HighLevel Is Giving AI Away Free. Neil Patel Calls It a Bold Bet (Full Episode)

Neil describes attending three events in Dallas, London, and Milan, where HighLevel CEO Sean told him the company was giving its AI away for free. Sean’s reasoning was that businesses sometimes have to “eat the cost” to do what is best for customers. Eric and Neil agree that funded or public companies rarely make genuinely expensive bets, especially when unlimited token usage could send costs soaring. Neil recalls that free UberSuggest usage once cost him roughly $320,000–$350,000 per month, illustrating how dangerous an uncapped offer can become.

Eric says he argued with a new managing director and his CTO that their agentic workflows, bots, and AI visibility reports should be free because anything built today may be outdated within three to six months. Free tools create usage and mindshare; a smaller group of users will then pay for implementation or services. His AI visibility report, which previously might have cost $1–$10 to produce, now costs about $0.18. At $10,000 monthly volume, that is $1,800, a reasonable acquisition cost if it generates qualified sales opportunities.

Neil warns that free AI becomes risky when a product has a large user base and no clear cost ceiling. Eric counters that falling costs and the need to retain mindshare justify the bet: he bought two Nvidia DGX Sparks for $4,000 each, and each had reportedly risen to about $8,000. He expects compute and electricity demand to increase, but believes companies must push usage before competitors achieve feature parity. They compare this with Meta’s enormous data-center commitments and Anthropic’s need for additional infrastructure, concluding that HighLevel’s decision is bold even if less frightening than a data-center gamble.

Eric revisits Meta’s CFO Susan Li repeatedly warning that the company could be in serious trouble if its AI investment failed. He estimates that only 0.2% of the world’s population—about 20 million people—currently works with AI at the intensity discussed by the hosts. Of 3.55 billion employed people, roughly 6% run companies, and only 4% of those employ 20 or more people, implying that perhaps 17,000 firms are genuinely AI-native. They expect much more usage as adoption expands and warn that top employees may leave traditional firms to work with other AI-native people.

Eric says a fully configured new MacBook Pro with the touch display would cost about $14,000, while a fully loaded Mac Studio with 512GB of RAM would cost about $21,000. Neil notes that the laptop starts around $2,899 and refuses to buy every upgrade before testing the built-in studio-grade microphone. Eric argues that expensive hardware sells because technophiles want the latest capabilities, especially when old machines make screen sharing and webinars slow. They also discuss a mutual friend warming to the Duo laptop despite cracks and other reported problems because it is faster and helps him accomplish more.

The hosts discuss a viral video from French riots showing people kicking, chairing, and striking a storefront window that would not break. The Chinese glass company used the footage as social-media marketing, reportedly receiving at least 22,000 views on one post and potentially hundreds of thousands overall. Neil says he bought security glass for his Beverly Hills home after earlier protests and crime concerns, paying roughly $200,000–$300,000 extra for glass resistant to kicking, ice picks, and ordinary gunfire. He says the product’s performance sold him more effectively than conventional advertising.

Eric says a friend who runs security for a publicly traded software company is receiving many more inquiries during “Black October,” amid hacks and AI-related security incidents. The friend’s practical recommendations are password managers with unique passwords, preferably passkeys, and strong two-factor authentication for important systems. Eric is seeking his advice on securing Single Brain Gateway, which connects agents to services and credits such as Apollo and Clay. He also says companies facing a breach should be transparent about what happened, who was affected, and what it means after completing their legal obligations; hiding incidents creates greater anger when the truth emerges.

Eric contrasts a YPO global marketing summit featuring Neil, which sold out only at the last minute, with his newer global AI summit, which sold out two weeks before the event. He attributes the difference to three factors: AI is a strong current trend, the summit is attached to an event already bringing about 1,000 people, and the timing was fast. He says the same rule applies to YouTube: waiting even two or three days after a trend begins can sharply reduce performance. Neil adds that being too early can also fail; AI agents once attracted little attention, but now resonate because products such as Muse, Dots, and Grokbot have made the category familiar.

Neil describes testing AI agents for local-service searches after Mike needed a Las Vegas water softener. A manual quote was about $7,600, while an agent found and scheduled appointments with reputable providers under $5,000; Mike chose one costing about $4,800. Neil’s early findings are that agents value not only total review counts but review recency, consistent messaging across platforms, and third-party mentions. Comparison articles perform best when they compare a business with the region’s genuinely recognized competitors, not obscure alternatives. Eric connects this to how a company brain would weight recent data and the velocity of people discussing a business.

Eric demonstrates a free AI visibility report that says how often Single Grain appears in AI answers, then unlocks the full report after an email address is submitted. He calls it a lead magnet and says the underlying infrastructure can produce it for cents; his JEV account showed 786 million tokens, 468,000 requests, and $31 spent. Neil says their tests found that highly detailed, chart-heavy reports often generate lower close rates. Simpler offers that identify a glaring problem and show the top two or three fixes convert better. Eric agrees that teams should iterate quickly, but not change 50 variables at once without enough data to know what improved conversions.

Eric cites Spawn, a reportedly 10-person company spending around $32 million annually—or 30 times payroll—on Claude tokens. Its permanent agents own areas such as engineering or finance, have permissions and memory, and helped ship 77,000 substantial changes in 30 days, about 108 per hour. The company reportedly eliminated pull requests and relies on agents coordinating through the living product and codebase. Eric urges teams to ask what “unreasonable” work their agents could perform, such as testing 20 landing pages or creating thousands of ads. He cautions that rapid iteration still needs controlled experiments and enough traffic to separate useful changes from harmful ones.

Eric describes “Money Prompt,” which analyzes code and the entire funnel to identify revenue leaks. It found missing activation emails, absent reminders for users running low on credits, and a page where 20% of users abandoned the process. Engineers could not infer those commercial issues from code alone, so Eric argues that judgment and years of pattern recognition remain valuable even when AI can build and ship quickly. The hosts then discuss Jen Abel’s sales advice: demos should focus on business problems and outcomes rather than features, preferably connecting results to money made or saved. They also favor a compelling, no-brainer offer over a long technical pitch.

Neil says founders improve early sales not necessarily because they are better presenters, but because they can hear objections, change the product, and unblock decisions faster. Eric cites Larry Page accelerating Google’s AI work after intervening directly. They agree that a win rate above 50% can signal underpricing, misleading expectations, or poor qualification. Neil recalls a salesperson selling SEO to a sombrero business for $250 per month, a price that could not produce a return; he refunded the money, personally helped the owner generate thousands in monthly revenue, and declined to keep her as a client. Their sales data shows discovery calls are about 50/50 speaking and listening, while proposal calls are closer to 60/40.

The hosts debate a FlexJet CEO’s claim that acquisitions often signal weakness, overpay for market share, and damage culture. Eric notes that Meta’s acquisition of Scale cost roughly $14.3 billion but helped push its market value up by more than $200 billion. Neil describes buying a Southeast Asian company for its client logos, reputation, and access to RFPs, despite knowing its culture was weak; employees prioritized ping-pong over customers, so the team had to be replaced or retrained. He says acquisitions can create “one plus one equals three” by cross-selling services and reducing churn, though a $400,000-revenue acquisition would not be worth the internal fees and distraction for his agency.

Eric proposes replacing annual performance reviews with daily AI-assisted check-ins: what high-leverage work was completed, what is blocked, and whether the work is genuinely valuable. Neil agrees that strong employees want timely feedback. They then discuss an investment thesis favoring companies that bear responsibility for consequential outcomes over companies performing easily verifiable work that can be commoditized. Neil says performance-based agency work carries the most risk because clients may resist paying after unexpectedly strong results, especially when enterprise customers prefer predictable budgets; a management fee plus capped bonuses is safer. They close by discussing Granola’s AI-search growth, Google’s 12% post-earnings stock drop, and Eric’s view that advertising remains viable while people and agents transact online.