Stage 4 seeks approval for the AI solution to be deployed. For internally developed solutions this means further evaluating performance on data not yet used or reporting results from the pilot, trial or prototype; for procurement the initiative owner reports the market engagement outcomes, assures delivery within endorsed scope, budget and time and the stated benefits, and seeks support to award, or notes the award of, the contract after negotiations, with any additional funding, timelines and resources approved. The team develops an implementation plan that addresses identified risks and issues before deployment and assesses delivery within budget and timelines; gains approval to deploy into production or wider rollout on the performance report; for procurement assesses prospective suppliers against business objectives and the benefits sought, examines vendor information on accuracy and transparency, awards the contract and signs service level agreements, and obtains legal, commercial and contractual advice on risks to mitigate; conducts an impact assessment including social and environmental impact; addresses the issues stakeholders have raised or reassesses their support; completes any assurance required for ICT-enabled projects; sets minimum quality standards for use; assesses reputational risk from the vendor or the technology; decides whether and how outputs will be audited; brings the change management strategy up to date and communicates it; obtains approval and stakeholder buy-in for minimum service availability standards covering accessibility, percentage uptime, response time and downtime procedures; tests and validates the technical integration plan for compatibility with workflows and data pipelines and tests the user interface with expected end users; ensures the risk management plan is approved and updates the risk assessment; and establishes minimum KPI standards defined at warning and critical levels. The decision maker and governance, satisfied with performance and integration test results and with clinical governance input for clinical applications, approve deployment unconditionally or conditionally, request further information and reassessment, or decline, giving an unsuccessful initiative enough information and an agreed timeframe to revisit earlier stages.
The graph holds this control, the 0 it maps to, and the evidence behind each claim, over MCP and REST.