02 November 2023

I Used AI To Build A Billion Dollar Business (#170)

1Frameworks

10Insights

Frameworks in this episode

Strategy7 steps

Contextualize, Focus, and Iterate

Turn curated source material into a tailored strategy through rapid AI iteration.

Read →

Insights & moments

The myth-busts, hot takes, explainers, and tools worth keeping.

Myth Buster· 1

Myth Buster16:00

AI Becomes Valuable Beyond Generic Best Practices

The hosts argue that impressive first-generation answers often amount to an average of internet best practices. Greater value emerges when AI pattern-matches across a distinctive private or curated dataset and connects those patterns to a precisely defined problem.

Generic AI answers can still be useful but are rarely differentiated.
AI is strong at pattern identification across large collections.
Curated data helps expose shared reasons behind successful practices.
Problem-specific questions turn pattern matching into actionable insight.

“AI is awesome at pattern matching and pattern identification and finding commonalities”
Kit Bodner · 16:00

“it is just giv you the average that best practices of the internet”
Kieran Flanagan · 16:30

#pattern matching#data#best practices#ai

Hot Take· 2

Hot Take17:00

The Future of AI Is One-to-One Personalization

Kieran uses Grammarly's access to a user's writing as an example of personalized AI. Rather than generating broadly acceptable prose, future tools can learn an individual's tone, voice, and habits and tailor every interaction to that person.

Personal writing history can train a tool on individual style.
Suggestions become more useful when they reflect the user's voice.
Unique personal datasets can differentiate otherwise similar AI tools.
One-to-one generation may replace much one-to-many content production.

“AI in the future is going to be one to one”
Kieran Flanagan · 17:00

“this is to me the future of every single AI interaction where it's trained on a unique data set for you”
Kieran Flanagan · 17:30

#personalization#writing#grammarly#future of ai

Hot Take18:00

AI Is a Great Co-Writer, Not a Great Writer

The hosts find Claude useful for progressively refining business principles but reject its early output as generic and conference-like. Their conclusion is that AI works best as an iterative partner whose drafts are shaped through human judgment rather than accepted as finished writing.

Early generated principles may sound polished but generic.
Examples such as Nike's principles can guide form and tone.
Multiple revisions can create more meaningful language.
Human judgment remains necessary to select and sharpen ideas.

“it is a great co-writing partner it is not a great writer”
Kieran Flanagan · 19:00

“a few more revs and you would have some pretty meaningful and very articulate like principles”
Kit Bodner · 19:00

#writing#collaboration#claude#editing

Explainer· 2

Explainer08:00

The Open-Source LLM Opportunity Claude Identified

Claude synthesizes the State of AI report and Buffett's principles into a proposed company focused on open, secure, and responsible language models. The hosts find a more concrete opportunity underneath the broad mission: managed infrastructure that lets ordinary companies adopt open-source models without maintaining the technical stack themselves.

Advanced AI access remains concentrated among large technology companies.
Open-source models can be cheaper and more specific to company needs.
Most companies do not want to maintain model infrastructure themselves.
A managed service could make open models easier to integrate and operate.

“that's basically making it easy for companies to adopt open source models in the same way they adopt SAS”
Kieran Flanagan · 09:30

“that business will exist”
Kieran Flanagan · 09:30

#open source#llms#startups#ai infrastructure

Explainer24:00

The Customer Pains Behind Managed Open-Source AI

When the hosts push Claude deeper, it identifies operational problems that make managed open-source AI commercially plausible. These include vendor lock-in, API throttling, closed architectures, and limited visibility into the safety techniques of proprietary black-box models.

Vendor lock-in constrains long-term flexibility.
API subscriptions can throttle builders' ability to scale.
Closed architectures limit control and customization.
Black-box models provide limited insight into safety techniques.

“it found like vendor lockin as a real problem”
Kit Bodner · 24:00

“blackbox proprietary models”
Kit Bodner · 24:30

#vendor lock-in#apis#open source#customer pain

Story· 1

Story21:30

Dyson Needed 5,000 Prototypes—AI Compresses the Feedback Loop

Kit recounts that James Dyson spent 14 years and more than 5,000 prototypes developing the production vacuum cleaner. The hosts contrast that physical manufacturing cycle with software and service work, where AI can dramatically accelerate drafting, feedback, and course correction.

Dyson began prototyping at 31 and reached production at 45.
Physical product iteration historically required years of experimentation.
AI can compress iteration cycles for software and service businesses.
Faster feedback loops help teams solve problems and improve output.

“he was 31 when he made his first prototype and he was 45 when the first Dyson production vacuum came off the line”
Kit Bodner · 22:00

“AI just like so extrapolates that so fast”
Kieran Flanagan · 22:30

#james dyson#iteration#prototyping#feedback loops

Tool· 2

Tool10:30

Build a Differentiated Sales Pitch From Competitor PDFs

Kieran proposes uploading competitors' sales materials and combining them with the company's principles. The AI can then help define the ideal customer, isolate current problems, and explain why the company's solution is meaningfully different.

Collect competitor materials into a single evidence set.
Define the target customer before drafting the pitch.
Make the AI explain differentiation explicitly.
Iterate the result into a practical sales narrative.

“get a entire like PDF spread of all of your competitors and other companies within the market and upload it”
Kieran Flanagan · 10:30

“ask how can I pitch my company in a differentiated way”
Kieran Flanagan · 10:30

#sales#positioning#competitive intelligence#ai

Tool14:30

Turn Newsletter Archives Into a Go-to-Market Database

Curated newsletters often contain years of company growth, marketing, and sales examples in consistent feeds. Those archives can become a queryable strategy database that helps users find common patterns and tactics appropriate to their company's size.

Newsletter archives contain structured collections of real strategies.
RSS or another standard feed can make the archive easier to ingest.
AI can identify common patterns across thousands of examples.
Queries can filter tactics by company size and situation.

“you can actually have a database of thousands of different like Marketing sales and growth strategies that you can just use AI to query”
Kieran Flanagan · 15:30

“it's basically like having an expert Council of marketing leaders in your pocket”
Kit Bodner · 15:30

#go-to-market#newsletters#rss#marketing

Takeaway· 2

Takeaway04:30

Why Nike's Offensive Mindset Echoes Warren Buffett

The hosts connect Nike's principle of staying on offense with Warren Buffett's advice to become greedy when others are fearful. Both ideas encourage selective aggression when competitors retreat, while acknowledging that defensive behavior may preserve market share but rarely expands it.

Economic pressure makes defensive behavior understandable.
Periods of widespread caution can create openings for aggressive moves.
Playing defense may preserve market share without gaining it.
Operating principles from different thinkers can encode similar strategic logic.

“we're on offense we're on offense all the time”
Kieran Flanagan · 05:00

“if you are playing defense you will get defensive”
Kieran Flanagan · 06:00

#strategy#warren buffett#nike#market share

Takeaway27:00

Tailor Every Investor Pitch to the Investor's Own Thesis

The hosts propose loading an AI model with a venture firm's mission, published writing, and investment criteria. A founder can then tailor the pitch to the investor's worldview instead of sending the same generalized deck to every firm.

Research the investor's public mission and decision criteria.
Use published material to model how the investor evaluates companies.
Connect the business to that specific investment worldview.
Generate one-to-one pitches rather than one-to-many content.

“I basically load up a model with all of that VC's Mission things that they've wrote about why they invest in certain companies”
Kieran Flanagan · 27:00

“we are going to be able to create one toone content”
Kieran Flanagan · 27:30

#fundraising#venture capital#pitch decks#personalization