Staff augmentation

The client owns priorities, architecture, coordination, acceptance, and outcomes; the vendor provides named skills and capacity.

AI delivery model decision

No. Staff augmentation provides capacity under client direction; outsourcing delivers against an agreed scope; an FDE owns a bounded operational problem, continuously connecting field learning, implementation, deployment, and evidence of outcomes.

Author Onyx Devs Lab · Published · Updated

Job titles and whether someone is on-site do not by themselves determine the engagement model.

The vendor delivers against a defined scope, milestones, and acceptance criteria; changes are handled through an agreed process.

When the requirements must be learned within real workflows, and the team owns the diagnosis-to-evidence loop for one bounded operational problem.

Choose the simplest model that also places accountability where the necessary decisions can be made.

The work is already decomposed, internal technical leadership is mature, and the main constraint is short-term capacity.

Users, workflows, interfaces, constraints, testing, and acceptance can be clearly defined with limited on-site exploration.

The high-value problem is known, but the boundaries of data, workflows, adoption, and intervention need to be discovered and tested together with operators.

Before choosing manual labeling, first write out the business decision, the current baseline, the people affected, the constraints, and the consequences of failure.

Decide whether the vendor is responsible for capacity, for specified deliverables, or for the evidence loop of an operational outcome.

Agree on field access, decision rights, product and engineering cadence, evidence review, and how new discoveries change the scope.

Specify when uncertainty is considered resolved, who receives the system and the knowledge, and which model is used after the FDE phase.

No. It requires proximity to the operating environment and fast decision loops; the on-site versus remote ratio depends on the workflows, data, systems, and users.

A bounded phase can have a fixed commercial scope, but uncertainty, evidence gates, change rules, and stop conditions must be clear; freezing a feature list can strip the FDE engagement of its meaning.

Consulting work can end at a decision or a roadmap; an FDE continues through implementation and production evidence for that bounded problem.

Yes. Once uncertainty has decreased and accountability is clear, the FDE should transition to stable product delivery, internal ownership, or ongoing operations.