23 January 2024

Tools & Strategies You MUST Use To Survive The 2024 AI Revolution

2Frameworks

11Insights

Frameworks in this episode

Communication6 steps

CRAFT Prompting Framework

Structure AI requests with context, role, action, format, and target

Read →

Innovation6 steps

Final Artifact Learning Loop

Learn AI by shipping one real artifact and studying the friction

Read →

Insights & moments

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

Myth Buster· 1

Myth Buster35:30

AI Personalization Is Not Valuable Unless the Recipient Benefits

The episode challenges the assumption that technically personalized communication automatically feels personal. Synthetic outreach can still work if it delivers genuine value or connection, while poorly tuned messages remain ineffective regardless of how advanced the AI is.

Authenticity problems predate generative AI
Recipients judge communication by relevance and value
Perfect production can trigger an uncanny response
Human imperfections may be deliberately restored to generated media

“AI is not the issue”
Nicholas Holland · 38:00

“great to me means that it basically gave value to me as the recipient”
Nicholas Holland · 38:00

#personalization#authenticity#ai outreach#communication

Hot Take· 3

Hot Take11:30

The Open Internet May Give Way to Gated Expertise

Holland argues that freely published expertise increasingly trains AI systems that can answer users directly. Marketers may therefore split content between public material that attracts attention and private material delivered through controlled audience relationships.

AI answers can absorb value without returning referral traffic
Public content may still reduce discovery friction
Unique perspectives may move behind gates
Direct audience channels could become more strategically valuable

“what I think will happen is that the open internet is over with”
Nicholas Holland · 11:30

“you'll learn what parts to make public because you don't want to create too much friction when people come to your site”
Nicholas Holland · 12:30

#content strategy#gated content#inbound marketing#ai training

Hot Take15:30

The Personal AI Assistant May Live on Your Phone

The conversation predicts that Apple or Android could run consumer models grounded in private device data. Photos, style preferences, health signals, and other personal inputs could turn a generic model into a Jarvis-like assistant.

Mobile platforms already hold rich personal context
Local models could improve privacy and personalization
Personal assistants may combine photos, wearable data, and instructions
Fine-tuning could shift AI from generic answers to individual guidance

“Apple or Android is going to run a local consumer model that is just going to be amazing”
Kip Bodnar · 15:30

“we are getting a Jarvis right”
Kieran Flanagan · 17:30

#edge ai#mobile ai#personal assistant#consumer technology

Hot Take49:30

The AI May Soon Assign the Marketer's Next Task

The conversation predicts a reversal in which humans stop directing every AI action and instead receive prioritized instructions from models. Marketers and salespeople could be told which weakness to address or which contacts to call next, with reasons grounded in available data.

Prediction quality may improve sharply with new model generations
AI systems could prioritize work rather than merely execute requests
Humans may shift from task assignment to judgment and approval
Model improvement can invalidate products built around temporary weaknesses

“at some point I think we flip the AI as like we're the intern the AI is telling us what to do”
Kieran Flanagan · 50:30

“you should call these next 10 people next because of this reason”
Kieran Flanagan · 50:30

#future of work#task prioritization#ai agents#marketing automation

Explainer· 4

Explainer16:30

Why Proprietary Data Is the Real Fine-Tuning Advantage

The hosts distinguish meaningful customization from merely uploading information already available online. Fine-tuning becomes strategically useful when a model can learn from proprietary customer, behavioral, stylistic, or personal data that the base model does not possess.

Public training material may produce little differentiation
Customer data can make models useful for company-specific work
Personal data could enable consumer assistants tailored to individual needs
Current customization often averages toward generic internet content

“so you need to have Pro priority data”
Kieran Flanagan · 17:00

“that's going to be a real unlock I think for marketers and customers”
Kieran Flanagan · 17:30

#fine-tuning#proprietary data#customer data#personalization

Explainer23:00

Cheap AI Audio Makes Micro-Audience Podcasts Viable

Tools combining language models, synthetic voices, editing interfaces, and licensed music can dramatically reduce podcast production costs. The strongest opportunity may not be mass-market shows but useful audio created for small customer or internal audiences.

Multimodal systems combine scripting, voice, music, and editing
Lower production costs make narrow formats economically viable
Internal updates can become audio instead of defaulting to text
Cheap production accelerates both good and bad content

“it will unlock a lot of micro audiences is my belief”
Nicholas Holland · 28:00

“the cost of creating audio video is just going to be just as cheap as text”
Kip Bodnar · 28:30

#podcasting#multimodal ai#synthetic voice#micro audiences

Explainer30:00

From Generated B-Roll to Licensed Digital Presenters

The speakers map a progression from today's generated B-roll to synthetic video that reproduces a person's writing style, voice, lips, appearance, and mannerisms. Narrow, approved use cases such as event promotions and product announcements could arrive before fully autonomous digital personalities.

AI video is already effective for evocative B-roll
Voice, text style, lip motion, and appearance are becoming separable layers
Approved likenesses could generate repetitive promotional videos
Audience trust will depend on who authorizes the digital representation

“Runway and a series of others are basically really good at b-roll right now”
Nicholas Holland · 31:30

“I will be able to use my likeness and then my audience will just have to trust that I was the filter of when it…”
Nicholas Holland · 34:00

#ai video#digital likeness#b-roll#synthetic media

Explainer51:30

Why Full-Spectrum Social Agents Could Outperform Individuals

Holland decomposes social media work into strategy, publishing, monitoring, response, and analytics, each with substantial depth. An AI agent may outperform an individual not by being the best at every component, but by maintaining competence and context across the entire operating spectrum.

Social media requires coordinated strategy, publishing, monitoring, and analytics
Each function contains many channel-specific decisions
Agents can retain more cross-functional context than one specialist
The human role may become supervising an agent with broader competence

“agents are the next wave”
Nicholas Holland · 54:00

“the combination of that is better than any exactly the amount the amount of uh knowledge it can retain and get through”
Kieran Flanagan · 54:00

#social media#ai agents#marketing operations#automation

Story· 1

Story04:30

How Conversational Search Replaced a Multi-Step Google Journey

Holland uses a winter-burn problem to show why conversational search feels fundamentally different from traditional search. He refined his question through dialogue, bought supplies, and forwarded the complete exchange to his landscaper without visiting a source link.

Conversational search supports rapid follow-up questions
A single dialogue can move from research to purchase and execution
Publishers may influence commerce without receiving traffic or attribution

“I didn't click on any Source links”
Nicholas Holland · 06:30

“I basically had to talk back and forth with it for a while to get what I was wanting”
Nicholas Holland · 06:30

#ai search#commerce#search behavior#inbound

Tool· 1

Tool42:00

Using LLM Pattern Matching to Prioritize CRM Work

The hosts describe LLMs as general pattern-matching engines that can interpret CRM records with less task-specific training than older predictive systems required. A model can inspect a CSV, propose which account to contact first, explain its reasoning, and draft an appropriate message.

Older predictive systems often required large specialized datasets
LLMs inherit broad patterns from language training
CRM context can support prioritization and message generation
Missing business constraints can still produce damaging recommendations

“we took HubSpot CRM database and we basically posted over just a CSV file into chat GPT”
Nicholas Holland · 47:00

“who do you think I should call First”
Nicholas Holland · 47:00

#crm#pattern matching#sales ai#lead prioritization

Takeaway· 1

Takeaway40:00

Synthetic Translation Can Make Global Communication Immediate

AI-generated video becomes especially valuable when it translates a speaker into languages the audience understands while preserving their appearance and delivery. This capability could help companies communicate internationally much faster than conventional localization workflows allow.

Lip-synced translation improves comprehension without reshooting
Audiences may accept synthetic delivery when it removes a language barrier
Leaders can address multiple markets from one recording
Translation tools could accelerate international expansion

“I think it's going to help companies go International Global much much faster”
Kieran Flanagan · 41:30

#translation#localization#global marketing#synthetic video