PROVOST BY IGNITE9 · THE GOVERNED AI OPERATING SYSTEM
Governed AI agents that get your business work done, inside your own walls.
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.
- Every system change waits for a named approver
- Runs in your cloud, on one server, or fully offline
- English and Arabic workspace
- Workday
- Oracle Fusion
- Avalara AvaTax
- SAP S/4HANA · planned
- Salesforce · planned
01 / WHY LEADERS CHOOSE IT
An AI workforce your board, auditors and regulators can live with.
Most enterprise AI pilots stall on the same questions: who approved this, where did our data go, and what will it cost at scale. Provost answers them in the design.
Every action under your control
A policy engine checks every agent action. Anything that changes a business system waits for a named person to approve it, and nobody approves their own request.
Inside your own walls
Run it in your own cloud or data center, on a single server, or fully offline with no internet connection: sovereign AI for regulated industries and government.
AI cost you can predict
Large business data stays inside your boundary and the model sees a compact view. In our benchmark one task over 4,004 invoices sent 256 tokens instead of 187,203.
Institutional knowledge that stays
Every request, approval and change is kept on a tamper-evident record, so how work was done and decided stays with the company when people move on.
02 / ONE CONVERSATION
Ask for the work in plain language. Approve what matters.
A planner hands each request to specialist agents that work across your systems. You see every step, and nothing changes in a system until a named person approves it.
Check this quarter’s supplier invoices in Oracle Fusion against the tax we should have paid, and draft the corrections.
- PlannerSplit the request into three tasks and checked each against policy
- Finance agentRead 4,004 supplier invoices in Oracle Fusion (read-only)
- Tax agentChecked the tax due on each invoice with Avalara AvaTax
- Finance agentDrafted correcting entries for review
Changes a system, so it waits. Approver: head of tax (not the requester). 0 of 2 approvals.
03 / AI COST OPTIMIZATION
Pay for the answer, not for shipping your data to a model.
A context firewall keeps large business data inside your boundary and gives the model a compact view of it. Fewer tokens means a smaller AI bill, and less of your data leaves the building.
AI cost control in detailOUR BENCHMARK · ONE TASK OVER 4,004 INVOICES
04 / ESTIMATE YOUR RETURN
What could governed AI agents return to your teams?
A quick estimate from your own numbers, with the formula in plain sight.
- Time returned to your teams
- Enter the first four values
- Value of that time
- —
- AI model spend avoided
- Optional
Hours returned a month = hours × agent share × (1 − review share)
Value a year = hours returned × cost per hour × 12
Model spend avoided a year = model spend × data share × reduction × 12
An estimate from the values you entered, not a guarantee, a quote or a measured result. It leaves out license, implementation and change costs, and your results depend on your work, systems and approvals. We will gladly build a real business case with you in a demo.
05 / SOVEREIGN AI, YOUR WAY
Deploy it where your data already lives.
For regulated industries and government, the whole system can run inside your boundary, including air-gapped sites. Releases are signed images with a software bill of materials.
Your Kubernetes
Install in your own cluster with a Helm chart, under your network and identity rules.
A single server
Start small on one server in your data center or private cloud.
Google Cloud reference
A documented reference deployment on Google Cloud for teams that want a managed starting point.
Fully air-gapped
An offline bundle with signature verification, for sites with no internet connection at all.
06 / TWO IGNITE9 PRODUCTS
Provost and Tenured do different jobs.
Tenured
AI production support. It keeps the enterprise applications you already run working and works their tickets, with your people approving production changes.
Explore TenuredProvost
The governed AI operating system through which your people get business work done, deployable inside your own boundary, including fully offline.
Security and control07 / WHERE WE ARE TODAY
Built and tested today. Clear about what is next.
The core platform is built and tested, with continuous integration on every change. Ignite9 runs its own operations on it as the first pilot, and design-partner pilots are open. We have no customer names, results or certifications to show you yet, and we will not pretend otherwise.
Today, the audit trail, approvals and change history keep how work was done and decided, and a knowledge-graph view shows agents, systems and record types.
ROADMAP · NOT AVAILABLE TODAY
- Built-in ERP modules
- Automated migration off legacy applications
- Self-hosted models in production use
- Micro-VM sandboxes
- Confidential computing
- The full ontology data layer with a living organizational memory
- SAP S/4HANA and Salesforce connectors
- Certifications: SOC 2 and ISO 27001 have not started
08 / QUESTIONS LEADERS ASK
Straight answers.
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.
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.