AI SDR Lead Qualification: The Scored Form Your Agent Uses Before It Books the Call (2026)

An AI SDR can write the email, follow up and book the meeting, but it still needs to know who deserves the meeting. Without that, it fills your calendar with calls that go nowhere. So the real question is not which agent sends the outreach. It is what the agent reads before it books.

The answer is a structured intake: a short form with fixed questions and a score on every answer. Your agent builds it, publishes it and reads the scored answers. Then it books the call only when the lead clears the bar you set.

This guide walks through that pattern step by step. You will see what goes into the form, how your agent creates it with dForms, and how the answers reach your outreach.

What an AI SDR does, and where qualification breaks

An SDR agent takes the top of the funnel off your plate. For example, 11x describes its inbound agent Julian as "voice agents built to qualify and convert inbound demand" on its homepage. Similarly, Artisan sells Ava as "the autonomous AI BDR" on its site. You may also be building your own agent with Claude and a few tools.

Whatever you run, the job has the same three moves. First, the agent finds or receives a lead. Next, it decides if the lead fits. Finally, it books the call or lets it go.

The second move is where things break. A free-text chat or an email thread gives the agent opinions, not data. One lead says "we are a big team" and another says "around 40 people". However, the agent cannot compare those two answers in the same way every time.

Why guessing costs you

When the agent guesses, two things happen. Good leads wait while your reps take calls with poor fits. Meanwhile, the agent learns nothing, because there is no record of why it booked what it booked.

A scored form fixes both problems. Every lead answers the same questions. Every answer carries points. As a result, the decision becomes a rule you can read and change.

Why AI SDR lead qualification needs a structured form

Good qualification depends on inputs the agent can trust. A form gives you three things a conversation does not.

• Same questions for everyone. You can compare this week's leads with last month's.
• Answers with a fixed shape. A choice between "1 to 10", "11 to 50" and "51 or more" beats a sentence about team size.
• A score that explains itself. When a lead gets booked, you can see which answers earned it.

This is the logic a human SDR uses, written down. If you want the full framework for picking criteria, read our guide on how to qualify a sales lead. Here we focus on turning those criteria into a form your agent can build and read.

Form first, then your AI SDR books the call

The form runs before the booking, not after. That way the calendar link appears only for leads that score high. Everyone else gets a polite ending and an email follow-up.

What goes into a lead qualification form your agent can score

A good lead qualification form is short. Five to seven questions is a sensible range, because every extra question costs you completions. Keep only the questions that change your decision.

The questions

Most B2B teams ask about company size, the problem, timing, budget and role. Use choice questions where you can, since choices are easy to score. Keep one open question for context, unscored.

The points

Next, give each answer a weight. Here is an example for a B2B software team. The numbers are an illustration, not a benchmark, so set your own.

The cutoff your AI SDR uses

Finally, decide what "hot" means. For example, you might send any lead at 9 points or more to the calendar. Leads below that get a resource and a follow-up. Write the rule down, because your agent will apply it exactly as written.

You can also add conditional logic. For instance, a lead who picks "Just exploring" can skip the budget question.

How your agent builds the form with dForms, step by step

dForms is the form builder for AI agents, the one your agent actually wants to use. With one API key, your agent builds the form, publishes it, scores every answer and sends hot leads to your calendar or CRM. It is also free and open source, with unlimited responses. Here is the workflow.

Step 1: Connect your agent

You have two ways in. First, create an API key on the Developers page of your workspace. The key is shown once, and your agent sends it as a Bearer token. Second, if you work in Claude, add the dForms MCP at mcp.dforms.ai. From there, your agent can create, publish, update and revert forms, list responses and read analytics.

A key acts as the person who created it, with that person's role. Therefore, create it from an account with the access you want the agent to have. For a broader view of this setup, see our post on the AI form generator for agents.

Step 2: Ask for the form in one message

Next, describe the form in plain words. Include the questions, the points and the cutoff. You can copy this prompt and adjust it:

Create a dForms lead qualification form called "Book a call".
Ask these questions, in this order:
1. How many people are on your team? Options: 1 to 10 (0 points),
   11 to 50 (2 points), 51 or more (3 points).
2. When do you want to start? This month (3), This quarter (2),
   Just exploring (0).
3. Who decides on this purchase? Me (3), My manager (2), A committee (1).
4. Do you have a budget for this? Yes (3), Not yet (1), No (0).
   Skip this question if they chose "Just exploring".
5. What problem are you solving? Open text, not scored.
Then ask for name and work email.
Send leads with 9 points or more to my calendar.
Leave the form as a draft and send me the preview link.

Step 3: Review the draft, then publish

Edits live in a draft until someone publishes them. That is why the prompt asks the agent to stop at the draft. Keep in mind that dForms does not force an approval step: an agent with a key can publish. So "a person approves the publish" is a habit you set in the prompt, not a lock in the product.

Once the preview looks right, tell the agent to publish. If a change goes wrong later, the agent can revert the published form.

Step 4: Read the scored answers

Now the form collects answers, and each one carries its score. Your agent can list responses through the API or the MCP. It can also export a CSV or read the form analytics.

Step 5: Route the hot leads

Leads above your cutoff go to your calendar. In addition, dForms can sync contacts to HubSpot and send each response to a webhook. That last part is how the form talks to the rest of your stack.

Wiring your AI SDR to the form: three patterns

Your agent does not need a special plugin to use the form. Instead, pick the pattern that matches how your agent already works. We are describing patterns here, not built-in integrations with any SDR product.

Pattern 1: Webhook, so the agent reacts

Point a webhook at the endpoint that starts your agent. When a lead submits, the response arrives with its answers. Then the agent reads the score and picks the next move.

Pattern 2: API, so the agent checks on its own schedule

Alternatively, your agent can pull new responses through the API with its key. This suits agents that run in batches, such as a morning review. Each key allows 120 calls per minute, which is plenty for a polling job.

Pattern 3: MCP, so you can ask in plain words

If you live in Claude, the MCP turns the form into something you can talk to. For example, you can ask "which leads from this week scored 9 or more?" and the agent lists them. Our post on using Notion MCP as agent memory with forms shows how to keep those answers around between sessions.

Which pattern to choose

You can also combine two: a webhook triggers the follow-up, and you use the MCP to review the week.

How AI lead qualification stays honest over time

Good AI lead qualification is not set and forget. Every few weeks, check the results. Did the booked calls turn into real conversations? Did good leads land below the cutoff?

When you spot a gap, change one weight or one question, then let the form run again. Because every rule lives in the form, the change is visible and easy to undo.

Ready to try it? Start from the dForms homepage and let your agent build the first version.

Frequently asked questions

What is an AI SDR?

It is an agent that does sales development work: it finds or receives leads, reaches out, qualifies them and books calls. Some teams buy a dedicated tool, while others build their own agent with a model like Claude.

Does dForms integrate directly with 11x or Artisan?

No, dForms has no built-in integration with any SDR product. Instead, it offers a webhook, an API and an MCP. Any agent that can call one of those can use the form.

Can my agent publish the form without me?

Yes, an agent with an API key can publish. dForms keeps edits in a draft until publish, and a published form can be reverted. If you want to approve first, tell your agent to leave the form as a draft and send you the preview.

How many questions should a lead qualification form have?

Keep it to the questions that change your decision, often five to seven. Every extra question adds friction, so use conditional logic to skip questions that do not apply.

Do I pay per response?

No, dForms is free and responses are not counted or billed. It is also open source under the MIT license, so you can self-host it.

Can I use the same form for hiring?

Yes, the same pattern works for candidate screening. You score answers about experience or availability instead of budget and timing.