Per-Action AI Model Configuration
Different AI jobs call for different models. Template building likes a strong generalist; structured extraction likes a cheaper, faster model that follows schemas well; onboarding and conversational insights sit somewhere in between. FolioReady now routes each AI action to its own configured model, so support can tune accuracy, cost, and latency for each job independently.
What's New
Independent action channels
Each AI-powered feature has its own service and model pairing:
| Action | What it powers |
|---|---|
| Builder | AI template builder panel that scaffolds fields, sections, and validations |
| Extraction | Post-upload field extraction from uploaded documents |
| Onboarding | Wizard that reads a client's documents and proposes a starter workspace |
Previously the same model was used for every AI action, which meant a model tuned for chat was also doing structured extraction — or vice versa.
Per-company overrides
Each company can run on a different model mix. Our support team can pin a premium model for extraction on an enterprise account while another company runs a faster, cheaper model across the board — no release required.
Safe defaults
Any action that isn't configured for your company uses the platform default, so nothing changes for existing customers. You'll only see different behaviour if you ask us to set an override.
What's Improved
Faster iteration on AI behaviour
Tuning a single action — say, cutting extraction cost by switching to a smaller model — no longer risks regressing chat quality, because the two are fully decoupled.
Clearer cost attribution
Usage and cost reporting now line up with the action taken, making it much easier to answer "why did this company's AI bill move this month?"
Related
- Folio Insights — Conversational AI over a folio's documents
- Document Synopsis — Per-file summaries consumed by Insights