Implementing AI in your organisation: path, challenges and cost
Most AI projects fail not because of the model but because of the path: a vague goal, unprepared data, and a deployment that never connects to daily work. The Miras path is designed to prevent exactly that.
Four steps, each with a defined output
Data audit and indicator
What data you have (sensors, documents, images, existing systems), where cost leaks, and which indicator is measurable today. Output: data map + one target indicator + a pilot proposal with a fixed price.
Limited pilot
The model is trained on your real data and delivers results beside the current process, without touching decisions. Output: the indicator before and after the pilot.
Deployment
On your server or on Miras GPU infrastructure in Tehran; connection to existing systems (ERP, SCADA, cameras, portal); user training. Output: the system inside daily work.
Monitoring and improvement
The model is refreshed with new data, the indicator reported monthly, and the scope grows on results.
The challenges of implementing AI in Iran, and the Miras answer
- Access and sanctions: foreign cloud services are unavailable or take data abroad. Answer: Miras’s own GPU infrastructure in Tehran and open-weight models.
- Persian: general-purpose models handle Persian poorly. Answer: Miras Persian language, vision-language and speech models tuned on industry data.
- Scattered data: data lives in Excel, on paper and in several systems. Answer: the data audit, starting where data is sufficient; the vision model for paper documents.
- User adoption: a system outside the workflow is not used. Answer: designed for the phone, no VPN, in the user’s language; training and support during deployment.
- Unknown cost: Answer: staged pricing — fixed-price pilot, deployment priced on results.
What is needed
- One project owner in your organisation.
- Read access to the relevant data (systems, historian, documents, cameras).
- One indicator that is measurable today.
- A server inside the organisation (optional) — otherwise Miras infrastructure.
How Miras prices
Audit
One working session with the Miras technical team; written output: data map, target indicator and a pilot proposal with a fixed price and duration.
Fixed-price pilot
Limited scope, a defined indicator, a fixed price. If the indicator does not move, full deployment is not proposed.
Deployment and maintenance
Priced on scope and the pilot result; periodic model maintenance and updates. The tuned model belongs to you.
Frequently asked questions
What does implementing AI in an organisation cost?
It depends on scope, data and infrastructure; a number given without seeing your organisation is not trustworthy. That is why Miras runs the audit with a written output and proposes the pilot at a fixed price; full deployment is priced on the pilot result.
How long does implementation take?
The audit is one session. The pilot length depends on data readiness and is committed in writing at the end of the audit. For example, the Mobarakeh Steel fuel-optimisation deployment took six months.
What if our data is not ready?
The audit shows exactly that. We usually start with a process that has data and design data collection for the others in parallel; paper documents and images are turned into data by the Miras vision model.
Who owns the model and the data?
The data is always yours. By default it stays in Iran, on your server or on Miras infrastructure in Tehran; a foreign fallback model is used only with the customer’s written consent. The model tuned for your organisation is, by contract, deliverable to your own server.
Bring intelligence to your plant or project
We start with a data audit: where your data is, where cost leaks, and which model pays back first. Tehran · Isfahan · Noshahr · Dubai.