31 May 2023

AI Shark Tank: 5 Founders Pitch Us Their Ai Products (#124)

2Frameworks

12Insights

Frameworks in this episode

Finance6 steps

Early-Stage AI Product Investment Scorecard

Rank AI startups by defensibility, use-case value, market potential, and team quality.

Read →

Marketing7 steps

Product-Led Demonstration Loop

Market a product by using it to create a public, measurable result.

Read →

Insights & moments

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

Myth Buster· 1

Myth Buster19:30

Travel AI Should Need Context, Not Endless Form Fields

Where To AI generates destinations and itineraries, but its click-heavy setup frustrated the hosts. They argue that an AI travel product should accept natural-language context or a hotel URL, infer the destination, and build an itinerary around where the traveler is actually considering staying.

AI interfaces should reduce repetitive clicking.
A hotel URL contains useful location and trip-planning context.
Itineraries should account for nearby activities and trip duration.
The strongest assistant would eventually handle packing and reservations.
Conversational input can replace rigid preference forms.

“We don't click anymore in AI.”
Kieran Flanagan · 20:00

“You should just be able to send it the URL of the hotel and then it will figure out the destination”
Kieran Flanagan · 21:30

#travel ai#user experience#conversational ui#itineraries

Hot Take· 3

Hot Take07:30

A Content Chatbot Is Magical but Not Yet Business-Critical

My AskAI can ingest a website and provide a ChatGPT-style interface that answers questions across its content. The hosts like the low-friction external use case but argue that content discovery alone is not transformative; product-page monetization, internal knowledge retrieval, and newsletter intelligence may create more valuable applications.

A complete blog can become a queryable conversational interface.
External content avoids some internal security and permissioning friction.
Blog search is useful but may not materially change growth or productivity.
Product information and internal organizational knowledge could support higher-value workflows.
The company needs a use case customers will reliably pay for.

“the external use case is really cool, I think the internal use case is even cooler and more valuable.”
Kipp Bodnar · 09:30

“It's not like a game-changer for your business growth or your productivity.”
Kipp Bodnar · 10:00

#knowledge base#chatbots#monetization#content discovery

Hot Take16:00

AI Could Rapidly Commoditize Successful Website Designs

The hosts note that sketch-to-code and Figma-to-code tools make it easy to reproduce the structure of high-performing pages, adapt them to another design system, and publish them quickly. This democratizes creation but also reduces the durability of design execution as a competitive advantage.

Successful page structures can be reproduced quickly.
Teams can translate copied formats into their own visual systems.
Code generation sharply reduces implementation time.
More people gain the ability to create products and websites.
Distinctive design execution may become easier for competitors to imitate.

“Anybody could literally just go copy the format of that page, just put that page in Figma, import it into Kajoo, and boom.”
Kipp Bodnar · 16:30

“you can take elements of any page and just quickly take the code and put them back together again.”
Kieran Flanagan · 16:30

#commoditization#web design#competition#code generation

Hot Take31:30

AI Will Free Us for Higher-Value Work—Then Automate That Too

The episode closes with an intentionally dark joke about the trajectory of automation. AI first removes disliked routine work so people can concentrate on higher-value tasks, but increasingly capable systems may eventually automate those higher-value activities as well.

Routine work is the first target for automation.
Workers initially move toward more valuable tasks.
AI capability may continue climbing the value chain.
The hosts treat eventual human displacement with humor rather than certainty.

“they've automated all the way, all that kind of work, so we can do higher value tasks”
Kieran Flanagan · 31:30

“and then they'll automate us away so they can do more high value tasks.”
Kieran Flanagan · 32:00

#automation#future of work#ai#productivity

Explainer· 4

Explainer02:30

Why Personalized Video Is an Early AI Marketing Breakthrough

Carvana generated 1.3 million personalized customer videos using purchase, location, timing, and preference data. The hosts argue that video, rather than text, could be the first major generative-AI unlock for marketers because complex technical work can produce a simple and memorable customer experience.

Carvana personalized videos at a scale of 1.3 million customers.
Each video incorporated customer and purchase context.
The infrastructure was complex, but the customer interaction felt effortless.
The hosts expect AI-generated video to transform marketing.

“Carvana used AI to hyper-personalize 1.3 million videos for their customers.”
Kipp Bodnar · 02:30

“there's a lot of crazy things happen in the background to make the one/two minute interaction feel completely amazing.”
Kipp Bodnar · 03:30

#personalization#video#ai marketing#customer experience

Explainer12:00

Turn Newsletter Archives into a Personal Research Bot

The hosts propose automatically importing a reader's favorite newsletters into a private database that can be queried conversationally. A public variation could let creators convert their archives into branded bots that keep subscribers engaged with years of accumulated ideas.

Aggregate newsletters from multiple publishing platforms.
Make the full archive searchable through natural-language questions.
Let users choose and combine trusted sources.
Offer creators an interactive way to reactivate old content.
Use public bots as acquisition and private databases as a paid service.

“I want them every week ported into-”
Kipp Bodnar · 12:00

“I just wanna be able to query this bot for anything I wanna learn from those people.”
Kipp Bodnar · 12:00

#newsletters#knowledge management#creators#ai search

Explainer26:00

Deepsheet Must Move Fast Before Platforms Absorb Its Feature

The hosts see a near-term opportunity for Deepsheet because OpenAI Code Interpreter was not yet broadly available. They also identify a major strategic threat: foundation-model providers and established analytics companies can build similar natural-language interfaces, making rapid user acquisition and deep integrations crucial.

Limited access to Code Interpreter creates a temporary opening.
OpenAI represents a direct substitution threat.
Analytics vendors can add conversational interfaces to existing products.
Integrations with Looker, Amplitude, Google Analytics, and Google Sheets could increase utility.
Fast execution is necessary while the market window remains open.

“If I was this gentleman, I would go fast here.”
Kipp Bodnar · 26:30

“Integrate to Looker, integrate to Amplitude, integrate to GA.”
Kieran Flanagan · 26:30

#platform risk#analytics#integrations#competitive strategy

Explainer27:00

Shared Data Analysis Could Create a Viral Alternative to Excel

Instead of exporting a static chart, Deepsheet could let recipients open the underlying analysis, manipulate the same data, and challenge another person's interpretation. This turns analytical disagreement into collaboration and gives each shared result a path back into the product.

Shared results should remain interactive.
Recipients should be able to test alternative interpretations.
Debate creates repeated engagement with the same dataset.
Every shared analysis can introduce another user.
The interaction would be difficult to reproduce with static Excel files.

“you can go and manipulate that same data the way you want and challenge how I looked at it.”
Kipp Bodnar · 27:00

“The back and forth that you could create on that would be awesome. That people just don't do in Excel.”
Kipp Bodnar · 27:30

#collaboration#data analysis#viral loops#sharing

Tool· 3

Tool04:00

Capsule Turns Transcripts and Prompts into Finished Video

Capsule transcribes uploaded footage, creates title cards, generates relevant images, and lets users refine edits through prompts. The hosts see it as a video-oriented counterpart to Descript that substantially lowers the editing-skill requirement, while noting that it cannot rescue fundamentally unengaging source footage.

The editor runs in a browser without extensions.
Users can edit video by selecting transcript passages.
AI generates summaries, animations, and supporting images.
Natural-language controls reduce the need for specialist editing skills.
Compelling source footage remains essential.

“what you can basically have is take raw video footage and turn it into a very professional looking video with AI.”
Kipp Bodnar · 05:00

“So you need a minimal amount of video editing skills to be able to do this stuff”
Kieran Flanagan · 05:30

#video editing#generative ai#natural language#content creation

Tool13:00

Kajoo Converts Sketches and Figma Designs into Responsive Code

Kajoo turns a rough hand-drawn wireframe or imported Figma design into responsive components and Next.js code. It can also apply an existing design system, allowing users with limited coding ability to move from an idea or established visual pattern to a working component within minutes.

Kajoo accepts both hand-drawn sketches and Figma designs.
AI makes initial responsive-layout decisions.
Imported design systems can restyle generated components.
The output includes editable Next.js code.
The workflow expands who can create functional web experiences.

“in a couple of minutes I have my component ready, either from a hand-drawn sketch or with Figma”
Presenter · 15:00

“Just, how many more people can create things.”
Kieran Flanagan · 17:00

#figma#code generation#web design#nextjs

Tool24:00

Deepsheet Lets Non-Analysts Question and Visualize Data

Deepsheet accepts an uploaded dataset, shows a sample for validation, and creates charts from plain-English instructions. The direct demonstration immediately convinced Kieran to seek out the product because it addressed an active data-analysis need with almost no apparent learning curve.

Users can validate a dataset immediately after upload.
Natural-language prompts replace manual chart construction.
The tool produces visualizations quickly.
A short, direct demo made the value obvious.
The capability is useful but exposed to platform competition.

“Deepsheet is an AI tool that lets you interrogate your data”
Dylan · 24:00

“you had my at, like, natural query of data and building charts that easily.”
Kieran Flanagan · 25:30

#data analysis#visualization#natural language#analytics

Takeaway· 1

Takeaway23:00

Trip Planning Products Should Be Social by Default

The hosts criticize the travel app for producing only a static sharing URL even though trips are commonly planned with other people. Invitations, shared decisions, and collaborative editing could improve the product while creating a built-in referral loop.

Travel is inherently collaborative.
Friends should be able to review and modify a proposed itinerary.
Invitations can introduce new users to the product.
A mobile workflow would make sharing hotel links and plans easier.
Basic growth mechanics could materially expand adoption.

“trips are inherently a social activity”
Kipp Bodnar · 23:00

“There should be a viral coefficient to this product, where people are inviting others to this trip planning”
Kipp Bodnar · 23:00

#travel#collaboration#viral growth#product design