The AI-Era Businesses Built on Trust, Access, and Community

Sam Parr’s starting point is that people who are not AI specialists can still build businesses that grow in an AI-shaped economy without selling AI. The examples he and Shaan Puri discuss are built around access, trust and participation: comparing business performance, sharing referrals, reaching a vetted pool of buyers, or solving a narrow personal problem.

NADA 20 Groups offer one model. Parr says the groups began in the 1970s. Dealers apply and pay roughly $6,000 to $8,000 a year to join a group of about 24 noncompeting dealerships. They meet in person three times a year. Each dealer brings a profit-and-loss statement, and the group goes through it line by line. Peers can ask how a dealership achieved a strong result, or suggest changes where its numbers lag.

The NADA material shown on screen describes grouping members by franchise, dealership department, sales volume, location, travel preferences and budget. It also says members get financial comparisons between their dealerships and average and best-in-class dealers. Parr says the model has expanded beyond dealerships to include owners of HVAC companies, marinas and boat-repair shops. Based on membership dues and pricing information, he estimates the business may bring in $13 million to $15 million a year.

That is more than a paid discussion group. Members pay for a structured comparison with peers who understand their business, while the group’s composition helps avoid direct competition. Parr’s proposed opportunity is to adapt the format for other less competitive fields—he mentions dental practices and law firms—and improve its presentation or marketing. Puri notes that NADA itself dates to 1917; that is the association’s history, not the start date Parr gives for its 20 Groups. The distinction matters: the longevity of the association does not establish how long this particular program has existed.

The White Coat Investor is a related example in medicine. Puri describes it as a community and media business for doctors, focused on their financial lives rather than medical research or clinical practice. A doctor might ask peers about buying a rental property or evaluating its cap rate. Parr notes that the business started as a blog. Puri says there are other businesses serving doctors and suggests there may be room for a newer version that reaches doctors and dentists through current media channels.

BNI applies the community idea across professions rather than within one. Local chapters bring together noncompeting business owners—a lawyer, realtor, garage-door company, landscaper or accountant—who can refer work to one another. Parr recalls seeing a meeting of about 30 people at a YMCA. Members took turns naming leads they could pass to the group. The BNI website shown in the discussion listed more than 355,000 global members, 11,800-plus chapters and $26.9 billion in member-generated business over the previous 12 months.

Parr says membership costs low thousands of dollars a year and that BNI had recently been acquired. He relays a rumor that the deal was valued above $1 billion; neither the website figures nor that rumor are presented as independently confirmed transaction terms.

The examples exchange different things. NADA members pay for a recurring forum to compare business performance. White Coat Investor gives doctors a place to discuss financial decisions. BNI members pay to participate in a network designed to produce referrals. In each, the group’s composition is part of the product: members either share enough context to help one another or are matched because they do not compete.

Parr connects these businesses to his interest in owning and operating multiple communities. He says his company may eventually acquire other communities, with the ambition of building a business with billions in revenue and hundreds of millions in profit. Puri asks whether that means building a peer group for every vertical. Parr’s answer is broader: he sees potential in combining multiple communities, rather than simply reproducing one format everywhere.

Sam Parr’s next example, MODA Communities, began with a watch enthusiast in Sioux Falls who wanted a better place to buy and sell watches. In 2018, the founder started a Facebook group. Parr says MODA has since grown to more than a dozen full-time employees and over 600,000 members across roughly 20 to 25 communities for watches, collectible cars, gold and other goods. The company’s website, shown during the discussion, describes 23 Facebook communities and more than 600,000 members.

The watch groups are divided by price and interest, including watches under $10,000, watches above $10,000 and a group for especially high-value collectors. Dealers cannot simply join and post. Parr says a dealer is interviewed, then must be vetted by another community member. Once accepted, dealers pay roughly $1,000 to $10,000 a year to list items in front of the group. Parr estimates annual revenue at $5 million to $10 million.

The distinction between membership and dealer access is important to the business model as Parr describes it. The audience is gathered in communities, but dealers pay to reach that audience and list their inventory. The vetting process is part of what makes that access valuable: buyers encounter sellers who have passed a screening process, while sellers get in front of people already interested in the category. Parr does not describe a transaction fee or claim that MODA itself operates a full marketplace checkout; the source of revenue he identifies is dealer payment for access and posting.

Puri identifies the buyer pool as the asset. Parr’s broader argument is that communities, networks and marketplaces can remain valuable even as AI makes technology itself a less reliable competitive advantage. MODA did not need to begin by building a standalone marketplace: a Facebook group gave the business a way to gather buyers and sellers around an existing interest.

Puri also points to the risk. A business built on Facebook groups depends on a platform whose rules, reach and ability to add members can change. The same platform provides access to a large network while limiting the community’s control over how that network works. MODA’s example combines specialist access and trust-building with that platform dependence.

Shaan Puri describes a pattern he calls the whale economy: in businesses ranging from collecting to mobile games, a small group of high-spending customers can account for a large share of revenue. He recalls Twitch’s gifting-subscription feature as a simple way to give those customers another outlet for spending. A viewer could buy 200 subscriptions, at roughly $5 or $6 each, for other people in a streamer’s chat. The streamer receives subscription revenue, recipients get subscriptions, and the buyer gets recognition from the community. Puri compares the gesture to buying a bottle at a club: the feature gave a high spender a visible way to spend more.

Puri says he saw a similar concentration in mobile games. He cites Candy Crush as having 82 million monthly active users, with 4% of players generating all the revenue. He describes the other 96% as part of the cost of finding those high spenders. From conversations with people in the games industry, he says, studios study how much they spend to acquire players, what different players generate, and which forms of targeting reach people likely to spend heavily across multiple games.

He describes game economies—currencies, rewards and the balance between spending, frustration and churn—as deliberately designed systems for keeping valuable players engaged. In one example, he says a lapsed payer might be given a run of success and then encounter a frustrating obstacle that makes paying again more tempting. Puri calls this sad, while also describing it as a revealing business mechanism: designers look for new ways to give high spenders reasons to spend.

The tension is not incidental to his account. Puri describes some mobile-game design as akin to a slot machine, but notes that mobile games do not have the payout constraints he associates with casino machines. A company can set its own balance between encouraging spending, frustrating players and losing them. He also mentions gambling products and customers whose combination of disposable income and addiction can make them especially lucrative. That observation points to a cost of the model, not simply a clever monetization opportunity: the same mechanics that increase spending can exploit people who are susceptible to them.

Parr says MODA’s founder described a similar revenue pattern: virtually everything the company does is free, while its top 1%, or even top 0.001%, pay to keep everything running. That does not mean a community should copy a game’s spending mechanics. It does show how free participation can help assemble a network while a small number of customers pay for unusually valuable access or experiences. In MODA’s case, Parr says, dealers pay to post in front of a vetted buyer pool; in Puri’s Twitch example, viewers pay for subscriptions that are given to others. Different products, but both make a small group’s spending consequential to the larger community.

Shaan Puri applies that spending logic to collectibles. In a traditional shop, a collector buys a pack of Pokémon cards and hopes to find something rare. Apps such as Triumph, he says, move the experience onto a phone: users open a digital pack, see what they received, and can keep the item in a digital collection or have it shipped. The Triumph screens shown during the discussion describe a collection that can remain vaulted until the user wants it shipped or sold.

The central mechanism in Puri’s account is buyback. If someone pays $50 for a pack and gets a card they do not want, the app may offer $43 to buy it back. The user can treat the amount as credit toward another pack, making another attempt feel as if it costs only $7 more. The card can go back into circulation rather than being shipped to the person who opened the pack.

That changes the economics of the interaction. Puri describes the company as receiving $50, returning $43 in value and retaining the difference, while keeping the card available to sell again. When Parr asks whether a reported figure should be understood as revenue or marketplace volume, Puri says it is a mix involving the cost and fulfillment of cards. He says Triumph had done $1 billion in revenue in the preceding year, but his qualification makes clear that the figure is not a simple measure of net revenue after card costs.

Puri also mentions Collector Crypt, describing a similar buyback or “vending machine” approach. He first says it had grown from about $50,000 a month to $25 million a month, then later cites data suggesting it was doing $100 million a month. He does not reconcile those figures. The Collector Crypt screens shown in the discussion describe cards being sent to vaults, authenticated and linked to a profile, then bought, sold or transferred.

The underlying attraction, Puri argues, combines a chance-based reveal with an item that has a resale value. He compares the apps with Gacha games, in which users repeatedly open mystery packages hoping for rare rewards. A collectible card differs from a purely cosmetic game item because it can have a real-world price. Puri says the top Pikachu card can sell for $15 million.

He describes the format as a gambling-like mechanic attached to physical collectibles: a user opens a pack, hopes for a valuable item and can try again. In his account, the chance of obtaining something valuable matters partly because the item can be sold. The buyback adds another step to that loop, making a disappointing pull feel less like a full loss and turning the returned value into a prompt to open another pack.

Puri extends the same logic to toys. A friend in the toy business, he says, showed him sales data in which the same products performed much better after being put into mystery boxes rather than sold individually. He reports that the mystery-box version generated seven times the revenue. In a conversation with a major intellectual-property holder, Puri says, the other side argued that customers wanted choice. He says the data across several categories showed the mystery format outperforming.

The explanation he offers is not that buyers dislike choice in general, but that the mystery format gives them a reason to keep opening packages in pursuit of a rare item. That creates a commercial advantage and an ethical tension. Puri calls the format addictive, compares it to gambling, and describes its appeal as a way to spend more on repeated attempts. The source presents the mechanics as a business opportunity, but also recognizes that the repeat-purchase loop can encourage people to continue after an unsatisfying result.

Sam Parr’s final business example is Private Photo Vault, an app for storing photos separately from a phone’s main photo roll. Its function is narrow: users move images into a locked vault and can remove them from the ordinary photo library. The conversation jokes about the app’s use for intimate photos, but Parr emphasizes that it is storage, not a photo-sending service.

Parr says the app has about a million reviews and that he met its founder at a Hampton event. The founder shared revenue and staffing figures that Parr would not disclose, but Parr described the company as a “behemoth” and said it had operated at that scale for a long time. The app’s promotional material shown on screen emphasizes that it requires no account, is offline by default and that the team cannot see stored photos or videos.

The example is useful partly because the product is so specific. It is not a broad social network or a new technology platform; it provides a separate, locked place for photos. Parr does not disclose enough to establish the company’s revenue or headcount, so the scale remains his characterization rather than a figure readers can inspect. But the app fits his opening premise: a non-AI business can still be substantial if it serves a durable, clearly defined use.

In a separate stretch of the discussion, Shaan Puri and Parr turn to crypto. Puri says many people who had focused on crypto have moved into AI, while a smaller group of high-conviction people remain. He also mentions concern that AI hacking or quantum computing could threaten encryption, and says a successful attack on Bitcoin’s encryption would cause panic.

The hack they discuss is not aligned neatly with the news item shown on screen. Puri and Parr describe a Binance hack as having happened “last week” and refer to losses of roughly $5 billion. The Associated Press screenshot displayed during the discussion is dated October 30, 2024, and its headline says Binance reported a $6.3 billion US hack. The speakers’ description and the dated screenshot are different pieces of information; the source does not explain the discrepancy.

Puri recounts that after an earlier hack, Binance created a fund called SAFU and began putting 10% of trading fees into it. He says it had grown to about $2 billion. Parr recalls hearing that Binance would cover the loss. Separately, an on-screen Binance update dated October 31, 2024 says the company had set aside more than $6.34 billion in company funds to cover hacked losses and that 90% of the impacted amount had been replenished. The document says the replenishment used a partial deployment of SAFU, a portion of the Futures insurance fund and Binance’s corporate treasury. The speakers’ estimate of the fund and the displayed statement about total company funds are not the same figure.

When Parr asks who took the money, Puri says the last account he had read attributed the theft to North Korea and described a woman in a romantic relationship with an engineer obtaining his laptop. He presents this as what he had read, not as firsthand knowledge. Parr asks how stolen crypto could be turned into spendable money if transactions are trackable.

Puri answers that funds can be split across wallets or moved through services, and describes a DeFi maneuver: the attacker put the stolen assets into a protocol, borrowed a stablecoin against them and withdrew the borrowed funds. In his account, that leaves the attacker with the stablecoin even if the original collateral is frozen. He describes DeFi as a permissionless form of banking in which users deposit collateral and borrow against a portion of it; if the collateral falls in value, it can be sold automatically to cover the loan.

Puri compares DeFi’s prospects with prediction markets. He says earlier attempts to build prediction markets attracted money and smart people but did not take off, before later examples such as Polymarket and Kalshi. He suggests DeFi may follow a similar path. That comparison is his view, alongside his account of how DeFi can be used to borrow against assets; it does not resolve the security questions raised in the exchange.