The Role Map for a Real AI Build
Every AI build needs four functions covered: an AI/ML engineer to build and evaluate the system, a data engineer to make data usable, IT/security for access and infrastructure, and a domain owner to define decision rules and approve outputs. Small teams combine roles; none is skippable.
The four functions
The four functions are AI/ML engineering (reliability), data engineering (trustworthy inputs), IT/security (access and infrastructure), and domain ownership (rules and approval). Coverage matters more than headcount, two or three people plus a partner can cover all four.
- AI / ML engineer, owns whether the system is reliable: retrieval, prompting, evaluation, guardrails.
- Data engineer, makes data reachable and clean; 28–41% of SMBs struggle here (adoption research, 2026), so it is rarely optional.
- IT / security, access control, infrastructure, integration; your existing IT provider’s home turf.
- Domain owner, defines the decision rules and approves high-stakes outputs; the accountable human.
Covering the map without over-hiring
A mid-market business rarely needs four hires. A specialist partner typically covers AI and data engineering, your IT provider covers access and infrastructure, and the owner or a manager holds the domain role, which is why 41% of firms create hybrid technical-business roles.
The one role you cannot outsource is the domain owner, the person who decides the rules the AI must follow and takes accountability for outputs. A vendor can build the system, but only you can own the decisions it automates.
Frequently asked questions.
How many people do I need to build an AI system?
Fewer than you think, what matters is covering four functions (AI engineering, data engineering, IT/security, domain ownership), not hiring four people. A mid-market build is often two or three people plus a specialist partner, with 41% of firms using hybrid technical-business roles.
Which AI role can we never outsource?
The domain owner, the person who defines the decision rules the AI follows and approves high-stakes outputs. A partner can build and even run the system, but accountability for the decisions it automates has to stay inside your business.