Google's lead-gen ad format is no longer a form — it is a conversation. At Google Marketing Live on May 20, 2026, Google introduced the Business Agent for Leads, a Gemini-powered chat unit that opens inside the search result. A prospect clicks, asks a question, and the agent answers using content pulled from your website — then, if they want to move forward, presents a pre-filled lead form. It is now in open beta for US advertisers and runs only inside AI Max for Search or Performance Max campaigns.
For B2B SaaS, this is a more consequential change than it first appears. The static native lead form was always a volume-versus-quality trade: remove friction, collect more submissions, accept lower average intent. The Business Agent inverts that logic. It adds a conversational qualification step in front of the form — the exact job a good landing page does — and it hands you a transcript and an AI-extracted summary of what the prospect actually wanted. That can raise lead quality. It can also quietly deflect buyers and expose thin website content. This guide covers what the format is, why the eligibility constraint matters, what it does to your SQL signal, and how to decide whether to opt in.
What the Business Agent for Leads actually is
The Business Agent for Leads is a conversational lead-generation ad unit: clicking the ad slides open a native Gemini chat window inside the Google search results page rather than sending the user to your site. The user types a question, Gemini answers using content grounded exclusively in your website, and the exchange continues until the prospect is ready to convert — at which point Google presents a lead form already populated with their account details. It is the same underlying lead-delivery plumbing as the existing native form, wrapped in a reasoning layer that talks to the prospect first.
Two design decisions define the format. First, grounding: the agent is restricted to answering from your own website copy, which is Google's mechanism for preventing hallucinated pricing or feature claims. That makes your site the agent's knowledge base — if a fact is not on your site, the agent cannot state it. Second, the qualification flow: rather than a passive form, the agent uses Google's reasoning models to ask and answer through a natural back-and-forth, reviewing criteria and, in some verticals, even booking consultations. For a B2B SaaS buyer with a specific question about integrations or seat pricing, that is a materially different experience from filling in name and email.
This sits alongside, not on top of, the native Google lead forms you may already run. The lead form reduces friction; the Business Agent adds an intelligence layer in front of it. Both still end in a submitted lead delivered to your systems — the difference is what happens in the seconds before submission.
How the conversational flow works, step by step
The user journey has four distinct stages, and understanding each one tells you where you can influence outcomes. Stage one: the ad renders in search with a chat entry point instead of a headline-plus-form. Stage two: the user opens the chat and asks a question in their own words — “does this integrate with Salesforce?”, “what does the Team plan cost for 30 seats?” Stage three: Gemini answers from your website content and, where relevant, asks clarifying questions that double as qualification. Stage four: when the prospect signals intent, the pre-filled lead form appears and they submit.
What you receive on the other end is richer than a form submission. Google delivers a standard lead row, the full conversation transcript, and a Gemini-extracted summary of the qualifying signals surfaced during the chat. In practice that means an SDR can open a new lead and already know the prospect runs a 30-person team, needs a Salesforce integration, and asked about SOC 2 — context that a name-and-email form never provided. The qualification that used to happen on the first discovery call has partly moved into the ad unit itself.
Why the AI Max / Performance Max requirement matters
The single most important constraint for B2B SaaS teams is eligibility: Business Agent for Leads runs only in AI Max for Search or Performance Max campaigns — standard Search is excluded. Opting into the agent therefore means opting into an AI Max or Performance Max structure, which is a bigger commitment than adding an asset. Many B2B SaaS accounts have deliberately stayed on standard Search precisely for the control it gives over match types, negative keyword lists, and query-level reporting. The agent is not available to them without migrating.
That trade-off connects to decisions you have likely already weighed elsewhere. Our analysis of AI Max search performance for B2B SaaS and the question of whether B2B SaaS should run Performance Max both apply directly here: the Business Agent is one more reason to be deliberate about that migration, not a reason to rush it. If you already run AI Max or Performance Max profitably, adding the agent is a low-cost experiment. If you do not, weigh the agent's upside against the query visibility and matching control you would be giving up to access it.
What it does to lead quality and your SQL signal
The Business Agent's effect on lead quality is genuinely double-edged, and the direction depends on how you configure it. On the upside, the conversational qualification step filters casual submissions the same way a well-built landing page does — a prospect who realizes mid-chat that the product targets 50-seat teams and they are a solo user will often exit before submitting, costing you no SDR time. The transcript and summary then let your team triage the leads that do come through far faster. This is the mechanism that can push cost per SQL down even if raw lead volume changes little.
On the downside, an aggressive or clumsy qualification flow can deflect real buyers who simply had a non-standard question, and because the agent can only ground on your website, thin content produces a thin, generic conversation that fails to convert high-intent visitors. There is also a measurement trap: if you optimize the campaign toward chat completions or form submissions, Google's bidding will chase whatever produces the most of them — not the ones that become pipeline. As we argue in optimizing for SQLs, not leads, the algorithm cannot tell a good lead from a bad one unless you feed it the downstream outcome. The Business Agent raises the stakes on that principle because it can generate more top-of-funnel volume, faster.
Your website is now the agent's knowledge base
Because the agent answers exclusively from your website, your site copy has quietly become ad creative. Every question the agent can answer well — pricing tiers, integration lists, security certifications, onboarding timelines, contract terms — has to exist in clear, specific, machine- readable form on your pages. Vague positioning copy that reads well to a human but states no concrete facts gives the agent nothing to work with. The pages that matter most are the ones a buyer interrogates: pricing, integrations, security/compliance, and comparison or use-case pages.
This is the same discipline that already pays off for AI answer engines and for the landing pages your campaigns point to: lead each page section with a direct, self-contained statement of fact, then elaborate. State plainly what the product is, who it is for, what it costs, and what it connects to. A practical pre-launch step before enabling the agent is to audit your key pages the way you would brief a new sales hire — if the answer to a common buyer question is not written down somewhere the agent can read it, the agent will either dodge the question or answer weakly, and you will lose the conversation at exactly the moment intent was highest.
Measurement: transcript, summary, and offline conversions
Treat Business Agent leads with the same closed-loop measurement you apply to every other lead source, plus two new inputs. First, capture the GCLID from the lead payload and store it with the CRM record so you can attribute downstream events back to the originating click. Second, feed the lead's progression — MQL, SQL, closed-won, each with a value — back to Google as offline conversions, exactly as described in our offline conversion import guide. Without that feedback, bidding optimizes toward the cheapest chat completions rather than the leads that become customers, and the quality upside of the format evaporates.
The genuinely new asset is the transcript-plus-summary. Use it two ways. Operationally, parse the Gemini summary fields into your lead-scoring model so routing and follow-up intensity are set before an SDR touches the lead. Strategically, read the transcripts in aggregate: the questions prospects ask the agent are an unfiltered record of buyer objections and confusion. If half the conversations stall on a pricing question your site answers ambiguously, that is a fix that improves both the agent and your human sales motion. The conversion-tracking foundation this all rests on is covered in our conversion tracking guide for SaaS.
Should B2B SaaS opt into the Business Agent now?
Opt in as a controlled test if three conditions hold: you already run AI Max or Performance Max, so the eligibility requirement costs you nothing new; your website has detailed, specific content the agent can ground on across pricing, integrations, and use cases; and you import offline conversions, so you can judge the leads by SQL rate rather than volume. For accounts that meet all three, the Business Agent is a low-risk experiment with real upside — better-qualified leads and richer context for sales, at no additional creative cost beyond the site content you should have anyway.
Hold off if you rely on standard Search for query control and are not ready to migrate, if your site content is too thin to power a credible conversation, or if you cannot yet measure cost per SQL. In every case, run the agent alongside your existing conversion path for four to six weeks and compare cost per SQL — not cost per lead, and not chat completion rate — before scaling. The format is new, the beta is US-only, and the honest answer to “does it work for us?” comes from your own downstream data, not from Google's launch framing. If you want a second opinion on whether your account and tracking are ready for it, our Google Ads audit evaluates campaign structure and conversion infrastructure as connected decisions.