TADA AI is the intelligence layer of the TADA Digital Twin Platform. Ask questions in plain language, get answers from your business data, and connect TADA to the AI tools you already use.
Table of Contents
How It Works
TADA AI sits between your users and your Digital Twin data. Users ask questions through one of several channels. TADA sends those questions to an AI model, which reasons over your data and returns an answer.
| Layer | What It Does |
|---|---|
| Users | Ask questions via the TADA App, Microsoft Teams, or your own AI agents |
| TADA AI | Routes questions to the right AI model and enforces permissions |
| AI Models | Reasons over your data — either TADA-provided models or your own (Azure OpenAI, Bedrock, etc.) |
| Digital Twin | Your business data, relationships, and context |
Getting Started
| Approach | Time to Value | Effort |
|---|---|---|
| TADA App AI | Immediate | None — already built in |
| Copilot | Days | Configure a Copilot Studio agent |
| MCP | Hours to days | Connect your agent to the MCP endpoint |
| BYOL | Hours | Provide your LLM endpoint and credentials |
Explainable AI - TADA App
AI is built into the TADA web app. Ask a question, get a streaming answer with data from your network. You can use the models TADA provides, or bring your own.
- Answers stream in real time
- Context-aware — the AI knows what you're looking at
- Specific Commands and Interactions
TADA-Provided
| Model | Best For | Cost |
|---|---|---|
| Ollama (Local) | Development, air-gapped environments | Free |
| Claude Haiku | Simple queries, high volume | Low |
| Claude Sonnet | Complex analysis, multi-step reasoning | Medium |
Bring Your Own LLM (BYOL)
Already have an AI provider? Use it. TADA handles the orchestration — your model handles the reasoning.
| Provider | Models | Runs In |
|---|---|---|
| Azure OpenAI | GPT-4o, GPT-4 Turbo | Your Azure |
| Azure AI Foundry | Custom fine-tuned models | Your Azure |
| AWS Bedrock | Claude, Llama, Mistral | Your AWS |
| Google Vertex AI | Gemini | Your GCP |
| Copilot Studio | M365 Copilot | Your M365 |
| Self-hosted Ollama | Any open model | Your infrastructure |
Setup Your Own Own AI Model
Benefits: your data stays in your cloud, you use models your compliance team has approved, and you can leverage existing enterprise agreements.
- Open TADA Admin Settings — Go to the AI configuration section
- Pick your provider — Azure OpenAI, Azure AI Foundry, AWS Bedrock, Google Vertex AI, Copilot Studio, or Self-hosted Ollama
-
Enter credentials — Each provider needs slightly different info:
- Azure OpenAI: endpoint URL, API key, deployment name
- AWS Bedrock: region, access key, secret key, model ID
- Google Vertex AI: project ID, region, service account credentials
- Ollama: server URL, model name
- Test the connection — Verify TADA can reach your model
- Set as default (optional) — Apply to all users or specific groups
You can switch providers at any time without losing saved sessions.
Security
- Authentication — All access goes through TADA SSO
- Permissions — Enforced at every layer. Users only see data they have access to
- Session isolation — Conversations are private to each user
- On-premise option — The entire stack can run inside your network
- No data leakage — BYOL models keep your data in your own infrastructure
Deployment
| Option | How It Works |
|---|---|
| Cloud-Hosted | TADA hosts everything. AI calls to cloud LLM providers are outbound-only HTTPS |
| On-Premise | Full stack in your network. Use local Ollama for fully air-gapped environments |
| Hybrid | Platform on-premise, AI model in your cloud (Azure OpenAI, Bedrock, etc.) |
Related: Ask TADA — AI Chat Assistant