AI Models in Pragma — Data Flow & Benefits

This page explains which AI models power Pragma, how they work together — and why that's better for your sales workflow than a generic CRM.

---

The AI data flow in Pragma

---

What does each AI module do?

1. Translate & Summarize

As soon as a message arrives, the AI automatically detects the language (Chinese, English, German, Japanese, Korean…) and creates a 2-sentence summary in your target language. You never need to manually translate again.

It also classifies the lead's intent:

• low — general interest, no concrete purchase intent
• medium — interest, but still unclear
• high — concrete buying signals (demo request, budget mention, appointment request)

2. Intelligence Pack

The deepest analysis step: the AI extracts structured information from the message — industry, company size, budget hints, buying stage, and gives a routing recommendation (which sales team should take this lead?). It also detects duplicates.

3. Lead score (0–100)

Every lead gets a score — calculated from intent, signals, and context. No more manual prioritization. Your team immediately sees which lead to contact first.

4. Follow-up draft & email draft

One click and Pragma generates a ready-to-send follow-up text — matched to the lead's language and tone. For emails, there's a complete draft with subject line.

5. Proposal draft

For qualified leads, Pragma automatically creates a first proposal draft — quote, pitch, or executive summary — based on lead context.

6. Next best action

The AI always recommends what to do next: "Request appointment", "Show demo", "Send proposal" — with reasoning. So you know where you stand even after two weeks away.

7. Dashboard narrative & KPI summary

Every day Pragma generates a short summary of your most important KPIs — as text, directly in the dashboard. No number-crunching, just read.

---

Which models does Pragma use?

Use case	Model (default)	Alternative
All AI tasks	Claude Sonnet 4.6 (Anthropic)	GPT-4o mini (OpenAI) · custom deployment
EU data protection	Azure OpenAI (EU region)	GDPR-compliant out of the box
Without API key	Heuristic fallback	For demos & tests

Auto-provider detection: Set ANTHROPIC_API_KEY for Claude (recommended), OPENAI_API_KEY for GPT-4o mini, or configure Azure OpenAI for EU-region GDPR compliance.

BYOK — Bring Your Own Key: You can use your own Anthropic, OpenAI or Azure API key. Pragma never makes API calls at your expense without a key configured.

---

What sets Pragma apart from other platforms?

Feature	Pragma	HubSpot	Salesforce	Pipedrive
Auto language detection (ZH/JA/KO)	✅	❌	❌	❌
AI score at first contact	✅	⚠️ Add-on	⚠️ Add-on	❌
WeCom / WeChat native	✅	❌	❌	❌
Follow-up draft in target language	✅	❌	❌	❌
BYOK (Anthropic / OpenAI / Azure)	✅	❌	❌	❌
Azure EU region GDPR	✅	⚠️ Extra cost	⚠️ Extra cost	❌
AI result cached (no duplicate costs)	✅	n/a	n/a	n/a
Price	💶 Flat	💶💶💶	💶💶💶💶	💶💶

---

AI on/off

Admins can control AI automation workspace-wide:

• Auto-analysis on/off (automatically analyze new leads)
• Batch size configurable (how many leads per cron run)

Settings: Admin → Integrations → AI configuration

---

Further guides

• AI scoring & follow-ups — understand scores, use Copilot
• Sales workflow overview — AI in the big picture
• Reporting & Analytics — measure AI coverage in reporting
• API keys & webhooks — add your own OpenAI key (BYOK)