AI that stays in its lane.
You decide what it knows, how it speaks, and when a human takes over. DigAI Micro-Brains ground every automated response in your private documents — eliminating ungrounded hallucinations.
How a DigAI Agent Answers Patient & Client Questions
Click through the 4 execution stages below to observe how Micro-Brains prevent ungrounded hallucinations.
Prospect or patient messages your WhatsApp number or website chat widget with a question.
Granular AI Guardrails You Control
Configure strict operating boundaries directly inside your Micro-Brain dashboard.
Custom Persona & Tone
Instruct your agent to be friendly, concise, or strictly clinical. Define custom greeting formulas.
Grounded Knowledge Boundaries
Configure the agent to say "I do not have that information, let me connect you with a team member" if a topic is absent in your docs.
Smart Human Hand-Off Triggers
Workflows automatically pause AI replies as soon as a human team member sends a message or a customer requests a real person.
Message Length & Pacing
Set concise reply constraints (1–3 sentences) to maintain clean readability on mobile WhatsApp screens.
Bring-Your-Own AI Key (BYO-AI)
Connect your own OpenAI, Anthropic, or private enterprise LLM API key with zero platform markup.
Complete Transcript Audit
Review every conversation thread in the Unified Inbox with highlighted citations of which document generated the answer.
What We Don’t Do With Your Data
Clear commitments on training, permissions, and non-substitute advisories.
Zero Foundational Model Training
Customer queries and uploaded knowledge documents are transmitted strictly as inference context. Your proprietary price lists and patient inquiries are never used to train public models.
No Autonomous Unbounded Actions
Agents can only perform explicit actions you permit: answering queries, tagging contacts, and scheduling booking calendar appointments.
Not a Substitute for Professional Counsel
DigAI agents assist with operational inquiries (pricing, slots, addresses, FAQs). High-stakes diagnostic or contractual determinations must always be routed to human professionals.
Known Technical Limitations & Operational Realities
Language models are probabilistic systems. Here are the edge cases every business owner should understand before going live:
Hallucination & Ambiguity Risks
While strict grounding minimizes hallucinations, complex multi-part questions or contradictory documents can confuse language models. Always keep knowledge bases updated.
Vernacular & Dialect Nuances
DigAI agents excel at English, Hindi, and common Indian Hinglish phrases. Regional dialects and informal spelling variations may occasionally require human escalation.
API Latency Constraints
Third-party LLM response generation typically takes 1.5 to 3.5 seconds. We include typing indicators so visitors recognize an automated reply is formulating.
Responsible AI Launch Checklist
Report an AI Safety or Hallucination Incident
If an automated response produced inappropriate, inaccurate, or unexpected output, report the snippet below. Our prompt engineering team investigates all safety flags.
Frequently Asked AI Safety Questions
When "Strict Grounding" is enabled, the agent is constrained by system prompts to refuse to answer queries not covered in your knowledge documents. It will politely inform the user that it does not have that information and route to staff.
Deploy grounded AI for your business
Connect your WhatsApp lines to Micro-Brains trained exclusively on your documents.