How signal-based targeting built an AI-visibility agency's go-to-market from scratch

Aug 13, 2026

By Artyom Jurkevich, Founder & CEO, RevSculpt

An agency came to us that audits how a brand appears inside AI answers: ChatGPT, Gemini, Perplexity. They score it against competitors and help the company fix the gap.

When they came to us, they had no packaged offer: no positioning, no landing page to explain the service, and no outbound built to sell it.

The goal was to build a steady flow of discovery calls with marketing leaders at B2B SaaS and e-commerce companies generating over $5M in revenue. To achieve this, we built the full go-to-market system, including positioning, a landing page, sending infrastructure, and an outbound engine.

Results:

An ICP database of ~60,000 qualified companies, built from scratch.
20,000 to 60,000 outbound emails a month, delivered clean.
A peak month of 80 held meetings on the first message system, proof the top of the funnel could fill at volume.
After we rebuilt the message: 20 to 30 held meetings a month, higher up the funnel, closing on retainer instead of one-off work.
One senior operator ran the whole channel, with no SDR team hired.

Diagnostic: why there was no pipeline yet

Before contacting a single prospect, we ran a full audit of the current setup and built a strategy meant to fill the client's pipeline. There were three problems, and each one made the others worse.

No packaged offer. The audit existed as something the founder could do, not as a product someone could buy: no positioning, no landing page to send a prospect to. The infrastructure was a standard startup setup — you simply can't run volume outbound from it without burning your main domain.
No portrait of the decision-maker. The team thought in terms of a company list: pull everyone above $5M in revenue and email them. But they couldn't answer the first question, why $5M at all - and a list of companies isn't a buyer yet. We needed to understand who inside actually makes the decision and what they struggle with.
The message talked about the product and missed the prospect's situation. The team took the usual market route: bury the person in a pitch and ask for a meeting. What it needed was market segmentation, with its own offer and its own specific pain for each segment - the narrower the segmentation, the higher the conversion, always.

How we built the machine: offer, infrastructure, signals

We built the go-to-market in the order it actually has to happen: first the thing we sell, then the system to sell it at volume, then the targeting that decides who even receives an email.

We packaged the offer first

We positioned the audit as a concrete product with clear boundaries and built a landing page that explains the visibility problem and makes the audit the obvious next step. Now the email had somewhere to lead.

Infrastructure with headroom

All outbound ran on secondary domains with full DNS authentication, and we provisioned surplus capacity so that even 60,000 emails a month never leaned on a single sending stream.

If one stream degrades, volume redistributes with no pause in the campaign. That headroom is what let us hold this volume without burning the setup.

Signal-based targeting instead of a list

We start every campaign from what is happening to a company's visibility right now, and we write only when there is a real reason to.

The signal model was built around the one pain this buyer genuinely feels: their potential customers increasingly look for solutions not in Google but in AI - ChatGPT, Gemini, Perplexity. And when the engine names competitors instead of their brand, the company is losing customers at the exact place where the decision is now being made.

The three signals we actually tracked

The temptation in this space is huge: track everything, every hire and every funding round. We tracked three signals, because signal quality matters more than quantity, and because all three tie back to one real pain, whereas hires and funding rounds are visible to anyone with a scraper.

A competitor pulling ahead in AI answers. A prospect's direct competitor starts showing up first in ChatGPT or Perplexity for the queries that matter. For the prospect, that means a customer being sent to the competitor.
A weakness on the prospect's own site. Low AI-readiness: thin content structure, missing schema, pages not optimized for how AI engines read them.
A decline in organic traffic. A public, verifiable downward trend the prospect has usually already noticed but cannot yet explain.

On its own, each signal stays a single data point. Together they form a specific picture: competitors are winning the answer, the site is not built for these engines to read, and traffic is already paying for it.

That is a reason to talk this quarter, not someday.

The AI auditor that scored every prospect

What made this work at volume was a custom auditing agent we built on Claude, and it is also the part most agencies have nothing to copy it with.

For each prospect the agent ran a scraper to pull the site, then an analyzer that scored the pages against our knowledge base: content structure, schema markup, and how well the page was optimized to appear in AI answers. The output was a concrete AI-readiness score for each prospect.

That score is what made the email specific. It drew on the real state of that prospect's own pages.

This is what "AI-powered" looks like for us in practice: a system that does individual research on every prospect. No human team does that research by hand at 60,000 emails a month.

Channels and the message

The email led with the prospect's own situation: the competitor that had pulled ahead, and the weakness the auditor found. Plain text, under 75 words, one concrete reason to reply.

The point of every send was for the recipient to feel the email was written after someone looked at their site. It was.

Once a prospect showed interest, the handoff was clean: the agency's SEO team picked it up from the discovery call, took GA4 and Search Console access on the spot, and moved into the work.

Our job was to fill the top of that funnel with the right conversations, reliably.

The build, phase by phase

Phase one: package and launch. Positioning, the landing page, secondary-domain infrastructure, and the auditor agent. Nothing goes out until there is an offer to send people to and a system that will hold the volume.
Phase two: volume. With the sending machine live, we climbed toward 60,000 emails a month, and the first message system delivered its peak: 80 held meetings in a single month. That proved the top of the funnel could be filled at scale.
Phase three: tune for quality. 80 discovery calls a month is a lot of conversations, and not all of them were the right ones. So we rebuilt the message, and the shape of the funnel changed with it.

How we rebuilt the message and started closing more

We tested three versions of the opening message, and the result is the most useful thing in this case.

The aggressive version. ("your competitors are first in AI, you are invisible, we will fix it") produced the most discovery calls. It also pulled in every curious reader, so many of those conversations did not convert.
The consultative version. led with data on the prospect's own site and offered a plan to improve it. It produced fewer meetings, in the 20 to 30 range instead of 80, but they sat higher in the funnel and closed on retainer, where before it had been a one-off audit.
The third version. tried to explain everything before the call. Conversions dropped roughly tenfold, and we turned it off.

With a marketing-leader audience, a message that filters is worth more than a message that maximizes replies. The consultative version said less and qualified harder, and that's what turned meetings into retainers.

The results in context

The 80-meeting month showed the ceiling on volume. Moving to the consultative message showed something more valuable: a funnel you can tune.

Once the message did the filtering, the same infrastructure produced a steadier flow of 20 to 30 held meetings a month with prospects who were ready for an ongoing engagement.

We report in held meetings, not booked. At this volume, that difference matters.

Client quote

"We came to you with no packaged product: we had the expertise and no clear way to sell it. The first surprise was that on the discovery call the prospect already understood why we were reaching out, because the email referenced their own site. Within the first month we saw a predictable flow of the right conversations for the first time, rather than random inbound." - Lina Ivanova, COO, BetterAnswer.ai

What the in-house version misses

Infrastructure comes first. Sending 60,000 emails a month without losing your domains takes its own discipline: which domains to warm, at what rate, how to spread volume, when to rotate. Most teams learn this the moment reply rate drops to zero, and by then the domain is already gone.
Tracking many signals is easy for anyone. It is far harder to know which of them actually move a decision, and for that you need to understand the buyer's business. That a competitor winning AI answers, plus weak AI-readiness, plus falling traffic together create urgency is visible only to someone who understands how that buyer's business works.
Individual research at volume rests on a system. The auditor scored every prospect's site separately. No team writes 60,000 emails by hand where each one references the recipient's real pages. A build like this is done once and correctly, or it does not hold.
The right message filters. The least obvious lesson here: the version that produced fewer meetings produced better clients. Until the message qualifies for you, volume stays a vanity metric.

When this approach doesn't fit

Product-market fit still in flux. If the offer shifts every couple of months, outbound only amplifies the confusion. Meetings will happen; conversions will not.
A visibility gap you can't yet measure. The model runs on real weaknesses: competitors ahead, weak AI-readiness, falling traffic. If a prospect is already in good shape on all of that, there is nothing to act on.
An ICP that hasn't been defined yet. The 60,000-company database worked because the target was specific: B2B SaaS and e-commerce above $5M in revenue, losing AI visibility. Without that specificity, the same infrastructure just sends more generic email.

FAQ

What is AI visibility (AEO/GEO), and why does it matter for B2B SaaS? AI visibility, also called answer engine optimizations (AEO) or generative engine optimizations (GEO), reflects how often and how prominently a brand appears inside AI answers on ChatGPT, Gemini, and Perplexity. It matters for B2B SaaS and e-commerce because buyer research has shifted: 51% of B2B software buyers now start their research with an AI chatbot more often than Google (G2, 2026). The brands an AI engine names when a buyer asks are the ones that make the shortlist, the rest don't get a conversation at all.

How do you run outbound for an SEO or AI-visibility agency without sounding generic? By replacing the contact list with a research step on every single prospect. In this project, a custom auditor agent built on Claude scraped each prospect's site, then scored the pages against a knowledge base covering content structure, schema markup, and AI-readiness. The email opened with that specific finding, the competitor pulling ahead in AI answers, or the exact gap on the prospect's own pages, instead of a generic pitch about the service. At 60,000 emails a month, that per-prospect research is what a small human team can't do by hand, and it's the difference between an email that reads as an ad and one that reads like someone actually looked at the site.

What signals show that a company needs help with AI visibility? Three signals, tracked deliberately instead of monitoring every hire and funding round: a direct competitor pulling ahead in AI answers for the queries that matter, low AI-readiness on the company's own site (thin content structure, missing schema), and a public, verifiable decline in organic traffic. Any one of these alone is just a data point a prospect may not have connected yet. Together, they describe a specific and urgent situation, which is what turns "someday" into "this quarter."

How many outbound emails a month can you send without hurting deliverability? In this case, the program ran between 20,000 and 60,000 emails a month on secondary domains with full DNS authentication and surplus sending capacity, so no single stream ever carried the full volume. If one stream's deliverability degrades, volume redistributes to the others automatically, without pausing the campaign. There's no universal ceiling, the limit is set by how disciplined the domain warm-up and volume-spreading are, not by a fixed number of emails.

Should a cold email push for the meeting or lead with value? It depends what you're optimizing for. A harder push produced more discovery calls in this case, the aggressive opener pulled in the most volume, curious readers included. A consultative message with a concrete, prospect-specific plan produced fewer meetings but better ones, and they closed on retainer instead of one-off work. If the goal is ongoing engagements rather than raw call volume, the qualifying message wins.