Ask in plain language. Answer from your own fields.
Grounded in your entities, your records, your documents — with citations. Sense runs your reports and dashboards, draws the chart when a chart is the answer, and drafts what does not exist yet; you approve it. On the web and in the field app.
Senseacting as Ravi K.One badge. Four ways to wear it.
Sense can look, too.
Budgets that stop overruns before the spend.
Seats, credit ledgers, per-model rate cards, and hard budgets — enforced up front, not discovered on the invoice. Bring your own key if you'd rather meter it yourself.
Build agents on your model, not around it.
Agents, knowledge bases, and tools composed in a studio — multi-provider models with failover, and the open MCP standard so external AI clients connect securely to your data.
Ask the question. Get the report, the dashboard, or the chart.
An answer is not always a paragraph. Sense runs the saved reports and dashboards your business already trusts, and when the question is about a population it draws the chart on the spot — from your live records, as you, with the working shown.
Sense, inside Claude and ChatGPT.
Connect the assistant your people already use to your xMatix organisation over one open endpoint. It asks your records, runs your reports and dashboards, searches your knowledge and — only if you allow it — acts. As the signed-in user, inside their permissions, with every call audited. Off by default; one switch, one ceiling, one log for IT.
How the connector works →Admins configure by chatting — with approval gates.
"Add a 'preferred delivery window' field to outlets and put it on the mobile visit form." Sense drafts the change set; you review and promote.
Agentic AI, stated as jobs.
Every use case as a person, their problems, the flow the agent runs, and where it stops for a human. Twelve of them, across claims, service, sales, inventory, finance, field sales, merchandising and administration.
Questions the AI-curious ask.
No. Customer Content is never used to train AI models. Prompts and the context retrieved to answer them are processed for that request, scoped to your tenant, and handled under the terms of the Data Processing Addendum; model credentials are stored encrypted. The full position — including which providers are involved and what is logged — is in the Responsible AI policy.
They get the answer their permissions allow, and nothing more. Sense executes as the signed-in user through the platform's own data services, so record-access policies, field-level permissions, team and role boundaries and tenant isolation apply exactly as they do in the application — records outside a user's access are never retrieved, so they cannot leak into a summary. There is no privileged AI service account. How access is decided · Sense Assist.
AI is licensed, metered and capped. Seats, credit allocations, per-model rates and per-tenant and per-user quotas are enforced before a request reaches a model, and a budget guard pauses requests at 100% of the allocation — so there is no surprise overage. Consumption is visible per tenant and per user, on the same product → feature → capability licensing the rest of the platform uses. Licensing and metering.
Yes. You choose the models behind your tenant and can supply your own provider keys — stored encrypted — while administrators decide which AI capabilities are enabled for which users. The studio composes agents over multiple providers with failover, so a model can be swapped without rebuilding what runs on top of it. Sense Studio · Responsible AI.
They can answer, watch and draft. Consequential operations — data changes, destructive steps, publishing — execute only on the user's explicit approval and only within that user's permissions: Sense drafts the change and lands it as an approval rather than writing silently, and proactive nudges wait for a tap. Autonomous and scheduled runs operate under an explicitly assigned identity with least-privilege defaults, and every tool call, approval and refusal is on the audit trail. Sense Assist · Sense Control.