Leads, qualification and cost per lead

Your agents talk to people on your website chat, your chat link, WhatsApp and email. Leads (in the sidebar, under Operate) keeps everyone who could become a client in one place: where they came from, whether they fit, what happens next, and what each one cost you in credits.

Anyone in the workspace can see Leads. Moving a lead, setting the next step and writing a note need the Owner, Admin or Builder role.

Where leads come from

A lead is saved when:

• an agent with Save a lead (under Tools) learns what someone needs and a phone number or email;
• a visitor fills in the name and contact form before a website chat or chat link conversation;
• an agent books a call for someone who wasn't a lead yet;
• a webhook from another tool starts an agent that saves the lead, such as IndiaMART's Lead Manager. See Get IndiaMART leads.

One lead per person. Leads belong to the project, not to one agent. When the same person writes again, with the same email or phone number (however it's typed, with or without +91), the new conversation is added to their existing lead instead of making a second one. Their newest details replace older ones, and an empty answer never wipes one out. Leads from test chats are kept apart and marked Test.

Stages

• Once someone on your team moves a lead, the agent leaves it where you put it.
• A Lost lead that writes again is flagged Came back. See Came back.
• Test leads never move by themselves.

Lost, with a reason

When you mark a lead Lost, pick why:

• Not a fit
• No budget
• Chose someone else
• No reply
• Other, with a few words on why

The lead shows its reason, and Analytics → Leads and cost counts lost leads by reason, so you can see whether you lose more on price, on timing or to silence. You can change the reason of a lead that's already Lost.

Suggestions to close

The AI never closes a lead. It can suggest closing one:

• Not a fit: the check found that a criterion that matters isn't met. If you move the lead anywhere other than Not a fit, you've disagreed, and the suggestion goes away.
• No reply: two follow-ups went out since the customer last wrote, and they haven't answered.

A suggested lead shows Suggest closing on its card. Open it and select Close as Lost… to close it with the suggested reason, or close several at once:

1. In Today, select leads to close. The list shows only the suggested leads.
2. Tick the ones to close, or select Select all.
3. Select Close … suggested as Lost, check the names and reasons, and confirm.

Only leads the AI still suggests closing are closed. Anyone who can move leads (Owner, Admin or Builder) can do this.

Came back

When a Lost lead writes again, with a new enquiry or a new message in its conversation, it's flagged Came back, with a line in its history. Today counts them. The stage stays Lost until you act:

• move it to another stage, such as New; or
• select Keep it Lost to clear the flag.

Merge duplicate leads

The same person can end up as two leads, for example one saved with a phone number and another with an email. An Owner or Admin can merge them:

• From the list: in List view, tick two or more leads and select Merge.
• From a lead: open it, find Merge with…, search for the other lead and select Merge….

A preview shows which value each field keeps. The newest value that isn't empty is picked for you; choose another lead's value for any field, the stage included. Then confirm that you understand it can't be undone, and select Merge.

• The oldest lead is kept. The others' conversations, history, notes and meeting suggestions move to it.
• Nothing is lost. The other leads are kept, hidden from the list, pointing to the one you kept. An old link to one of them opens the kept lead, and a new enquiry with their old email or phone finds it too.
• It's recorded. The history says who merged which leads. There's no undo in the app.

What makes a good lead

Each agent that saves leads has up to five criteria, plus a sentence on who a good lead is. Open the agent's Build tab and find What makes a good lead. Agents made from a template start with that template's suggestions, such as need, budget, timeline, who decides and how to reach them for the Lead qualifier. Change anything to make them yours; Use the template's again brings the suggestions back.

For each criterion you set:

• what it is, such as "Budget";
• when it counts as met, such as "₹50,000 or more";
• how the assistant might ask, such as "Do you have a budget in mind?";
• how much it weighs: Must have, Important or Nice to have.

The agent asks about these naturally while it helps, one or two at a time and never as a form.

How a lead is checked

When a lead is saved, and again once the customer has said more, a model reads the conversation and answers each criterion met, not met or not known, with a quote. Then:

1. The quote must be the customer's own words. It's kept only if it appears word for word in one of the customer's messages. Case, spacing and the style of quote marks don't matter; anything else does. Words the assistant said, a paraphrase, a translation, or words joined from two messages are thrown away, and that criterion becomes not known.
2. The verdict and the score are worked out by code from the weights, never by the model:
   • Not a fit: a Must have isn't met, or less than 40% of the weight is met and something isn't.
   • Qualified: 70% or more of the weight is met.
   • Partly: 40% or more is met.
   • Not checked yet: too little is known so far.
   • The score is 1 to 5, from the share of the weight that's met. A Not a fit lead scores at most 2.

A lead's card shows its two strongest reasons, for example ✓ Budget: “we can spend about 60k”. Hover a reason to see what counts as met. Open the lead to see every criterion, and select Check again to run the check now.

Why quotes?

A score you can't check is a guess. Tying every tick to the customer's own words means you can see exactly why a lead was rated as it was, the AI can't talk a lead into Qualified, and a wrong verdict is easy to spot and fix: change the criterion, or move the lead yourself.

Each check is a small model call, charged like any other and counted in that lead's cost. A lead is checked at most 6 times a day.

The Leads page

• Today shows new leads today, follow-ups due, follow-ups to approve (select either to see only those) and calls booked for today.
• Board has a column per stage. List shows the same leads in rows. On a phone you always get the list.
• Filters: stage, where they came from (the channel, and the place tag or code on your chat link), agent and verdict. Search finds a name, email or phone number.
• A card shows the name, where they came from, what they need, the verdict and score, the top two reasons, the next step with its date, and the cost so far. When an email to them waits for your OK, it says Follow-up to approve, with Approve, Edit and Skip.
• Open a lead to see its conversations, every criterion with its quote, the history (who did what: the AI, your team or Dhanur AI), the next step and a note. Move it with the stage buttons. Won asks for an optional value, and Lost for a reason.
• Time to first reply shows on each lead: how long after their first message they got a reply.

The agent suggests a next step after each check, such as "Call to confirm the budget", with a date. Once you set your own, it stops suggesting. See Next step for the ones it can send.

Next step

Each lead has one next step at a time: what happens next, when, and how it reaches them.

• Email: from your connected Gmail, or from the agent's own email address if you haven't connected Gmail. The AI drafts the email once the conversation goes quiet.
• Call: a reminder for your team. Nothing is sent.
• Nothing to send: for example, waiting for a booked call.

Open the lead to see the draft under Next step:

• Approve and send sends it. It's the same approval as in Tasks, decided by you, so your workspace's rules apply. Approved outside sending hours (Monday to Saturday, 10:00 to 18:00 India time), it waits for them.
• Edit lets you change the step, the date, how it reaches them, and the email itself before it goes. You can also write an email yourself, or select Write it for me, which shows about what it costs before you select it.
• Skip drops it. You can say why; it's kept in the lead's history.
• For a call, Done marks it done.

How each email goes out is the agent's setting Each email to a lead, under Build → Emails to leads and reminders:

• Ask first (the default): it waits for your OK. On its due day it also shows in Tasks.
• Ask, then send after some hours: it goes by itself if nobody decides.
• Sends on its own: on its due day, in sending hours.

What keeps it safe:

• An email only ever goes to that lead's own address.
• The agent's Email must be on (Build → Email), so their reply comes back to the conversation.
• A reply stops anything waiting. When they answer, by email or in any chat with the agent, a follow-up that was waiting is cancelled and the reply is recorded in the lead's history. The agent answers them as usual.
• STOP means stop. A reply of STOP, or the unsubscribe link, ends emails from every agent in your workspace to that address.
• At most 2 follow-ups per lead written by the AI (you can choose fewer). After an unanswered email, the AI proposes the next follow-up 3 days later. Emails you write yourself don't count.
• Never two follow-ups for one silence. If the agent also has Follow up when a customer goes quiet on, the lead's next step takes over once an email to them is written here, and follow-ups of both kinds count towards the 2.
• The daily limit and sending hours are the same as for outreach from your Gmail.
• Test leads never get an email.

Each email's history, sent or skipped, is under Emails before on the lead. Each one sent is charged at the outreach email price, and writing it uses AI like a short reply, counted in that lead's cost.

Speed-to-lead

A lead from a form, HubSpot, a webhook or your API gets a first reply drafted within minutes: a short, personal email about what they asked, offering free times when the agent books calls. A lead from a chat already got a reply, so it gets none.

• Off (the default): the reply waits for your OK, and Owners and Admins get an email saying it's ready. It goes out in sending hours.
• Reply to new leads at any hour (under Build → Emails to leads and reminders): the reply sends itself 5 minutes after it's written, day or night, unless you stop it.

To get leads from a form into Dhanur AI, use a webhook, HubSpot, or the form on your website chat. Each lead shows its time to first reply, from any channel.

An agent's Leads tab shows its recent leads and opens this page filtered to that agent. That includes leads another agent saved first that this agent talked to later.

Meetings with a lead

When the meeting notetaker records a meeting with a lead, the meeting is added to the lead. That happens when the meeting's link came from the lead's booking, or when the lead's email is among the people in the meeting.

The meeting is checked against the agent's criteria, and for the lead's need, budget and timeline. What it finds waits under From meetings in the lead, as suggestions such as Budget: ₹80,000, with the quote and the meeting's date. Nothing changes until you accept it:

• Accept sets that field, or that criterion's answer and quote. The verdict and score are then worked out again from the weights.
• Reject leaves the lead as it is.

The quote rule applies to the transcript too: a quote is kept only if it's word for word in what was said.

• When the transcript names the speakers, only the lead's own words count, and the suggestion says their words in the meeting on 2 Oct.
• When it doesn't, the whole meeting counts, and the suggestion says from the meeting on 2 Oct, so you know it may be someone else's words.

A check you accepted from a meeting stays when the conversations are checked again later, unless they answer that criterion themselves. Each meeting check is a small model call, charged like any other check and counted in the lead's cost.

The morning brief and what your AI team noticed

On Home, Your AI team noticed shows up to three things worth doing today, each with one button:

• a lead who has waited a day or more for a reply;
• a lead who has been Qualified for 2 days with no call booked;
• emails to leads waiting a few hours for your OK;
• a conversation your agent handed to your team;
• a lead you marked Lost who wrote again;
• a festival post waiting for your OK;
• leads the AI suggests closing;
• a lead you won without a deal value.

We find these with simple rules on what's already in your workspace. Finding them uses no AI, so it isn't charged. Problems, such as a task that failed, stay in Needs attention.

Nothing happens until you choose. A button that writes, such as Write a reply, shows about what it costs first: the price of one reply. It works like Write it for me on the lead. The AI writes the email, the lead opens with it under Next step, and it waits for your OK like any other (unless you set that agent to send emails on its own). Other buttons open the right page.

Select Not now to hide a card for a week, for everyone in the project. It comes back sooner if something new happens, for example the lead writes again. Viewers can open each card; the buttons that write, and Not now, need the Builder role or higher.

The morning brief. At 9:00 each morning, Owners and Admins get one email per project. It starts with what your AI team noticed, each with a link into the app, then:

• new leads since yesterday;
• leads newly qualified;
• calls booked for today;
• follow-ups due, and messages to leads waiting for your approval;
• leads waiting over 24 hours with no reply and no next step.

The email itself never writes anything, and you aren't charged for it. It never includes phone numbers, email addresses or quotes; numbers in what someone needs are hidden. On a day with nothing to say, it doesn't come. To stop it for yourself, turn off Morning brief at the top of the Leads page. It's your own setting; your teammates keep theirs.

Cost per lead

Every lead shows what it has cost so far: the credits every conversation with that person used, including checking them against your criteria. It's worked out when you look, from the same numbers as Tasks, so it never drifts.

Analytics → Leads and cost shows, for leads created in the period:

• leads, qualified, calls booked and won;
• cost per lead, per qualified lead, per booked call and per won client;
• the same broken down by where they came from, by agent and by week;
• the average time to first reply, by where they came from;
• lost leads by reason;
• once you enter values on Won leads, what you won against what you spent.

Home shows a line for the week, such as "12 leads · 5 qualified · 3 calls booked · ₹46 spent (₹15 per booked call)", once you have leads.

Export

Select Export CSV on the Leads page to download the leads that match your filters, with the stage, verdict, reasons and quotes, next step, value, cost and a link to each lead. Excel opens it with Hindi and other scripts intact.

What Viewers see

When an agent has Hide contact details from Viewers on, Viewers see phone numbers and email addresses hidden on its leads, in the quotes, the history and the export too, and can search those leads by name only.

HubSpot

To keep HubSpot in step, turn on Update lead status in HubSpot in the agent's CRM settings. See Copy leads to HubSpot.

Start an agent when a lead moves

A lead changes stage is an event another agent can start from, for example to email your sales team when a lead is qualified, or to post a won deal to your other tools with a web request. See Start an agent when something happens.