Joel Horwitz says AI agents let marketing teams research accounts, detect buying signals, and create personalized outreach at a scale that manual work cannot match.
2
AI agents do not remove salespeople or marketers because human approval, authenticity, trust, and conversation still matter.
3
Marketing work is becoming systems integration, with teams connecting data, agents, campaign tools, landing pages, and communication channels into tailored workflows.
Summary
Joel Horwitz describes marketing as an early but fast-moving use case for AI agents. His team uses Clay and related workflows to detect signals such as free trials, job postings, public filings, and account activity. Agents research prospects, identify likely buying situations, prepare account context, and generate tailored outreach such as websites, LinkedIn ads, memes, and founder voice messages. Horwitz argues that this work improves the quality of interactions rather than simply increasing the volume of cold email. He also expects AI to change search and lead qualification, including traffic from ChatGPT and Gemini and a possible move from MQLs toward richer, signal-based qualification. He is clear about the limits. People still need to review outputs, correct mistakes, build relationships, and create authentic thought leadership. His current role is closer to a marketing systems integrator because no single tool can yet generate and run an entire campaign from one prompt. AI lowers the cost of experimentation and gives teams more time to understand customers.
AI marketing is moving from bulk content toward richer interactions
Joel Horwitz says early AI marketing focused on producing large volumes of cold email and outreach, often accepting a low response rate. That approach created spam and became the public image of AI marketing. He sees a different pattern emerging, where agents help create better interactions across channels. The aim is to meet a target audience with a timely message that fits its situation. Horwitz believes AI can now create customer experiences that would be impractical to produce manually. He frames this as a shift from quantity toward value creation, with agents supporting more specific and relevant communication instead of simply generating more content.
Agents can turn account signals into personalized outreach
Horwitz describes workflows built with Clay for clients. An intent signal might be someone starting a free trial or an account posting a relevant job opportunity. An agent then researches the person and company, writes an outreach sequence, and creates material for different channels. Examples include a meme using the prospect's image, a personalized message in a founder's recorded voice, and account-specific websites or LinkedIn ads. The agent handles the research and production work that would otherwise take an SDR or marketer substantial time. Horwitz says the purpose is to help a team act quickly without reducing every interaction to a generic template.
Human salespeople still provide judgment and personal style
Demetrios Brinkmann asks whether these workflows make SDRs unnecessary. Horwitz rejects that conclusion. Each SDR has a personal style, and the generated work is not simply copy and paste. Agents can research an account, find its likely pain points, and prepare creative material, while a person decides how to use it and continues the conversation. Horwitz also focuses on the account rather than a single contact because larger software purchases involve several decision-makers. Research may include company filings, Crunchbase, PitchBook, competitors, and the people involved in a buying decision. Agents reduce the time needed for this work, but they do not replace the relationship.
Personalized pages and ads can meet prospects before a sales call
Horwitz says a prospect may not be ready for a meeting even when the account shows interest. The immediate task is to educate the buyer about how a product fits its existing environment. His team has used Mutiny to create personalized websites that change the messaging and customer examples for a target account. They also create LinkedIn ads tied to account research and specific pain points. Horwitz recalls seeing target accounts react positively to ads because the message was unusually relevant. The workflow uses research to decide what a prospect should see before asking for a conversation.
AI is changing search, advertising, and lead qualification
The conversation moves to advertising and search. Demetrios Brinkmann describes using language models to update stale Google AdWords keywords and find SEO opportunities through tools such as Ahrefs and Semrush. Horwitz adds that agents can classify converting and non-converting keywords and manage negative keywords. He is also seeing more referral traffic from ChatGPT and expects teams to think about optimization for answer engines such as ChatGPT and Gemini. He argues that traditional top-of-funnel metrics often produce vanity traffic, bots, or low-value signups. Agents can research seemingly weak email leads, identify competitors or other non-buyers, and route them differently.
Signal-based qualification could replace rigid MQL scoring
Horwitz questions qualification models based on actions such as clicking four links, visiting a website, attending a webinar, or downloading a white paper. Those actions do not prove that a person is ready to buy. He proposes the term AIQL for a qualification model that combines more meaningful signals and reduces some human subjectivity. Inputs can include job postings, public filings, organizational information, budget clues, existing competitors, and relationships among decision-makers. Agents can also investigate personal email addresses, although Horwitz says the identification process is incomplete and has an industry average hit rate of about 40 to 45 percent. The model still depends on the quality and amount of signal provided.
A shared inbox and Slack bot connect research to sales action
Horwitz explains that his team sends outreach into a unified inbox that the company can view. When a prospect responds, someone can join the conversation immediately. For higher-priority signals, an agent sends a message in Slack to the account executive or assigned representative. The message includes the new job posting, the hiring manager's LinkedIn contact, and relevant facts about the account or person. The representative can then write a more informed note. Demetrios suggests that this could become a marketing equivalent of Cursor, where a marketer asks for a campaign and receives the connected assets. Horwitz agrees that this product does not yet exist in a complete form.
Horwitz says today's marketer often has to connect separate tools and bespoke workflows. Clay, Webflow, and Mutiny cover parts of the process, but no single interface can yet turn a campaign request into ad copy, creative, audience definitions, landing pages, white papers, and supporting content. That gap is why his agency acts as a marketing systems integrator. Demetrios connects this to a broader change in software work, where people want the finished result while AI handles more of the intermediate tasks. Horwitz agrees that AI lowers the threshold for experimentation. A prompt can produce a working prototype quickly, giving engineers and marketers something concrete to refine.
AI can support thought leadership without creating authenticity
Demetrios distinguishes targeted campaign work from brand building, podcasts, talks, and long-term trust. Horwitz agrees that AI agents are assistive rather than replacements for people. In an aquaculture startup project, his team used agents to find relevant operations, arrange conversations, process interview transcripts, and identify themes for product and go-to-market decisions. The agents helped synthesize many conversations and suggest subjects for thought leadership. They did not remove the need to speak with customers or develop original ideas. Horwitz says people respond to people, and founders still need to build trust and show up personally. AI gives them a sparring partner and helps distill information, but it does not create a relationship on its own.
"The key to using AI agents in your workflow and in marketing and frankly in business is assistive versus replacement."Joel Horwitz39:55
Who should watch
You are building account-based marketing workflows and need a practical view of where agents can handle research, personalization, and campaign preparation.
Your marketing team relies on MQLs, broad top-of-funnel traffic, or manual account research and you want to reconsider which signals deserve attention.
You are evaluating AI-assisted marketing tools but need to preserve human review, founder authenticity, and direct customer conversations.
