PROVOST BY IGNITE9 · FREQUENTLY ASKED QUESTIONS
Straight answers for leaders evaluating governed AI agents.
Provost by Ignite9 is the governed AI operating system. If your question is not here, ask us directly.
What is Provost?
Provost by Ignite9 is the governed AI operating system: your people ask for business work in plain language, specialist AI agents carry it across your systems, and anything that changes a system waits for a named approver. The long-term vision is one conversational system that a company talks to instead of working through hundreds of applications.
How is Provost different from Tenured?
Tenured is AI production support: it keeps your existing enterprise applications running and works their tickets. Provost is the governed AI operating system through which people get business work done, and it can run inside your own boundary, including fully offline. Both are Ignite9 products.
Can it run without an internet connection?
Yes. Besides your own Kubernetes cluster, a single server or a Google Cloud reference deployment, it can be installed fully air-gapped from an offline bundle with signature verification. Releases are signed images with a software bill of materials.
Which AI models does it use?
You choose per client: your own model endpoint, a self-hosted open-weight model as an option, or frontier models such as Claude, DeepSeek or Gemini, only through the data boundary gateway and with recorded consent. Each agent role can use a different model, and complex steps can be escalated to a stronger one.
Does our business data go to outside AI models?
Only if you allow it. Before anything reaches an outside model, names, emails, bank and account numbers, tax IDs, card numbers, amounts and your own term lists are replaced with placeholders and restored only in the answer. Each client chooses off, code-only or consented mode, and every call is audited.
Can an AI agent change our systems on its own?
No. A policy engine checks every action, and anything that changes a system waits for a named approver. Nobody approves their own request, sensitive changes can need several approvers, and every connection is read-only until a second approver grants write access.
How does it keep AI costs down?
A context firewall keeps large data inside your boundary and gives the model a compact view. In our benchmark, one task over 4,004 invoices sent 256 tokens to the model instead of 187,203 (99.9% fewer), and a task starts with about 583 tokens of context whether 100 or 10,000 tools are connected. Your results will depend on your work.
Which business systems does it connect to?
Workday, Oracle Fusion and Avalara AvaTax connectors are built and contract-tested against vendor-specification test services; connecting to a real tenant is a pilot step. SAP S/4HANA and Salesforce are planned.
Is it certified?
Not yet. SOC 2 and ISO 27001 work has not started, and we do not claim either. The core platform is built and tested, with continuous integration on every change.
Who uses it today?
Ignite9 runs its own operations on it as the first pilot, and design-partner pilots are open. There are no customer names or results to cite yet.
What is on the roadmap?
Built-in ERP modules, automated migration off legacy applications, self-hosted models in production use, micro-VM sandboxes, confidential computing and a full organizational memory are on the roadmap. They are not available today.
NEXT STEP
See Provost on one of your own processes.
In a briefing we walk through Provost on a process you choose, answer your security and risk team’s first questions, and agree whether a design-partner pilot makes sense.