AI Agents in DMs: How to Keep Leads from Chats Measurable

Daniel Kalcher

CTO, lead-metrics.com

7 min read

Until now, lead generation with ads mostly meant: click, form, call. Now TikTok and Meta are rebuilding that path. Instead of a form, an AI agent in the chat answers prospects' questions, asks qualifying questions itself and collects contact details. TikTok introduced Agentic Leads for this in early October, and Meta has offered Business Agent for WhatsApp, Messenger and Instagram since June.

For advertisers, that sounds like more leads that are better pre-qualified. But there's a catch: leads that start in a chat never reach your website. They're missing from any click-based tracking, and your CRM often just says "WhatsApp" instead of campaign and ad. This article covers what the agents can do and how to keep leads from chats measurable all the way to the deal. As of November 2026.

Key points

  • TikTok Agentic Leads: An AI agent qualifies prospects in TikTok direct messages and in lead forms on your own website. So far it's only in testing for selected advertisers and, according to TikTok, not yet available in Europe.
  • Meta Business Agent: Answers messages in WhatsApp, Messenger and Instagram, qualifies leads and books appointments.
  • The measurement problem: Chat leads don't visit your website. UTM parameters, click IDs and website tracking don't see them.
  • The fix is an unbroken chain: get the ad and the contact details from the chat into your CRM, track qualification and deals there and send results back to the platform.
  • Meta Conversion Leads only works with Instant Forms, according to Meta, not with chat leads. Chats have their own Conversions API for Business Messaging.

What the new agents do

TikTok Agentic Leads

On October 5, 2026, TikTok announced a set of AI features for advertisers, including Agentic Leads. The so-called Lead Agent is meant to run in TikTok direct messages and in lead forms on advertisers' websites, capturing and qualifying prospects in guided conversations (TikTok Newsroom).

Advertisers set up a knowledge base with approved information and define which qualifying questions matter most. The agent answers questions, collects contact details with consent and hands over the lead together with the conversation context (TikTok for Business).

So far, it's only available through tests that advertisers have to be approved for. In its Italian newsroom, TikTok states explicitly that the new agent solutions aren't yet available in Europe (TikTok Newsroom Italy). TikTok hasn't described how the leads get into a CRM.

Meta Business Agent

Meta introduced Business Agent in June 2026. It replies in WhatsApp, Messenger and Instagram and, according to Meta, can book appointments, qualify incoming leads and close sales. Businesses decide when a team member takes over (Meta Newsroom). Meta doesn't publish a list of countries where the agent is available. Some industries such as finance and health are excluded on the WhatsApp platform (Meta for Developers).

Separately from the agent, Meta has long offered click-to-message ads with the leads objective. They work with Messenger, Instagram Direct and WhatsApp, and the qualifying questions run as an automated chat (Meta for Business). Meta also notes that some messaging campaign features and metrics aren't available in Europe due to privacy rules.

Why chat leads go missing in reporting

A classic lead through a landing page leaves traces: UTM parameters in the URL, a click ID such as fbclid or ttclid, a session in your tracking. The form is matched to that session, so you know the campaign and ad.

With a chat lead, none of that happens:

Lead via landing pageLead from a chat
Visit to your websiteyesno
UTM parameters and click IDyesno
Attribution to the advia trackingonly via platform data
Contact detailsfrom the formfrom the chat or the lead export
Typical CRM entrycampaign, ad"WhatsApp", "Instagram" or no source at all

The result: the ad platform counts the lead, your CRM knows the person, but nobody connects the two. Whether the chat campaign brings customers or just conversations stays unanswered.

How to keep chat leads measurable

  1. Store the source with the lead

    Make sure every chat lead lands in your CRM with campaign and ad. For Meta's lead ads with messaging, leads go to a connected CRM, Leads Center or the download from Ads Manager, according to Meta. Check whether campaign and ad names are passed along.

  2. Collect contact details

    Email address or phone number is the bridge between chat, CRM and later deals. An agent that qualifies leads should always ask for them. With WhatsApp, you already have the phone number.

  3. Put the next step on your website

    If the agent books the appointment through a booking page or landing page, the person does reach your website after all. With UTM parameters in the link and tracking on the page, the lead is attributed like any other.

  4. Track qualification in the CRM

    The agent can pre-qualify, but sales makes the call. Track stages such as qualified, meeting and closed for chat leads just like for form leads.

  5. Send results back to the platform

    Meta offers the Conversions API for Business Messaging for this. It accepts events such as lead, qualified lead and purchase and matches them via the click-to-WhatsApp ID ctwa_clid or the user ID in Messenger and Instagram (Meta for Developers).

⚠️
Meta's optimization for qualified leads from the CRM, Conversion Leads, only works with Instant Forms, according to Meta (Meta for Developers). Through the Conversions API for Business Messaging, Meta's documentation offers purchase optimization for click-to-Messenger and click-to-WhatsApp ads. Optimization for lead events isn't described there.

What the agent optimizes for

An AI agent in the chat mainly speeds up the first response. That's valuable, because how fast a lead gets an answer often decides whether it closes. More on that in the article on speed to lead.

At the same time, the same rule applies as with any automation: the agent is only as good as the goal you give it. If what counts in the end is the number of conversations or phone numbers collected, the system optimizes for people who like to chat, not for customers. The platforms promise better lead quality, but there's no reliable independent data yet comparing chat leads with form leads. So check for yourself how many chat leads get qualified and close in sales, and compare that with your form leads.

Which roles AI agents can sensibly take over in marketing and where humans should stay involved is covered in the guide on the AI marketing team.

Checklist for chat leads

  • Every chat lead lands in the CRM with campaign and ad
  • The agent asks for an email address or phone number
  • Links from the chat to a booking page or landing page carry UTM parameters
  • Qualification and deals are tracked for chat leads in the CRM
  • Chat leads and form leads are compared separately by close rate
  • Qualified leads and purchases are sent back to the platform

How LeadMetrics helps

LeadMetrics attributes leads through your website and your CRM. For chat leads, that means:

  • Website as the bridge: If the agent sends people to a booking page or landing page, lead tracking captures UTMs and click IDs and attributes the lead to campaign and ad.
  • Matching by contact details: If leads arrive via webhook, Zapier or Make with an email or phone number, LeadMetrics matches them to an earlier visit up to 90 days back, as long as the person already entered their email or phone number there, for example in a form. Leads that never visited your website can't be attributed to an ad this way.
  • Deals and revenue: Deals from HubSpot and Close and payments from Stripe, CopeCart and Digistore24 are matched to the original lead. That way you compare chat and form leads by deals, not by conversations.

Ready for better tracking?

Discover how LeadMetrics helps you understand the true ROI of your marketing channels.

Start a 10 day trial