You ask“What's our current price for Acme?”

RAG

Retrieval returns every price mention—old quotes, new quotes, drafts—and leaves the model to re-adjudicate the whole history from scratch, on every question, for every person, in every tool.

re-derived every time

Ontonym

One current price, adjudicated once when the evidence arrived. The superseded quotes are still there—marked superseded, with the email that replaced them.

resolved once · sourced

You ask“Which promises to customers are still open?”

RAG

Returns the passages that rank highest—a sample, not a set. And when the answer is wrong, there is nothing to fix: you can't correct an index, only pile more text on top.

top-k ≠ all · uncorrectable

Ontonym

The complete, counted list—7 open, 2 overdue—each with an owner, a due date, and the message where it was made. A wrong record gets corrected once, for everyone.

counted · correctable

You ask“What changed since last week?”

RAG

An index holds passages, not state. The only way to answer is to re-read everything and reconstruct the timeline from scratch—every time anyone asks.

no state · no diff

Ontonym

A diff of your business: the delivery that moved, the price that changed, and the records and people each change affects.

live state

Ontonym uses retrieval too—as evidence under the record, not as the record. Facts are reconciled against history once, when they arrive; permissioned per person; corrected in one place for everyone; and kept current. That is the part no index can do.

01Current state

Know what is true now, not just what was written.

Ontonym resolves new evidence against history. Current facts stay distinct from stale quotes, disputed claims, and superseded dates—so AI doesn't treat every email in your inbox as equally true.

02Commitments

Turn promises into owned, dated work.

A promise made in a call, email, or chat thread becomes a durable commitment with an owner, due date, status, and source. When it slips, you see it before the customer does.

03Change impact

When one fact changes, see what it affects.

A delayed delivery can invalidate calendar events, customer promises, invoices, and follow-up tasks. Ontonym traces the affected records and owners before stale context becomes another mistake.

04Permissions

Useful context, without privacy laundering.

Source permissions travel with the facts extracted from them. AI can reason over your whole business while only showing each person—employee, accountant, contractor—what they're allowed to see.

Custom memories

Start with a sample, build your own memory.

Your private memory becomes a live model of the people, customers, systems, decisions, and obligations your company runs on. Browse a public sample, then connect the sources your team and agents already use.

Connected systems

Read every system you already run.

Ontonym connects to the systems your company lives in — email, docs, calendar, scheduling, reminders — and turns each into structured, sourced records. It keeps the link back to the source, never a copy of the content, and flags when a record needs another look.

Living intelligence

Data changes. The graph shows what changed.

Owners change, commitments slip, and yesterday's plan survives in calendars and decks. Ontonym keeps sources attached, surfaces conflicts, and shows which customers, documents, tasks, and people a change affects.

80%+of strong signals extracted into structured records
Refreshautomated
Reviewhuman
Approvehuman
Monitorongoing
Built for small teams

No data team. No six-month rollout.

Connect the email, docs, and calendar you already use and Ontonym starts building your business model the same afternoon. Everyone works from the same memory—each person seeing only what they're allowed to see—and growing teams can move to private-cloud or on-premise later.

What you can ask

Questions AI should never get wrong about your business. Counted, cited, current.

Ask across everything you've connected — email, docs, calendar, scheduling, reminders. Get the answer, the full set behind it, and the records that still need work. No context pasting, no re-explaining.

ask → answer
What did we promise this customer?3, with sources
Which follow-ups are overdue?2 past due
What changed while I was out?1 date moved + 1 conflict

Explain your business once. AI stays current forever.

Ontonym keeps a live, permission-aware model of the customers, prices, promises, and plans your business runs on—so every AI answer, for every person on your team, starts from today's truth, not last month's documents.