A B2B technology hardware team in India uses an AI voice agent to qualify business buyers on budget, authority, need and timeline (BANT) before an account manager ever calls. The agent asks quantity, budget per unit, timeline and who decides, one question at a time. It finds the right contact if it reaches the wrong person, and closes by promising a quote from an account manager within a day or two. Account managers now call buyers who are already qualified. This is the AI sales agent running a classic B2B framework.
Build a BANT agent: Start free on Edesy — Rs 50 free credit, no demo needed.
The BANT call, question by question
| BANT | What the agent asks | Field |
|---|---|---|
| Need | "What are you looking to buy, and roughly how many units?" | product, quantity |
| Budget | "Do you have a budget range per unit in mind?" | budget_per_unit |
| Timeline | "When do you need them: this month, or within 90 days?" | timeline |
| Authority | "Are you the person who approves this purchase, or should our account manager also speak to someone else?" | decision_maker |
The close:
Thank you. An account manager will call you within 24-48 hours with a quotation. Is this the best number to reach you?
Finding the right person
B2B lists are messy: reception numbers, old contacts, people who moved roles. When the agent reaches someone who doesn't handle purchases, it asks:
Could you tell me who handles IT purchasing at your company, and the best number to reach them?
The new contact is saved and called separately. A wrong-person call still produces a lead.
What went wrong at first, and the fix
In early calls the agent hung up too soon. It treated a polite "thank you" in the middle of the call, or "no" to a choice question ("this month or 90 days?"), as the end of the conversation. The fix was a strict closing checklist:
- The call has run a minimum time.
- The closing question was asked in exactly those words.
- The buyer gave an explicit "no" to "anything else?".
- The goodbye is spoken as its own turn.
- Only then does the agent end the call.
"Not sure" answers are now marked callback pending, and the call continues. The full story is in why your AI voice agent hangs up too early.
Other rules that helped
- Ignore "this call is being recorded" announcements and business call-screening assistants; wait for a person.
- Switch to Hindi only after two full Hindi sentences, not after one "haan". Keep verb forms consistent with the agent's voice.
- Always speak before using a tool, so the buyer never hears silence. See how to fix dead air on AI calls.
- "Are you an AI?" Answer honestly and briefly, and continue.
- Remove long pauses ("...") from the script; they sound like the call dropped.
Where the answers go
Every call saves need, quantity, budget, timeline, decision-maker and a summary. Push them to HubSpot, Salesforce, Zoho CRM or GoHighLevel, or to a Google Sheet, so the account manager opens the record and sees the full picture. Hot leads can be transferred live during working hours.
Cost
On Edesy, calls cost Rs 4-6 per started minute depending on your prepaid top-up, and unanswered calls are free. A two-minute BANT call costs about Rs 8-12 when answered.
Try it on your pipeline
Create a free account, start from the lead qualification template, and call your own phone as a test buyer. Also read how a manufacturer qualifies leads with AI calls and AI outbound calling for small business sales.