All guides

Graph lookup

Your voice agent should know how your business connects.

A caller asks about a course, then its next intake, then fees. Another asks about a property, its locality and whether it matches their budget. Swarit graph lookup brings connected business knowledge into the conversation while the step harness keeps the customer journey moving.

Human-like voice agents. 11 Indian languages and English.

Your voice agent should know how your business connects.

Give the agent connected knowledge

Your business information has relationships. A programme belongs to an institution and has intakes. A property has a location and attributes. A service is available at particular branches. Graph lookup lets the agent work with those connections as context for its answers.

The current knowledge tool set supports context queries, entity lookup, search and traversal. Configure the agent’s access explicitly so it can retrieve the information relevant to its task. Your team owns the facts and the relationships the agent can use.

Keep the answer close to the business fact

A Hyderabad student asking about a course needs the right programme information. A Pune buyer asking about a home needs the relevant listing context. Connect the information the team would consult, then review whether the agent can retrieve it for those questions.

Graph lookup is useful because the next question often depends on the previous answer. A branch question can lead to a service question; a course question can lead to an intake question. Keep the information current and preserve an unresolved question when a fact is unavailable.

Pair knowledge with a step harness

Knowing an answer and completing an enquiry are different tasks. The graph supplies context. The step harness tracks required fields and validates progression. Together, they support a conversation that can answer questions while collecting the details your team needs.

For a property workflow, the graph can supply listing context while the harness tracks locality, property type, budget and visit preference. If a buyer changes the budget, review the updated capture and any recommendation that depends on it. The next advisor should receive a consistent record.

Use the same business context across 12 languages

Swarit’s voice offer covers 11 Indian languages and English. The caller’s chosen language should lead to the relevant business information and a useful enquiry record. Marathi, Kannada or Telugu changes how the caller speaks; it should not make the business process disappear.

Build examples around the words your customers use for courses, products, locations and preferences. Review retrieval and capture together. A natural-sounding answer is strongest when the team can connect it to the right information.

Build your first connected conversation

Start with one business area: a course catalogue, a property set or a service directory. Define the entities, relevant relationships and who maintains them. Add the required enquiry fields and decide who receives the record.

Run the conversation with a question, a follow-up that needs related context and a correction. Inspect what was retrieved and captured. Bring your business knowledge and put the first workflow in motion.

Questions before you start

Is the knowledge graph the workflow?

No. The graph supplies connected business context. The step harness separately controls required information and accepted progression.

Does lookup allow the agent to modify my systems?

The current knowledge tools are read-only. External writes, bookings or payments need separately supported integrations and explicit permissions.