About us
A Demo of Salespeak for B2B Marketers with Omer Gotlieb
Airfleet’s demo with Salespeak founder Omer Gotlieb: a B2B marketer’s look at how the platform closes the gap between buyers, AI agents, and the slow “request a demo” wait.
Quick Jump
The B2B buyer journey is broken. Most marketers know it.
A buyer lands on your website after you spent real money getting them there. Then they hit a wall: dozens of generic pages that look like every competitor, or a “request a demo” button that starts a slow chain of calls beginning with your most junior sellers before they get a single useful answer.
AI did not create this problem. It sped it up and punctuated it with a giant exclamation mark. Buyers now expect instant, expert-level answers in seconds. They will not wait days to find out whether your product is worth a closer look.
In this Airfleet demo, Salespeak founder and CEO Omer Gotlieb walks through how his product closes that gap. Camela Thompson, Head of Content Marketing at Airfleet, hosts.
TL;DR
- Salespeak’s pitch: put an intelligent conversation between the buyer and the company, whether that buyer is a human visitor or an AI agent like ChatGPT crawling your site.
- The platform splits into two products: Visitor Conversations (an on-site AI chat) and Agent Conversations (LLM optimization at the CDN level). They can be bought separately or bundled.
- Setup is fast (live in an hour for basic, a few days for a thoughtful launch). Pricing is month-to-month with free tiers in both products.
- Best fit: B2B marketing teams ready to maintain a knowledge base. Worst fit: companies with outdated websites and no resourcing to clean them up.
How Salespeak Solves the Broken B2B Buyer Journey Problem
Omer opens the demo with a personal story. The morning of the recording, he was researching alternatives for his company’s HR and payroll vendor. He saw prompts to leave his email on a competitor’s website. In return, he would expect two emails, a LinkedIn message, and a phone call. All of them pushing him toward a 45-minute call with a product expert before he could find out whether the product was even worth a closer look.
I need two minutes to understand whether I want to go further or not. Give me those two minutes, and then I’ll be able to spend time with you. But I don’t have time to spend 45 minutes.
Salespeak’s approach to this problem is at least threefold.
1. Optimize content for the AI agents already reading your site.
Most B2B websites are written for humans and tuned for SEO. Increasingly, the first visitor to your site isn’t a human; it’s an AI agent answering a buyer’s question in ChatGPT, Claude, Perplexity, or Gemini. Those agents usually hit blog pages first, and blog pages weren’t written for this. As Omer puts it: “It doesn’t have enough information about your competitive advantage. It doesn’t have enough information about your ROI. It mainly was written for SEO.”
Salespeak installs at the CDN level, identifies AI agents in real time, and serves them structured content (FAQs, grounded answers from your knowledge base) instead of the marketing copy meant for humans. Google’s bot sees no change. Human visitors see no change. Only the AI agent gets content formatted for its job.
2. Train an AI agent to answer buyer questions on your website.
Salespeak’s other half is an AI chat layer that lives on the site itself. It learns from your website, internal documents (decks, battle cards, ICP), and video content. It answers technical, integration, and compliance questions in real time, qualifies the buyer based on rules you set, and routes high-intent buyers to a calendar via Calendly or Chili Piper.
The product isn’t a chatbot in the old sense. It’s not following a script tree. Omer explains the design choice plainly: “We don’t upload the content of the document as we don’t upload the content of the website. We actually analyze it. We analyze the content, we analyze the document, we analyze the video, and then we convert that to a customer-facing content.”
3. Continue the conversation in the buyer’s inbox.
B2B evaluations don’t finish in one visit. If a buyer leaves an email mid-conversation but doesn’t book a meeting, Salespeak’s email agent picks up where they left off. It can send a follow-up summary, answer replies, and run on the same knowledge layer as the on-site chat. You can also point it at a shared inbox (example: [email protected]) and the AI will respond to incoming emails automatically.
None of this is meant to replace a sales team. Omer is explicit: “I’m not undermining the power and importance of human relationship. Human relationship is critical in making decisions.” The pitch is that Salespeak handles the research stage of the buyer journey, where most websites currently fail.
The Salespeak Product
How hard is it to get Salespeak running?
Based on the demo: not very. The implementation path looks like this.
- Point Salespeak at your website. Choose which pages to include (you can exclude careers, privacy policy, and the like). The AI ingests the content and produces hundreds or thousands of question-and-answer pairs, each scored for readiness. In the demo, the Wiz sample instance scored 93% from website training alone.
- Upload supporting documents and videos. Sales decks, battle cards, ICP. Video matters because the AI can play a relevant clip mid-conversation, which Omer says converts better than text alone.
- Set the AI’s goal and qualification rules in plain English. The goal is what a good conversation with a qualified buyer should end in (usually a demo booking). The qualification rule decides who reaches that goal. “Don’t qualify companies under 500 people,” for example. Or, as Omer offers in the demo: “It could be any company that a CMO name starts with a D. Whatever you want, just write in plain English.”
- Test in the simulator before launch. Omer is direct on this point. “Nobody’s going to put anything on the website, definitely not customer facing, without testing it first.” You run as many buyer simulations as you want, correct any answer you don’t like, and click update.
- Deploy via JavaScript snippet or Google Tag Manager.
Technical setup is fast. Omer cites customers going live the next day, though he doesn’t recommend it. His guidance: “I do recommend to spend a few days, not more than that, but not just 30 minutes, just to make sure that the simulations are running properly, that you feel that the responses are great.” A couple of days to a couple of weeks is realistic, depending on how thorough your review is and how clean your existing content is to start with.
What do you get with Salespeak?
It depends on the bundle you buy. The product splits into two major categories that can be bought bundled together or separately. Here’s how Salespeak’s own AI describes them when asked:
“Visitor Conversations refer to interactions with human visitors on your website, handled by the Website AI Agent. This agent answers technical questions, qualifies leads, and books meetings in real time. The Website Inbound Agent page best represents this category, focusing on replacing static forms and engaging buyers directly.
βAgent Conversations are interactions with AI agents (like ChatGPT, Claude, Perplexity, Gemini) that scan your site for information. The Agent Interaction Platform and Agent Optimizer pages highlight this category, showing how Salespeak serves optimized content to AI agents and captures buyer signals from their research.”
Product 1: Visitor Conversations
The on-site agent. What it does:
- Answers technical, integration, and compliance questions in branded chat, voice, or email.
- Captures buyer criteria, objections, and decision stage during the conversation.
- Routes high-intent buyers to your calendar via Calendly or Chili Piper.
- Pushes full conversation context into Salesforce, HubSpot, Slack, and Gmail. Custom integrations via webhook.
- Lets marketers query every conversation in plain English. Omer’s examples from the demo: “Show me everyone who asked about a competitor”; “Show me conversations from Fortune 500 companies in North America that mentioned my competition, had more than five questions, and didn’t convert”; “We released a new feature three months ago. Is anyone asking about it?”
- Surfaces missing information. When the AI can’t answer a question well, it flags the gap so you know what content to write next.
- Includes an email agent that continues the conversation in the buyer’s inbox using the same knowledge layer.
What it doesn’t do, per Salespeak’s own page: it doesn’t run outbound, replace your SDR team, do CRM lead scoring or routing logic, or replace your content team. It’s an inbound qualification tool.
More details: https://salespeak.ai/website-inbound-agent/
Product 2: Agent Conversations
The CDN-level layer for AI agents. What it does:
- Agent Analytics: a free CDN-level tool that captures every AI agent visit to your site in real time, broken down by bot (GPTBot, ClaudeBot, PerplexityBot, Gemini), per page, per minute. No JavaScript, no performance overhead.
- Agent Optimizer: detects AI agents at the edge and serves grounded answers from your structured content instead of your marketing copy. Per Omer: “We basically tell ChatGPT, take the page that you want, but here’s a set of FAQs that we’ve generated automatically for you that will allow you to provide a better answer for your customers.”
- Captures every question an agent asks (answered or not) as a buyer intent signal.
- Runs on Cloudflare, Vercel, Netlify, CloudFront, and Nginx.
More details: https://salespeak.ai/agent-interaction-platform/
How much does Salespeak cost?
Salespeak’s pricing is transparent and easy to access from their home page. The model is month-to-month with no annual lock-in.
There are three buying paths: Visitor Conversations only, Agent Conversations only, or a Visitor + Agent bundle. Each path has four tiers that scale with conversation volume. Free tiers exist on both Visitor Conversations (25 conversations per month) and Agent Conversations (analytics only).
We’re oversimplifying, but the pattern across paid tiers in the bundle looks like this. The entry tier covers basic chat plus CRM sync and agent analytics. The mid tier adds an actionable insights module and a full security and privacy compliance audit. The growth tier scales conversation volume and AI query limits. The enterprise tier moves to custom annual contracts and a custom support tier.
Pricing ranges from a free Agent Analytics dashboard to $3,000 per month for the Growth bundle before you hit Enterprise (which is custom).
A few useful billing details: conversations from internal IPs don’t count toward your plan, spam and malicious conversations aren’t billed, and bot/crawler traffic is excluded from billing.
For current pricing and overage rates, check the pricing page directly.
Who on the marketing team should use Salespeak
To get the most out of the product, it needs accurate, current information to train its agents. If your website hasn’t been updated lately or you have outdated messaging on the site, plan for a content cleanup phase before launch. You can use offline materials to train the model in the meantime, but the website itself will still be what AI agents read.
A realistic rollout looks like four phases.
Phase 1 (before purchase): collect and clean the source material. Pull together everything you’d want the AI to know: current product docs, sales collateral, ICP, battle cards. Dedicate a resource to correcting the misinformation that’s already on the website. This belongs to the content marketing team in close collaboration with whoever owns messaging. Budget anywhere from a few hours to a few weeks, depending on the state of the content.
Phase 2: testing. Your product marketing lead and demand generation lead should set aside four to eight hours to run buyer simulations and pressure-test how the AI handles different inquiries. This is the time to find the answers you don’t want it giving customers.
Phase 3 (concurrent with phase 2): integration. Marketing operations should connect the tool to the CRM and decide how the intent signals captured in conversations get routed. Salespeak passes context; your team owns the scoring and routing on the receiving end.
Phase 4 (ongoing): weekly review. Budget at least an hour or two per week to review questions buyers asked, mine the platform for content gaps, and escalate technical or content issues to the website team, especially anything specific to LLM bot traffic.
Omer’s framing of the kind of operational discipline this requires is worth pointing out.
A great marketing team is being measured by the number of experiments that they can run.
“A great marketing team is being measured by the number of experiments that they can run.”
Omer Gotlieb Founder & CEO, Salespeak
The tool is built to support that. The team still has to do it.
What Stood Out About the Demo
etup is fast and the testing layer is real; the simulator catches problems before they reach customers. The month-to-month plan with no annual lock-in is rare in this category and useful for teams that want to pilot before committing.
The platform is a well-orchestrated way to stay on top of content sprawl and outdated information that tends to proliferate over time, while also improving the buyer experience on the website. And the LLM optimization at the CDN level is one of the more concrete answers to GEO we’ve seen demoed; most platforms in this space are still describing the problem, this one operates on it.
Before buying, schedule a demo, which you can do by clicking here.
Thanks for registering!