Give AI every capability.
Give it zero authority.
Imperium decouples what AI can work out from what it is allowed to do. The AI proposes; an independent, patent-pending governance layer authorises every consequential action; the system of record executes. No agent, however capable, ever holds the keys.
Everyone is racing to hand AI the keys
Every week brings a more capable agent — and a new way for it to act on its own. That is exactly backwards. An agent that can book, pay, message, cancel and commit on your behalf becomes a liability the first time it is confidently wrong. And it will be. Bigger models do not fix this; they raise the stakes.
Guardrails live inside the model
Prompt it the right way and they bend. The thing meant to hold the AI back is the same thing you are trusting to behave.
Permissions were built for people
Role-based access asks who you are, once. It has no answer for an autonomous agent taking thousands of actions an hour.
The audit trail is an afterthought
When it goes wrong, ‘which system, which decision, on whose authority?’ is nearly impossible to reconstruct after the fact.
Capability and authority are not the same thing. The whole industry is fusing them — a more capable agent is handed more freedom to act. We split them apart, and put the authority somewhere the AI cannot reach.
That separation is the difference between a clever demo and software you can put in front of customers, regulators and auditors.
Authority, held somewhere the AI can’t reach
Imperium introduces a separate runtime that sits between intent and execution. Every consequential action — whether it came from a person, a scheduled automation or an AI agent — has to pass through it before it takes effect. It applies the right limits, lets routine actions through, holds risky ones for a human, blocks anything outside policy, and writes a tamper-evident record of every decision. It is independent by design: that is what makes its answer to “was this allowed?” consistent, provable, and identical for humans, automations and AI alike. The technology is the subject of seven patents pending.
- One gate for every actor — human, automation or agent
- Allow, hold-for-approval or block — decided per action, in context
- Fails safe: in doubt, it holds rather than risks
- A tamper-evident record of who decided what, and why
What Imperium is not
Guardrails live inside the AI and travel with it. Imperium lives outside it — it cannot be prompted, fine-tuned or jailbroken away.
RBAC asks who you are. Imperium asks whether this specific action, right now, in this context, is allowed — and can stop it for a human.
Routine work flows through instantly. Only genuine risk pauses for approval, so governance speeds AI up instead of strangling it.
A single gate governs every system, automation and agent you run — swap the model underneath and the guarantees still hold.
Three roles every action passes through
Keeping these responsibilities separate is the whole design — one action, one governance decision, one audit entry.
Execute · Operational products
Systems of record that hold the real data and carry out every action deterministically — bookings, jobs, invoices, payments, stock, communications. Nothing actually happens anywhere else.
Judge · AgentFlowMe
A fleet of specialist AI agents that interpret messy real-world input, weigh trade-offs and make expert recommendations. The AI proposes — it never executes.
Govern · Agent Shield
An independent governance layer that authorises every consequential action: applies the right limits, holds risky actions for human approval, blocks policy breaches, and records a tamper-evident audit trail — for humans, automations and AI alike.
Four parts, one governed platform
Operational platforms
Complete production SaaS that runs the core of a business — JobFlowMe (field-service & motor trade) and IMAISA (field-service & helpdesk with human-in-the-loop approval). Deterministic, reliable, auditable, always in control of what changes.
AgentFlowMe
The governed AI workforce — 200+ specialist agents across facilities management, accountancy, legal and the motor trade. Each is purpose-built for a real task and proposes its work back to the operational product for execution and human oversight.
Agent Shield
The patent-pending trust layer. It treats agents like employees — defined authority, accountability, and consequential decisions checked before they take effect. Allows routine actions, holds higher-risk ones for approval, blocks anything outside policy, and fails safe.
AISight
Visibility across the whole suite — what the agents are doing, what was decided, held or escalated, and the value delivered — so leaders can see and trust the automation at a glance.
One action, start to finish
- 1
Something happens — a new enquiry, an overdue invoice, an unassigned job, a customer message.
- 2
AgentFlowMe judges it — the specialist agent interprets the situation and prepares a recommendation or drafted action. It does not carry anything out.
- 3
The operational product takes the proposal — and, at the moment of executing, asks Agent Shield for authorisation.
- 4
Agent Shield decides — allow, hold for a human, or block — and records the decision.
- 5
The action is carried out (or held) — once, with a single audit entry, fully accountable.
An independent gate, protected by design
A human action, an automated process and an AI agent are all judged by the same gate — so an organisation gets one consistent, provable answer to “was this allowed?” across everything it does. The governance technology that makes that possible is the subject of seven patents pending.
AI you can put in charge of real operations
Safe automation
AI proposes; it never has the keys. Every meaningful action is authorised and logged.
One source of truth for ‘allowed’
Humans, automations and AI are all governed the same way.
Built for accountability
A tamper-evident audit trail and clear human-approval points — for sectors where that isn't optional.
Fails safe by default
When in doubt, the system holds rather than risks.
Proven across sectors
Facilities management, accountancy, legal and the motor trade, with hundreds of governed agents.
Putting AI into operations? Build the brakes before the engine.
The Imperium Suite is live across facilities management, accountancy, legal and the motor trade. Get a briefing and see how the governance layer fits your stack.