Model Selection Agent
A one-stop assessment to determine whether Data Scan is needed alongside AITDS (i.e., prepare training data for AI), identify the right AI model for the use case, and seamlessly progress to implementation..
Business Fit · Data Need · Delivery Path
Recommended Delivery Pathway
Pending
Required
Consider
Not Required
Business fit
Process Scan
Required
Start here to confirm business fit, value, risk, and workflow impact.
Data gate
Data Scan
Pending
Submit to confirm whether data readiness, quality, and governance checks are needed.
AI enablement
AITDS
Pending
Submit to confirm whether specialist AI training and design support is needed.
Model change
Model Training
Pending
Submit to confirm whether the model itself must be tuned or trained.
Delivery
Use Case Development
Pending
Submit to confirm the right route into solution design, build, test, and deployment.
Submit the questionnaire to see the recommended delivery route.
Final recommendation
Click Submit to calculate
--The result will explain whether an existing model is enough, whether RAG is needed, or whether tuning/training should be considered.
Recommendation Summary
Complete the questionnaire and submit to see the recommended approach.
Why this result?The tool compares business need, knowledge source, behaviour change, accuracy risk, scale, cost, data readiness, and RAG suitability.
Approach Strength
Questionnaire
Questions 1–10 assess the use case, data need, risk, scale, and readiness. Q3A checks whether specialist domain understanding is required. Questions 11–20 test whether RAG with an existing model can work before considering tuning or pre-training.
Decision Logic
Move directly to Use Case Development whenThe process scan confirms AI fit and the solution can use an existing model without enterprise, client, sensitive, or governed data.
Add Data Scan before development whenThe solution uses enterprise knowledge, client data, sensitive data, records, documents, or governed data, even if no model training is needed.
Add AITDS and Model Training whenThe model must learn repeatable behaviour, specialist patterns, strict output formats, or domain understanding that prompting and RAG cannot deliver.