Deployment Service Life Cycle

The following graphic summarises the Deployment Service Life Cycle, which can be followed by startups and larger organisations alike:

While the considerations for startups and larger organisations are generally similar, they do vary slightly. Startups and organisations that are generally smaller should take additional care when working through each phase of the Deployment Service Life Cycle, as the cost of certain resources may put them beyond their scope.

The considerations per step of the Deployment Service Life Cycle framework for both types of organisations are listed below:

	Startups	Larger Organisations
Gather	The requirements, goals and objectives Understand the AI/ML use case(s) Understand the current infrastructure requirements Understand the deployment expectations (Budget, timescales, resources)	As with startups, but including understand the AI/ML use case(s) and their connectivity to organisational strategy.
Assess	Current AI/ML model Maturity of the code Current available infrastructure Current data pipeline (if any) Data dependencies Size of the current dataset If current data is real or synthetic The data source for model in production Skills availability within the organisation	As with startups, but including: Past AI/ML models (if any) already in production The MLOps Maturity level The scalability requirements.
Plan / Design	Potential deployment pipeline Deployment strategy Metrics to be monitored
Implement	Deploy the model Review the deployment strategy
Evaluate / Review	Monitor the metrics of the model Observability - Model health, Data health and Service health