Nowoslawski. AI-built lists need 8+ search filters
The rule came out of auditing why an autonomous campaign agent kept producing messaging that read fine but landed thin.
Eric Nowoslawski says the rule that made his autonomous campaign agent work was about the list, not the copy. Any campaign the agent launched needed at least 8 search filters, or as many filters as it could hold while still retaining 3,000 contacts in the list.
The context is a build he started earlier this year, an AI agent that could look at a website and generate a cold email campaign with no human in the loop that got leads. He says it was far harder to build than he expected.
What the audits found
When he and his team audited why campaigns were not working, the messaging was generally correct but missing punch when they read it contact by contact. The lists the AI was producing were the ones most people make by default. Employee headcount, job title, location, industry, some company keywords. Broad enough that generated copy had nothing specific to attach to.
The fix was to hand the agent the full set of filters available in a Clay search and force it to tighten. Nowoslawski calls that the silver bullet that cracked the effort, and says the resulting copy spoke to prospect needs and problems rather than hovering above them.
The filter list
The filters he published include industry, company size band, exact employee count, social follower count, year founded, company type, annual revenue range, and employee growth over 3, 6, 12 and 24 months. Also B2B versus B2C model, Clay's AI industry classification, AI sub-industry, revenue stream such as SaaS or transaction fees, products and services matched by meaning rather than keywords, NAICS codes, parent company versus subsidiary, keywords in the company description, and specific domains to include or exclude.
Location can be set by country, state, city, ZIP, region, HQ or any office. Tech stack can be filtered by vendor, product and category. Job postings can be filtered by title, description keywords, seniority, employment type, location, date posted, whether the role is still open, and recruiter. There is also a people filter for how many current employees match a title, seniority or location, his example being 3 or more people in sales.
For people sending
The claimed lift is his own and it comes with a condition. He frames the 2x as what happens if you have never run a cold email campaign, not as a result measured across accounts, and the filter list is specific to searching inside Clay. The transferable part is the threshold itself. If your targeting collapses to headcount, title, location and industry, the copy has nothing left to be specific about, no matter who or what writes it.
any campaign we launched needed to have at least 8 search filters (or the most you could have while retaining 3,000 contacts in the list) in order for an agent to be successful
Know what is landing before you burn another domain
One email a day. Under three minutes. What the operators found out yesterday.
