17 January 2023

6 Marketing Opportunities That A.I. Unlocks For Businesses

2Frameworks

9Insights

Frameworks in this episode

Marketing6 steps

Goal-Driven AI Iteration Loop

Give AI a measurable goal and iterate on the variables that drive it

Read →

Innovation6 steps

Impossible-to-Possible AI Opportunity Scan

Find valuable AI opportunities by testing previously impossible outcomes

Read →

Insights & moments

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

Myth Buster· 1

Myth Buster11:30

AI Does Not Automatically Deliver Nuanced Answers

Kieran challenges the idea that AI inherently restores nuance to online information. Current systems can present definitive answers without making uncertainty visible, creating a risk that users will trust incorrect outputs after seeing several correct ones.

AI may state uncertain information as a definitive answer.
Users cannot always distinguish correct outputs from errors.
Repeated correct answers can create misplaced trust.
Confidence indicators could help communicate uncertainty.

“I actually think the problematic thing about AI right now is it does provide definitive answers, and at times you actually can't tell whether they're…”
Kieran Flanagan · 12:00

“And you're going to be so confident in this when you're getting the right answers, you're not going to question the wrong answers.”
Kieran Flanagan · 12:30

#ai#accuracy#hallucinations#critical thinking

Hot Take· 2

Hot Take20:30

AI Could Turn All Media Into a Personalized On-Demand Product

Kieran predicts that audiences will eventually specify the subject, format, and intellectual style of the media they want. Instead of choosing among fixed articles, courses, or films, each person could generate a one-to-one version tailored to the immediate request.

AI may generate media around an individual's exact request.
Users could specify both subject matter and presentation style.
Education could be generated as a personalized course on demand.
The model shifts media from one-to-many publishing toward one-to-one creation.

“The Uber version of this is actually very similar to what Uber did for taxis is like media on demand.”
Kieran Flanagan · 20:30

“And the thing that I can't get my head around is like the way that we win today as marketers is we create one to…”
Kieran Flanagan · 21:30

#media#personalization#content#education

Hot Take22:00

Instant Style Copying Could Commoditize Brand Differentiation

The ability to imitate a distinctive creator's style creates a profound branding challenge. If audiences can generate new ideas through another person's recognizable voice or perspective, media differentiation may weaken and copyright disputes may intensify.

Distinctive writing styles may become reproducible without the original creator.
Instant imitation threatens perspective-based differentiation.
Media, content, and art are especially exposed.
Copyright rules will face pressure from style-trained generation.

“Like how are you different? How do you have a brand? How do you have a perspective, a point of view when anybody can copy…”
Kipp Bodnar · 22:00

“This is like why the copyright wars are going to be so intense.”
Kieran Flanagan · 22:30

#branding#copyright#differentiation#ai

Explainer· 2

Explainer02:00

Why AI Capabilities Are Advancing So Quickly

The hosts use the rapid expansion in model parameters to illustrate the extraordinary pace of AI development. They argue that businesses should expect assumptions about what AI can do to become outdated quickly.

GPT-2 was described as having 1.5 billion trainable parameters.
GPT-3 was described as having 175 billion trainable parameters.
The hosts emphasize that the increase occurred over roughly two years.
Rapid capability growth makes static AI strategies risky.

“GPT 3 has 175 billion.”
Kieran Flanagan · 02:30

“And that's what I keep trying to tell people is I don't think you understand the rate of evolution here.”
Kieran Flanagan · 02:30

#ai#language models#innovation#technology

Explainer19:00

Why AI Content Still Requires Heavy Human Iteration

The hosts describe current creative AI as a human-driven, AI-assisted production loop. AI can accelerate drafts and widen access to creative work, but quality still depends on selection, revision, and specialist tools.

AI provides useful starting points for content and design.
Many generated outputs still need substantial editing.
Creative systems increase both production speed and creator access.
A competition-winning AI artwork reportedly involved 900 iterations and further Adobe work.

“And so there's still like, it's still quite human driven AI assisted and then human edited, right?”
Kieran Flanagan · 19:30

“So he went through 900 versions and he iterated, iterated, iterated, and then he actually ended up doing the rest of it in Adobe.”
Kieran Flanagan · 20:00

#content#generative ai#creativity#workflow

Tool· 2

Tool30:00

Build an Always-Available Synthetic Customer With Private Data

A company with sufficient customer data could train an AI to approximate how its target segment thinks and responds. Teams could then consult that model during product and market decisions while using the aggregate perspective to reduce the influence of vocal outliers.

Proprietary customer data can support a segment-specific model.
Teams could query the model during frequent decisions.
Aggregated patterns may balance feedback from unusual customers.
The model should supplement rather than eliminate direct customer research.

“you can train an AI to act and behave in the same way as your customer would.”
Kieran Flanagan · 30:00

“Like it's like your customer in your pocket and you can ask it all the time.”
Kieran Flanagan · 30:30

#customer research#proprietary data#decision making#ai

Tool31:30

Turn Internal Documentation Into an Employee Onboarding Assistant

The hosts propose an AI assistant trained on a company's internal wiki and onboarding documentation. Employees could ask questions conversationally instead of manually searching scattered pages, making institutional knowledge easier to access in remote organizations.

The assistant would ingest approved internal documentation.
Employees could retrieve answers through a chat interface.
The concept is particularly useful for remote onboarding.
Its usefulness depends on complete and organized source material.

“So AI can completely solve employee onboarding, especially in remote world.”
Kieran Flanagan · 31:30

“Do you not think that would be incredible for employees to be able to access that information through a chat?”
Kieran Flanagan · 32:00

#onboarding#knowledge management#remote work#ai

Takeaway· 2

Takeaway26:30

AI Could Counter the Internet's Recency Bias

Kipp argues that AI can make historical ideas immediately accessible and relevant rather than favoring only recent material. Businesses and educators could rapidly connect older lessons with modern contexts, although the quality of that repackaging remains important.

Humans and search systems tend to favor recent information.
AI can aggregate lessons from a much wider historical range.
Older ideas can be reframed for modern audiences and problems.
Historical accessibility may create new teaching and marketing opportunities.

“I think AI is going to even that bias out because it's going to bring some of the most important lessons from history up into…”
Kipp Bodnar · 27:30

“And I think that forever we thought about search engines as the best way to access data.”
Kieran Flanagan · 29:00

#history#recency bias#knowledge#education

Takeaway32:30

Proprietary Data Will Determine Who Builds Better AI Experiences

AI does not remove the need to create, organize, and maintain valuable information. The hosts argue that companies with cleaner proprietary data and intellectual property will be better positioned to train useful models and build differentiated experiences.

AI assistants depend on accurate underlying documentation.
Organizations should organize their data before selecting use cases.
Unique proprietary data can create a competitive advantage.
Poor information hygiene limits the value of advanced models.

“right now get your data together, understand that the data that you have available, whatever you choose to leverage this technology is going to be…”
Kipp Bodnar · 32:30

“And the companies with better data and more unique and proprietary data and intellectual property will have an advantage of how they can train and…”
Kipp Bodnar · 33:00

#data strategy#competitive advantage#intellectual property#ai