The advertisement for an AI specialist goes out, dozens of CVs arrive, and after a few interviews it becomes clear that none of them is quite what the organisation needs. The problem did not start with the labour market; it started with an advertisement that never defined the role.

AI is not a single discipline. The person who builds the data pipeline, the person who trains the model and the person who turns model output into something usable are three different skill sets, and they rarely live in one individual. Before any decision about hiring locally or remotely, it must be clear which of them you want.

Define the role first, then advertise

In most business projects the first priority is a data engineer, not a model researcher. As long as data is scattered, inconsistent and without history, even the best machine-learning specialist cannot move things forward. The second role belongs to someone who knows the business domain and can say which output is actually usable in practice. The third is an engineer who makes the solution stable and maintainable. That is usually the right hiring order.

Local or remote: the deciding criteria

• Data sensitivity: if the data may not leave the organisation's network, on-site presence or tightly controlled access is a precondition.
• Internal process maturity: a remote team loses its productivity in an organisation without documentation and clearly defined responsibilities.
• Nature of the work: research and project work suits remote arrangements; work that needs daily conversation with operating units suits it far less.
• Cost versus coordination: remote hiring usually widens access to talent, but raises the cost of coordination and management.

Building from within or bringing in from outside

In many organisations the best candidate is already inside the company: an analyst who knows the process deeply and has analytical skill. Training that person is usually faster and more durable than recruiting a specialist who will spend months learning the business. Hiring externally is justified where the required skill genuinely does not exist in-house and the timeline leaves no room for training. A combination of the two, one experienced specialist alongside two internal people, tends to give the most durable result.

The real challenge in Iran: retention, not recruitment

Capable technical talent is not scarce in Iran; keeping it is the hard part. A specialist trained inside your organisation quickly becomes attractive to the outside market and to employers abroad. Salary is only one factor, and it is usually the one a small organisation cannot compete on. What actually retains people is serious work they can be proud of, authority over technical decisions and a clear path for growth. If your specialist spends months cleaning data by hand, do not expect them to stay.

A three-month path to building the team

• Month one: list which work genuinely has to happen inside the organisation and which can go to a contractor. Define roles only for the first category.
• Month two: hire one person with practical experience and at the same time assign two internal analysts to work alongside them. Base the interview assessment on a real problem from your own data, not on theoretical questions.
• Month three: give this group ownership of one specific process along with authority over technical decisions. Make knowledge transfer mandatory from day one so no single-person dependency forms.

Common hiring mistakes

• Searching for one person who is simultaneously a data engineer, a model specialist and a software developer; such an advertisement usually goes unanswered or attracts the wrong candidate.
• Assessing candidates with theoretical questions instead of a short exercise on the organisation's real data.
• Hiring before the data is ready; the new specialist sits idle for months or ends up doing work they were not hired for.
• Having no knowledge-transfer plan, which ties the organisation to one individual and turns their departure into a crisis.

How to tell the choice was right

The right measure is not the number of models built. Three signals say more: the time between defining a problem and producing the first assessable output; the number of processes that keep running without that particular person; and whether internal analysts have gradually become able to carry part of the work independently.

Frequently asked questions

• How many people are needed to start?
  At small and mid-size scale, one experienced person alongside one or two internal analysts is enough for a first project.
• Is outsourcing to a contractor a sound choice?
  For a first pilot, yes, provided ownership of the data, code and documentation stays with you and knowledge transfer is written into the contract.
• Which matters more, technical skill or business knowledge?
  In small organisations business knowledge is usually scarcer and slower to acquire; foundational technical skill can be filled in more quickly.

Takeaway

Once the role is precisely defined, choosing between local and remote hiring is a straightforward decision. What determines the outcome is the right hiring order, giving the team serious work, and a clear plan for keeping them. A small stable team delivers more than a large one with high turnover.

Glossary

• Data engineer: The person who builds and maintains the path along which data is collected, moved and prepared.
• Machine-learning engineer: The person who trains a model and readies it for durable use in a live environment.
• Data pipeline: The automated chain of steps that carries data from its source to the point of use.
• Build or buy: The choice between developing a capability internally and sourcing it externally.
• Retention: An organisation's ability to keep key people after investing in their training.
• Knowledge transfer: Documentation and training that lift work out of dependence on one individual.
• AI agent: Software that takes a goal and carries out the steps needed to reach a result on its own.