15 AI Case Studies + Latest Trends: Live Webinar with Richard C. Wilson

1. Use AI for the impossible, not just the tedious. The most powerful leverage of AI is for tasks that were physically impossible a year ago—like analyzing 18,000 sent emails to capture your writing voice, running sentiment analysis across 10,000 Slack messages, or categorizing 152 newsletters in 15 minutes—rather than just speeding up work a human could already do.
2. Build Claude skills and projects. Claude 'skills' are framework files (.md profiles) that encode best practices and activate dynamically when relevant; loaded into a Claude 'project,' a skill stays always-on. Both are available on free Claude accounts since February, making them accessible to nearly everyone.
3. Avoid the 'blob of best practices.' Generic AI draws on everything from Wikipedia to a 17-year-old podcaster, giving you down-the-fairway answers. To get top 1% results you must feed it proprietary data, your mental models, your preferences, and instructions to play devil's advocate rather than just agreeing with you.
4. Migrate software and cut costs with AI. Wilson's team used AI to switch membership platforms in nine days (versus nine months and $25,000 with a human consultant for a prior Kajabi migration) and cut their software stack from $40,000 to $17,000 per year.
5. Cross-check important questions across multiple AI platforms. For critical decisions, run the same question on Claude, Gemini, Grok Heavy, and ChatGPT, then paste all results into your most-trusted platform for synthesis—leveraging trillions of data points and reducing single-model bias.
6. Use Cowork to actually do things, not just chat. Anthropic's Cowork (a downloaded tool separate from Claude.ai) can manipulate Excel spreadsheets, sort and draft emails, generate documents, and run weekly lead-generation processes—taking real actions rather than just providing feedback.
7. Always have AI audit its own work. AI output is rarely perfect on the first pass; instruct it to audit, debug, and re-audit in multiple directions until everything is cleaned up, and always keep a human in the loop to check the results.
8. Physical presence is the new moat. As everything—data rooms, videos, even entire companies—can be faked, software moats collapse and the real differentiators become physical assets and in-person trust, which is why Wilson built a 'verified member' tier with background checks.
9. Apply extreme focus and 'overdo it.' Echoing the founder of Scale AI, Wilson stresses choosing highly unique projects, applying more focus than everyone else, and overdoing quality and thoroughness as the path to outsized results—and uses an AI accountability tool to grade himself daily against this standard.

Webinar Framing and the Family Office Club

Richard C. Wilson opened by distinguishing this webinar from past AI sessions: rather than rushing through as many tools as possible, the focus would be entirely on practical, day-to-day uses of AI by himself and his team. The explicit goal was to send participants away with many more concrete use cases for AI as founders and investors, helping their teams be more thorough, save time, and get more done per hour. He repeatedly stressed that nothing covered would be speculative or sensational ('not five years from now, just imagine the magical things'), but exactly what they are doing right now.

Wilson introduced himself and his organization: he founded the investor club 19 years ago (later in the episode he cited 18 years), has hosted over 300 live events at a pace of about 30 per year, runs a monthly webinar series, and operates the most-visited website in the family office industry (familyoffices.com) plus billionaires.com on billionaire strategies and mental models. His team has produced several thousand YouTube videos, including 250 new videos in the current year alone.

He defined a 'family office' as an ultra-wealthy family—which could be a decamillionaire, centimillionaire, or billionaire—and described the Family Office Club as a hybrid of an investor club and a mastermind format, built on the premise that spending time around people worth $10 million-plus and $100 million-plus helps members think like them and build a powerful network. He noted the club's reach: 16 million social media community members plus another 2.5 million followers (about 18.5 million total), the largest network of any investor club, and over $1 billion in deals done within the club.

Events and Membership Tiers

The club hosts events in four rotating cities: New York City, Dallas, Fort Lauderdale, and Los Angeles, while Wilson himself is based in Oahu, Hawaii. Event formats include one-day masterminds (10am–3pm) and larger one-to-three-day investor summits; he encourages new members to attend both over time. Speed networking events were scheduled for the week of April 21st across all four cities, and a Single Family Office Summit was planned for May 28–29 in New York City featuring 75 speakers over two days—billed as having more family offices on stage than any other 2026 New York event.

Three membership tiers were outlined. Charter members get six months of access to live events (virtual and in-person) plus core AI tools for just under $2,000 for six months. VIP members pay $5,000 for a full year and also receive 22,000 investor profiles, Excel spreadsheets, and the investor pipeline networking tool. Verified members—the most vetted tier at $10,000/year—undergo a background check (2–4 business days) for a blue checkmark, get the most platform exposure, two panel talks on stage, one exhibit table, a custom-produced one-minute founder pitch video, three guest tickets, and sponsorship exposure that normally sells for over $16,000. Wilson offered verified members a bonus of customized AI tools (like the Dewey due diligence tool trained on their own criteria).

The Family Office Club AI Tool Suite

Wilson explained that his team has built 40 to 65 AI tools (depending on how they are counted), with about a dozen used internally and 40-plus available to members. Crucially, he built all of these himself—no contractors, no consultants, no white-labeled tools—codifying nearly two decades of experience custom-structuring over 100 deals and watching over $1 billion in transactions close. The tools also draw on a proprietary dataset of 3 million words from 1,500 investor talks delivered on stage at events, data not available anywhere else.

He demonstrated where the tools live in the new member portal: a 'Start Here' section with a welcome video and letter, an event photo gallery (featuring a Shark Tank shark, a UFC fighter, NFL/NBA athletes, Grant Cardone, and Tony Robbins interviews), and the Artificial Intelligence Hub. Guest-pass users see unlocked tools (no lock symbol) versus locked premium tools. The investor portal contains 11 AI tools targeted at investors; the member portal unlocks the full set for paid members. The video library houses over 1,500 videos sortable by billionaires, mini-series, capital-raising, due diligence, deal structuring, fireside chats, and multi-hour workshops.

Featured Tools

Several flagship tools were highlighted. 'Dewey' has completed over 2,500 due diligence reports, analyzing a deal 20 different ways and custom-scripting seven questions about any opportunity to help users be thorough, save time, and avoid 30-minute phone calls on deals that may be dead on arrival.

'Deal Structure Ninja' produces 10 deal structures for any deal—covering collateral, preferred equity, gross revenue royalty, buyback provisions, and optionality—and iterates: a user can say 'I like number four and six, give me 10 more like that,' repeatedly. 'Billionaire Collective Intelligence' captures wisdom from a thousand public billionaire talks plus 45 exclusive interviews, allowing users to converse with the collective wisdom of billionaires. Wilson also referenced a 'Mental Model Intelligence' tool loaded with 2,500 mental models and an 'Extreme Focus Finder' tool.

Case Study: AI-Driven Software Migration and Cost Savings

Wilson recounted a stark before-and-after. Roughly four years ago, migrating from a WordPress membership site to Kajabi—with a so-called Kajabi 'expert' who went off the rails and billed unauthorized overtime—cost $25,000 and took nine months. By contrast, using AI to migrate off Kajabi and Whova into their new membership portal took just nine days for transferring and importing members and building out the platform; the only remaining piece was uploading 1,500 videos (about 800 in at the time, with all 1,500 expected by May 1st).

The payoff was both speed and cost: the new portal, with its own branded app coming to the app store, will cost $17,000 per year instead of $40,000. Wilson emphasized that AI eliminated the bottleneck of emailing customer service for every question—instead he got instant answers based on public forums and help documentation. He framed software migration as a prime AI use case and predicted big software companies will be massively disrupted in real time, prompting his recurring question: 'Could we build our own AI tool to do that?'

Similarly, moving from custom GPTs to their own in-portal AI agents required scripts totaling over 250,000 words of instructions and blocks—equivalent to two full-length novels—a task no human could realistically QA and customize in a reasonable timeframe. Claude, which he noted is among the top platforms for coders on technical projects, made this feasible.

Claude Skills and Projects Deep Dive

Wilson devoted significant time to Claude, noting that 80%+ of attendees were not primarily using it and were missing capabilities hard to replicate in ChatGPT. He explained that Claude offers a free account that, since February, supports 'skills.' Skills are frameworks (.md profile files) encoding best practices and ways of thinking through problems from someone with deep niche experience. His team converted the most powerful of their 50 AI tools into 10 downloadable skill profiles available in the AI Hub under the Claude logo (also downloadable as a zip).

Installation takes about four clicks: in Claude, go to 'Customize' on the left, then 'Skills,' then the plus symbol, then upload the downloaded file. Once loaded at the base account level, Claude reads each skill's description and 'leans into' the relevant skill dynamically whenever a day-to-day request matches—following, for example, your due diligence process, a choke-point strategy, the investor advantage skill (built from 1,500 investor transcripts), or a deal-structuring skill.

Wilson then explained the difference with 'projects' (which ChatGPT also has). Projects act as sandboxes that retain context across all past chats on a topic—he keeps separate projects for email blasts, the membership portal, and his 'Investor Superintelligence' AI platform. Within a project, an added skill is always on, ensuring the framework is never forgotten. He demonstrated adding the Billionaire Collective Intelligence skill to his billionaires.com project so that every work session uses a framework built from a thousand billionaire talks and 45 interviews.

Clarifying Skills vs. Private LLMs

Responding to a question, Wilson clarified that a skill is not a private LLM. A private LLM is typically a locally downloaded model (a mini ChatGPT/Grok/Claude) kept on your own computer so data isn't shared with the internet. A skill instead equips your existing LLM account with frameworks and best-practice intelligence. He said his team is not currently focused on private AI given the downsides of available local models and setup, arguing that using any big-tech platform involves a trust trade-off—Microsoft, Google, and others can scrape data from Word, Drive, Gmail, etc.—that each person must resolve individually, especially since most people already use these platforms.

Free Skill Bonuses

As a webinar bonus, Wilson dropped two free skills into the chat for all attendees (not just paid members): a 'First Principles Thinking' skill that strips problems down to base premises, removes noise and assumptions, and then figures out how to scale and automate; and a 'Choke Point' skill. He offered to email these skills to anyone listening later who emails richard@familyoffices.com.

Cowork: AI That Takes Action

Wilson explained that AI for the masses is moving beyond chatting and feedback toward actually doing things—going out to the web, looking up prices, sending emails, recalculating spreadsheets, and changing formulas. Anthropic's 'Cowork' is a separate piece of software downloaded and run on your computer (distinct from Claude.ai the LLM) that can manipulate things rather than just talk. He likened it to OpenClaw (formerly Clodbot then Moldbot) and parallel tools like Perplexity Computer and Manus—all converging in a race to build secure, highly functional autonomous AI that operates in a sandbox without access to sensitive items like bank accounts. He explained the boom in Mac minis stems partly from people wanting always-on machines to run OpenClaw/Cowork autonomously, and partly because Mac processors handle AI requests well.

A practical limitation he flagged: Cowork's memory history is separate from Claude.ai and they don't re-sync, so he must upload his dashboards, goals, and voice.md file into both environments and update both whenever something changes. He also noted Cowork can be controlled from a phone via 'dispatch on Claude,' but the computer must be running for it to work—another reason people buy Mac minis (so processes keep running even when a laptop is shut down on an airplane).

Excel Plugin Demonstration

Wilson live-demonstrated the Claude Excel plugin (an orange symbol at the top of Excel after installation). With a short prompt on a 600-row spreadsheet of sent emails, he asked it to combine columns A and B, add columns C and D, color any resulting cell over 4,000 red, and professionally format the whole sheet. Cowork read the data, asked permission before manipulating ('allow once' or 'allow always'—he uses 'allow once'), and completed the task in about two minutes. He stressed the value scales: imagine 5,000 rows or a complex discounted-cash-flow or investment-banking formula. He said the plugin worked roughly 12 of 13 times, with the lone failure due to an unclear prompt, underscoring the need to be explicit about what you want and don't want.

Daily Email Triage

Every morning Wilson opens Cowork to analyze all incoming emails, sorting them into seven categories: personal emails, contracts to sign/legal matters, revenue opportunities, VIP/whale clients, and others. It then drafts an email to himself flagging the top five emails in each category, and finally surfaces the five most time-sensitive, important emails to answer first. He gave a concrete example: the founder of Keller Williams emailed an introduction to a billionaire, which the report correctly identified. He framed this as applying the principle that 'you'll never have enough time to do everything, but you always have enough time to do the most important things,' freeing energy for in-person meetings, health, family, and building more AI systems. Crucially, he does NOT let AI send emails on his behalf—only draft and sort them—citing cases where an AI agent sent a rude or nonsensical generic reply to a billionaire from the Pritzker family, missing nuanced context that matters greatly in his business.

Billionaire Interview Lead Generation

Wilson runs a weekly Cowork process scanning social media and the web for new or interviewable billionaires—especially those who recently wrote a book, may be less private, or are young and thus potentially more open to being interviewed for billionaires.com (target: 100 billionaire interviews). The output categorizes leads as hot, warm, strategic targets, and young billionaires based on his criteria. He noted the same approach applies to any sales team, deal origination, or deal-flow generation.

Newsletter Categorization

Wilson demonstrated using Cowork on his LinkedIn newsletter 'Deca Millionaire Accelerator' (about 68,000 subscribers, 152 editions). He asked it to surf through, categorize, and describe all editions, applying keywords and suggesting better category names. It produced a navigable resource with categories including AI, video content, interviews (billionaires, centimillionaires, Robert Cialdini, Tony Robbins, Mark Cuban, Tim Draper, Jeff Hoffman), mental models, growth-oriented tools, capital-raising strategies, investors and investment strategies, events and community, and performance/ultra-healthy living—plus an intro, conclusion, and table of contents—in about 15 minutes (a half-day of human work).

Capturing Your Voice: The 18,000-Email Project

One of Wilson's most striking case studies illustrated his 'overdo it' philosophy. He started simply by asking AI to mimic his voice based on past sent emails. He escalated to 1,000 emails, then realized that wasn't enough because he sends about 10 distinct types of emails and wanted many examples of each across different prospect types over a long period. Ultimately he had AI process 18,000 sent emails—excluding reminders to himself and trivial messages, focusing on substantive back-and-forth and important VIP emails—to build a comprehensive 'voice.md' file.

The resulting file documents his categories, best practices, who he forwards emails to and why, how he phrases things, his preferences on doing deals, what he never wants to do, words he never uses, and how he treats potential time-wasters and phone calls. He instructed it to weight 20% of the sample toward the last six months and to understand his current routing (forwarding certain emails to team members Ellie, Jennifer, Xander, or Daphne). The voice.md skill is now permanently installed in his email-writing Claude project and at his base account level, and also trains his daily Cowork inbox-analysis drafts—producing replies in his voice that route correctly. From the same data he generated a one-page inbox-management cheat sheet plus a 15-page training guide for his executive assistant.

Other Practical AI Use Cases

Wilson rattled off numerous additional applications. Connecting AI to Slack enables sentiment analysis across thousands of messages to identify which of 15 team members aren't getting enough attention, which stated projects show no evidence of completion, where balls were dropped, and who is showing constant frustration and may be unhappy—picking up trends a human would miss. It also helps instantly locate a single paragraph buried in hundreds of Slack messages during a live client call when Slack's own search fails.

He uses AI to score leads and contacts on credibility, net worth/wealth estimates, and industry heavy-hitter status, and has internal tools that assess whether someone is an investor, capital raiser, or potential event speaker. He noted that feeding a name, city, company, or LinkedIn link into AI yields more in 30–45 seconds than five minutes of manual research, and the AI can tailor a suggested outreach message and recommend which team member should handle the contact.

For payroll, a Cowork process runs spreadsheet calculations on payroll day, with his assistant double-checking the math and flagging items—aiming toward AI handling 90% reliably while never fully trusting it. He uses the AI PowerPoint capability (mostly via Cowork rather than a plugin) to brainstorm and build slides—about three to five of the webinar's case study ideas came from Cowork analyzing their projects and Slack. He also iterates ideas using AI as a thinking partner: for a new 'connect with founders and fund managers' directory (a modernized version of their Blue Pages, structured like a cleaner Craigslist with drill-down categories such as manufacturing founders, a member with 33 car washes, a member with a billion-dollar credit fund, hedge funds), he thought about it himself 4–5 times, bounced it off AI twice, then off team members—building higher conviction.

WhisperFlow for Voice Transcription

Wilson strongly recommended WhisperFlow (whisperflow.ai) for real-time dictation. He reprogrammed its default two-key-hold shortcut to use the middle mouse scroll button, so he can record without holding keys, hit it again to stop, and have transcribed text appear instantly. He types fast (75–80 words per minute) but speaks at 120–160 words per minute, and WhisperFlow auto-corrects spelling and grammar on the fly. He tied this to a broader point: AI has intensified the workload on founders (faster information, more to improve, more new systems to learn), so people must consciously slow down on important things and use tools like WhisperFlow to make fewer mistakes. He demonstrated it live during the webinar's chat responses. For longer audio he mentioned Rev (rev.com), Dropbox auto-transcription, and youtubetranscript.com for YouTube videos.

Daily Dashboard Alignment / Accountability Tool

Wilson built a private 'Daily Dashboard Alignment Grade' tool in Cowork. He uploads his monthly/quarterly/annual goals document with 30–40 mental models he wants to live by; the tool pulls activity from Google Calendar, Slack, and Gmail, then scores how his energy was actually invested versus his stated priorities. The report includes a one-line verdict, an overall score (he typically gets 40–75), pillar scores, five things done right, five things not done right, something he should almost certainly stop doing, a blind spot he might not see, and how he could be faster and apply his mental models more consistently. He runs it nearly every day, sometimes twice. He argued that consistent, objective daily accountability against how you intend to live wasn't possible before AI, and described his laminated one-page document of goals and mental models that he reads each morning while shaving—models he runs on automatically aren't listed, only ones he wants to reinforce or absorb.

Cross-Platform Verification and Devil's Advocate Prompting

For high-stakes questions, Wilson queries multiple platforms (Cowork, Gemini, Grok Heavy, ChatGPT) because each cites different data points, then copies all results into his most-trusted platform for synthesis—leveraging trillions of data points. He always instructs AI to base responses on data, identify who dominates an industry and has the most expertise, back conclusions with data, give percentage-confidence and numeric scores, and explicitly NOT just agree with him. He cited a quote: 'the dumbest person you know right now is being told by the LLM, yes, you're 100% right'—warning that AI is programmed to be likable, can make people delusional, and must be instructed to play devil's advocate.

AI Industry Trends and Macro Observations

Wilson shared several trend data points. Enterprise AI applications with AI agents are projected to jump from under 5% in 2025 to 40% by the end of 2026—an 8x (800%) shift in a year. He cited that the U.S. Treasury prevented and recovered $4 billion in fraud in 2024, up from $652 million the year before—six times more, largely due to AI. He referenced a news report that morning about a 'mythos' model so good at finding security threats in technology (90x better than the prior model) that it can't be publicly released without causing chaos, and would only go to entities like the Defense Department.

He sees the economy recovering in a bifurcated way: tech-enabled, AI-enabled versions of old business models grow much faster while outdated businesses grow much slower. He noted multiple people have argued AI's invention is more important than the discovery of fire or electricity. On disruption, he cited Oracle laying off about 27,000 people via a 6am email, plus layoffs at companies like Fox and Jack Dorsey's company, attributing tech-sector layoffs and record profits to deep, early AI adoption—predicting more displacement as normal companies catch up, alongside eventual net job creation from AI and robotics during a turbulent transition.

He referenced Elon Musk's XAI announcement and the ambition to have AI replicate software like Microsoft Word from scratch (via Grok, Google Studio, Claude, or Dorsey's new platform), and a participant coding seven apps at once in Claude, as evidence that software moats are collapsing. His conclusion: in a world where data rooms, Zoom videos, YouTube content, and even entire AI 'influencers' (one fabricated AI persona has gained hundreds of thousands of followers) can be faked, the real moat becomes physical assets and in-person trust—people increasingly want to walk a manufacturing floor or meet face-to-face to verify authenticity. This directly motivated the verified-membership background-check tier.

Capital Raising in the Current Climate

Wilson noted family offices manage somewhere between $3 and $14 trillion—more than all private equity and venture capital combined. He cautioned members to have realistic expectations: relationships with family offices rarely close at a first event. He told of a well-known brand founder who came to one event expecting to raise $10 million, got upset when he didn't, and was reminded that if such outcomes were guaranteed, the club would charge $250,000 per event rather than $2,000–$10,000.

He balanced this with success stories: $30 million of seed capital secured from a dinner organized around an event; a member growing from $30 million to $130 million AUM after a publicly-traded-company billionaire backed them; an $800 million family making five allocations into club members; and a Shark Tank shark (Kevin Harrington, whose family office Wilson helped formalize) investing in three transactions and partnering with members. His point: a $4 million check equals forty $100,000 checks (40x efficiency) but doesn't take 40x as long—though larger checks require longer relationship-building, trust, in-person meetings, planting seeds, uniqueness, dialed-in positioning, a one-liner, a one-pager, a one-minute founder video, a DDQ, value-add first, and the ability to do direct deals, JV, and co-GP structures rather than just pitching a fund with a large minimum.

On the macro environment, Wilson said 2.5x as much capital was raised in 2023 as in 2025, so fundraising is literally about two and a half times harder—not imagination. He blamed high interest rates more than war or inflation, expressing surprise rates hadn't fallen further, and predicted AI-driven economic slowdown and uncertainty will eventually force rates down. On politics, he argued both left and right are being fed extreme edge cases that make each see the other as having lost their minds, fueling uncertainty and fear—but that interest rates remain the bigger drag on capital raising.

Q&A Highlights and Recommended Tech Stack

Asked whether Claude beats ChatGPT, Wilson said they are simply different; he adopted Claude for its superior, consistent technical abilities, while praising Google for catching up well and possibly pulling ahead. On token limits, he uses Claude Max ($200/month) and rarely hits walls despite heavy use and migrating 35 AI tools, though a participant (James) noted coding seven apps in parallel maxes quickly; another (Angie) suggested using ChatGPT for lighter work and Claude for heavy lifting to save credits.

On confidentiality, Wilson explained most of his portal tools run on Sonnet 4.6 and occasionally Gemini 3.6, and that he has no visibility into what users submit or the results—the data is treated as if uploaded directly to Claude or Gemini. He recommended TerraMind (teramind.co) as an AI-enabled alternative to Hubstaff for managing virtual teams, noting it gives more transparency than sitting next to someone in an office and that requiring physical office presence limits the talent pool. He also praised Grok for being relatively objective, data-driven, and built with devil's-advocate thinking.

His recommended tech stack for a family office or founder, costing about $600/month, comprised: Claude Max ($200/month) with Cowork plus the Excel and PowerPoint plugins; Grok Premium; and Google AI Ultra—stating this combination enables everything demonstrated. He also mentioned subscribing to ChatGPT Pro. Critically, he advised configuring each platform's personal-preferences field (Claude allows up to 1,500 words/characters): feed extensive context, then ask the AI to compress it, ensuring it includes instructions to not always agree, to base answers on data with confidence percentages and numeric scores, to leverage your AI tools, mental models, and goals, and to know your company, priorities, and VIP clients.

Member-Suggested Use Cases

• Tristan: scanning over 10,000 business websites by user experience, with the AI getting stronger each time and then building improved websites for businesses.
• A participant: unsubscribing from spam, scanning websites on PageSpeed.
• James: running different agent groups for each email type with voice and video agents answering 24/7 (Wilson agreed it's possible but declined for his nuanced business).
• Salimah: enriching detailed Notion investor lists by having Cowork research each investor's public statements, then customizing value-first outreach.
• Jay (philanthropy): finding wealthy hospital patients/affiliates impacted by a cause and screening lists by wealth indicators using data sets like Apollo or ZoomInfo.
• CRM: Wilson uses ActiveCampaign and admits using AI for subject-line optimization, email writing, and trend analysis, but not yet for deep CRM cleanup; AI tools integrate more readily with HubSpot and Salesforce.

The 'Overdo It' Philosophy and First Principles

Wilson played a 30-second clip from the founder of Scale AI (described as the youngest billionaire at age 21, a physics/math/chess genius who sold his AI company for $3 billion). The clip's message: if you are hyper-focused and invest enormous time and effort—'go the extra mile, go the extra 10 miles, constantly overdoing things'—you improve faster than anybody else by many times, attributing accomplishment to focus and overdoing it. Wilson watches this clip every single morning as a daily reminder.

He tied this to his core thesis: choose highly unique projects (be one out of 500 or 1,000), apply extreme focus, and overdo it on quality and thoroughness—the same mindset that drives success in investing, capital raising, and platform building. He encouraged everyone to spend an hour brainstorming 'a list of 20 things AI could do that no human could ever do,' suggesting one or two ideas might transform their business. He repeatedly framed AI's greatest leverage as enabling the previously impossible, and stressed that the experts who can 'call AI's bluff'—noticing when output is generic or wrong—will hold the most power in the marketplace.