Procurement and security pack
Mirage Metrics builds and operates AI agents inside a client's operations. This pack gathers, in one place and one PDF, what procurement, IT and legal reviewers ask before a first contract. It covers who we are, how an engagement runs and what it costs, where the agents run and who holds the data, and how access and approvals work. It also gives our answers to the questions IT teams send us. Everything in it is what our security and enterprise pages state, word for word.
- Published
- Reading time
- Reference
- Written by
- Mirage Metrics team
The pack is updated when /security or /enterprise changes. It lists no certification or audit report, and we answer a security questionnaire on request.
The PDF holds the same content as this page, for a review file or a joint-venture partner.
Mirage Metrics at a glance
| Item | Detail |
|---|---|
| Legal entity | Mirage Metrics Corp., registered address 1111B S Governors Ave, STE 3123, Dover, DE 19904, United States |
| Founded | 2024, by Mehdi Yacoubi (CEO) and Ayyoub El Amrani (CTO) |
| Offices | Paris, Barcelona, Casablanca |
| People | Around fifteen employees, all engineers |
| Clients | More than 30 clients in 6 countries, 35+ deployments in production |
| Contact | mehdi@miragemetrics.com |
How an engagement runs, and what it costs
Forward-deployed engineers build each workflow inside your systems, with your team, then Mirage operates, monitors and improves it in production.
| Item | Detail |
|---|---|
| First reply | Within two business hours. |
| Proof of concept | On your own documents and data, one week after the first meeting. |
| Pilot | Engineers on site, one workflow, at a fixed price. You can stop after it. |
| Production | The first workflow in production in 4 to 6 weeks, operated, monitored and improved by Mirage. |
| Next workflows | Added on the same platform. Transwin started with one document workflow and now runs five systems in production. |
A fixed-price pilot first. From $10,000 to launch, then an annual subscription from $1,000 a month, priced by volume; the exact amount depends on the workflow and the volume and is quoted in the first conversation.
| Item | Detail |
|---|---|
| Pilot | Fixed price, on one workflow, with no commitment beyond it. |
| Subscription | Annual, priced by the volume of work the agents handle, not by the number of users. It only starts if you decide to continue after the pilot. |
| What it covers | The platform, the agents running on it, and Mirage's engineers operating, monitoring and improving them. |
| What it does not include | Licences for your own systems, and hardware where a deployment needs cameras, probes or terminals; both are named in the quote. |
| Compared with per-seat tools | A per-user licence grows with the number of people who open the tool. A per-volume subscription grows with the work the agents do, which is what you are paying for; a workflow that ten people review costs the same as one that two people review. |
| Stage | When | What you pay | You can |
|---|---|---|---|
| Proof of concept | One week after the first meeting | Nothing: it runs on a sample of your own documents or data | Stop, or scope the pilot |
| Pilot | Engineers on site, one workflow | A fixed price, from $10,000 to launch | Stop after it, with the output and the comparison in hand |
| Production | First workflow live in 4 to 6 weeks | An annual subscription by volume, from $1,000 a month | Add the next workflow on the same platform |
Security commitments
- Zero data retention
- Your data never trains a model, ours or anyone else's. The model providers a deployment calls work under zero-retention agreements: they keep nothing once a request is answered.
- Deployment and isolation
- Mirage runs where your policies require, from the cloud to a server on your premises. Each deployment is isolated: its own environment and its own data, shared with no other client.
- Full audit trail
- Every agent action is logged with the inputs it received, the sources it relied on and the result it produced, so any figure can be traced back to the document it came from.
- Access and human approval
- Access rights are set by role. Before an agent writes anything into your systems, a named person on your side approves it.
Where the agents run, and who holds the data
Deployment options
- Cloud
- Private cloud
- Your cloud account
- Sovereign cloud
- On premises
- Edge and offline
Mirage can run in your infrastructure, in your own cloud account, or hosted by Mirage. The architecture decides who holds the data, the keys and the access.
| Item | Detail |
|---|---|
| Data | In your infrastructure or your cloud, data and processing stay in the environment you control, and nothing is transferred outside it. |
| Keys | In those deployments, the encryption keys are your responsibility. |
| Access | Mirage's access is strictly limited, or removed after deployment, depending on the architecture you choose. |
| Review | A non-disclosure agreement on request, and answers to your security questionnaire for the architecture you are considering. No certification or audit report is listed. |
How it connects to your systems
Mirage connects the ERP, TMS, WMS, maintenance, telematics and document systems already in place, more than 200 so far, and writes back into them. Nothing is migrated and nothing is replaced.
| Item | Detail |
|---|---|
| With an interface | Where a documented interface exists, Mirage connects to it. |
| Without one | Where none exists, our engineers build the connector during deployment. |
| Construction | Project management and document control platforms such as Procore and Autodesk Construction Cloud, estimating spreadsheets, and ERP and accounting systems such as SAP, Oracle and Sage. |
| Logistics | TMS and freight systems: in production with Logitude World at OFRET Groupe, and CargoWise through its eAdaptor interface. |
| Manufacturing | ERPs: in production with SAP Business One at Biodéal, and Dynamics 365 Supply Chain Management. |
Approvals, access and audit
Mirage models your operation's own objects and rules: orders, shipments, assets, sites, suppliers, quantities. The agents follow the way your team already works rather than a generic template.
| Item | Detail |
|---|---|
| What you give | Your workflow, its rules and exceptions, your codes and document conventions, and the checks your team does by hand today. |
| Autonomy | Configurable for each workflow: what runs on its own and what waits for a person's approval. |
| Approval | Role-based access and human approval flows. A person on your side decides the exceptions. |
| Audit | Every agent action is traceable to the input and the rule that produced it. |
Questions your IT team will ask
- What data does Mirage keep, and does it train a model?
- Your data never trains a model, ours or anyone else's. The model providers a deployment calls work under zero-retention agreements and keep nothing once a request is answered. When Mirage runs in your infrastructure or your cloud account, data and processing stay in the environment you control.
- Where can Mirage run?
- In the cloud, a private cloud, your own cloud account, a sovereign cloud, on your premises, or at the edge and offline. Each deployment is isolated: its own environment and its own data, shared with no other client.
- How does it connect to our systems?
- Through the interface each vendor provides: an API, a file exchange, eAdaptor for CargoWise, data entities over OData for Dynamics 365. Where a system has no API, Mirage builds the connector during the deployment. Nothing is migrated.
- Who at Mirage can access our environment?
- Mirage's access is limited to what the deployment needs, and you can remove it. In a deployment in your infrastructure or your cloud, the encryption keys are yours.
- How are user access and approvals controlled?
- Access rights are set by role. Approvals and the thresholds that send an item to a person are set for each workflow, and every agent action is logged with its inputs, sources and result.
- Will you complete our security questionnaire?
- Yes. Send it to mehdi@miragemetrics.com.
- Do you have a template for our own AI policy?
- Yes. The AI policy guide gives contractors and joint ventures a free policy template (.docx), the clauses an owner can put in an RFP, and a vendor evaluation checklist. It is a starting point for your own policy, not legal advice.
Questions to put to any AI agent vendor
Nine questions IT, procurement and operations leaders ask Mirage, with the page that answers each one. Put the same nine to every vendor you compare.
- Which of our systems does it write into, through which interface, and who enables the access?
- Named per workflow in the first meeting; where no documented interface exists, Mirage's engineers build the connector. See the integration section above.
- Where does it run, who holds the data and the keys?
- Your infrastructure, your cloud account or Mirage-hosted; in the first two, data and keys stay with you. See security and data handling.
- What does a person approve, and can we change that per workflow?
- Configurable autonomy per workflow, role-based access, human approval on the exceptions you define.
- What is logged, and can we export it?
- Every agent action is traceable to the input and the rule that produced it; audit logs stay under your organisation's control.
- What happens to a document the agent cannot settle?
- It is held with the reason and routed to a named person; nothing is dropped or guessed silently. At Biodéal, about 7% of orders are flagged for a person.
- Who runs it after go-live, and who changes it when our ERP is upgraded?
- Mirage operates, monitors and improves it in production, with the engineers who built it.
- What results can you show from a named deployment, with the definition behind each figure?
- Nine published case studies. Six state their figures with the definition behind each: Biodéal, OFRET Groupe, Transwin, John Cockerill, OCBM, Mass Céréales Al Maghreb. Rouandi describes its deployment without figures, and SGTM and G3C describe missions that are just starting.
- What does it cost, on what basis, and what can we stop?
- A fixed-price pilot you can stop after. From $10,000 to launch, then an annual subscription from $1,000 a month, priced by volume.
- What will you not do?
- Mirage does not sell self-serve software, does not check drawings against US building codes, and keeps no material catalogue or labour database: it feeds yours.
Documents and reviews
- A non-disclosure agreement is available on request.
- We answer your security questionnaire for the architecture you are considering: send it to mehdi@miragemetrics.com.
- No certification or audit report is listed on this site today, and this pack claims none.
Questions this pack does not answer?
Send them with your questionnaire, or bring one workflow to a first call. We will show you where the agent acts, where your team approves, and what the log keeps.