Real-Time Release Testing Surrogate Models
Uses inline spectroscopy and process signals to estimate CQAs earlier in batch disposition workflows Evidence basis: Published RTRT studies demonstrate ML surrogate models can predict dissolution and support near-real-time quality decisions; FDA PAT guidance provides a framework for model-based control when validation and lifecycle management are robust
The Problem
“Real-Time Release Testing Surrogate Models”
Organizations face these key challenges:
Uses inline spectroscopy and process signals to estimate CQAs earlier in batch disposition workflows
Impact When Solved
The Shift
Human Does
- •Review batch records, lab results, and process notes to assess release readiness
- •Coordinate quality checks and follow-ups across manual spreadsheets and handoffs
- •Investigate deviations or late-emerging quality concerns before disposition
- •Decide batch disposition based on retrospective evidence and expert judgment
Automation
- •No AI-driven analysis in the legacy workflow
- •No automated prioritization of quality risk or release readiness
- •No continuous estimation of CQAs from inline process information
Human Does
- •Approve surrogate model use within defined release decision workflows
- •Review predicted CQA risk and decide disposition actions with supporting evidence
- •Handle exceptions, model challenges, and batches flagged for additional review
AI Handles
- •Estimate CQAs earlier using inline spectroscopy and process signals
- •Monitor batches in near real time and flag elevated quality risk or uncertainty
- •Prioritize batches for review based on predicted release readiness
- •Generate decision-support summaries for batch disposition workflows
Operating Intelligence
How Real-Time Release Testing Surrogate Models runs once it is live
AI runs the first three steps autonomously.
Humans own every decision.
The system gets smarter each cycle.
Who is in control at each step
Each column marks the operating owner for that step. AI-led actions sit above the divider, human decisions and feedback loops sit below it.
Step 1
Assemble Context
Step 2
Analyze
Step 3
Recommend
Step 4
Human Decision
Step 5
Execute
Step 6
Feedback
AI lead
Autonomous execution
Human lead
Approval, override, feedback
AI handles assembly, analysis, and execution. The human gate sits at the decision point. Every cycle refines future recommendations.
The Loop
6 steps
Assemble Context
Combine the relevant records, signals, and constraints.
Analyze
Evaluate options, risk, and likely outcomes.
Recommend
Present a ranked recommendation with supporting rationale.
Human Decision
A human accepts, edits, or rejects the recommendation.
Authority gates · 1
The system must not make a final batch disposition decision without review and approval from the designated quality decision-maker [S1].
Why this step is human
The decision carries real-world consequences that require professional judgment and accountability.
Execute
Carry out the approved action in the operating workflow.
Feedback
Outcome data improves future recommendations.
1 operating angles mapped
Operational Depth
Technologies
Technologies commonly used in Real-Time Release Testing Surrogate Models implementations: