Generating Success: How Generative AI Can Elevate Your Startup's Growth
Generating Success: How Generative AI Can Elevate Your Startup's Growth
Learn about generative AI, ChatGPT unit economics, AI prompt engineering, and beyond as we cover the key trends driving the future of startup growth
Hi Friends 👋
Welcome to a brand-new edition of Growthstore. xyz! 💥
Every other week, we bring real-life startup experiences and knowledge to help you take your business to the next level. Our community is a hub for startups, PMs, marketers, and designers to connect and learn from one another. 💡
If you've missed any of our latest articles, here's a quick recap of what we've covered recently:
Be sure to subscribe to receive all our full articles and support Growthstore.xyz
Today, I want to talk about something on my mind lately: Generative AI. 🤖
I'll be honest, when I first heard about ChatGPT, I thought it was just another tool for writing SEO articles. But when I got my hands on it, I was blown away by its capabilities. It made me realize that large language models like GPT-3 can potentially transform how startups operate and grow.
Sam Altman, the CEO of OpenAI, made a prediction about AI that caught my attention.
This got me thinking about the implications of AI on the future of work and how startups can leverage generative AI to stay ahead of the curve.
But here's the thing: I've noticed that many people still view generative AI as a novelty or toy, not realizing the actual value it can bring to businesses.
This essay will explore the power of generative AI, startup opportunities, and how to implement it in your business. We'll also discuss building relevant products and the importance of AI prompt engineering in the age of current AI models.
Topics covered:
Advantage ChatGPT: Unit economics
AI and ChatGPT - The New 'iPhone' Moment for Tech Software
AI Prompt Engineering — The skill you need in the age of AI
What is a Prompt?
How ChatGPT's API is Disrupting the Text Generation Startups
Conclusion: Building AI-Native SaaS products that won't be obsolete
This article aims to help experienced SaaS operators and newcomers understand generative AI and adjust their startup strategies in a rapidly evolving tech landscape.
So, let's dive in!
Advantage ChatGPT: Unit Economics
It's incredible how far LLM (Large Language Models) technology has come in just a few short years.
What is LLM? - A Large Language Model (LLM) is a class of artificial intelligence models designed to process and generate natural language text data. These models are usually large deep neural networks that use complex algorithms to analyze vast amounts of text data, learning patterns and structures within that data in order to generate new text.
Back in 2021, when GPT-3 was released, the cost for 1,000 tokens was a staggering $0.06.
That's a pretty penny, especially for bootstrapped startups and entrepreneurs trying to realize their ideas.
But when we thought affordable language processing was just a pipe dream, OpenAI stunned us all in August 2022 by introducing text-davinci-003, a finely-tuned version of GPT-3 that excels at following instructions and lowering the cost to $0.02/1,000 tokens.
But that's not all.
The real breakthrough came on March 1st, 2023 —when OpenAI set the price for ChatGPT at just $0.002/1,000 tokens.
That's 1/10th of the cost of the GPT-3 API!
If you're still worried about the cost of using ChatGPT's API, fear not! Dr. Jim Fan reminded us that 1k tokens only cost $0.002.
According to OpenAI:
The new model family, called gpt-3.5-turbo, is the “best model for many non-chat use cases
Quoting Max Woolf:
Let's remember the impact on companies like Jasper and Copy.ai. These companies wouldn't exist without the APIs for GPT-3
Now, with ChatGPT, the possibilities are endless. Startups can truly level the playing field and compete with established players in the market.
ChatGPT's API is so good and cheap that it makes most text-generating AI a commodity
This is a game-changer for startups and entrepreneurs who previously may have been priced out of the market for such cutting-edge technology. With these new unit economics, even the smallest startup can leverage the power of AI to create innovative products and services that were previously out of reach.
AI and ChatGPT - The New 'iPhone' Moment for Tech Software
OpenAI's ChatGPT is a game-changer for the AI industry, just like the iPhone was for smartphones. With its user-friendly interface and powerful features, it has the potential to revolutionize the way we work and live.
However, the real power of the AI revolution lies in its trickle-down effect.
As more and more people adopt these tools and incorporate them into their workflows, we'll see new use cases emerge, new industries disrupted, and new possibilities unlocked. Just like the iPhone's impact was felt not just by early adopters but by everyone who used smartphones, the effects of ChatGPT and AI will be felt by everyone who interacts with technology.
The upcoming release of a multimodal GPT-4 will allow AI to process text inputs, images, audio, and video.
Here are some potential use cases:
Personalized healthcare diagnoses and treatments
More accurate e-commerce product recommendations
Immersive entertainment experiences with personalized interactive video games and movies
The widespread adoption of ChatGPT by companies like Snapchat, Slack, and Instacart is proof of its potential.
Quoting Max Woolf:
It won't surprise me if every consumer-facing tech company does something with ChatGPT to look like they're cutting edge to their investors. Some have compared the sudden mass adoption of AI to chasing a fad, like how companies were randomly embracing web3/crypto/metaverse/NFTs a year ago. But unlike those which were a solution for a problem that didn't exist, generative text AI does work, and there is an actual demand from people outside of its die-hard supporters for it to work.
The impact of ChatGPT is not just limited to big companies.
Startups like Copy.ai, Jasper, and Tably use it to disrupt their respective industries. Copy.ai is revolutionizing the world of copywriting, while Jasper provides a more efficient way of conducting customer research. Tably is enabling users to generate investment insights through natural language conversations.
The possibilities of ChatGPT and AI are endless.
And the best part is that you don't need to be an AI expert to use it.
All you need to do is learn how to use prompts. Prompting is providing ChatGPT with specific information to generate a response. With a bit of practice, anyone can use ChatGPT to their advantage.
So why not start exploring the possibilities of AI and ChatGPT today? The next big idea could be just a prompt away.
AI Prompt Engineering — The skill you need in the age of AI
Generative AI is gaining popularity. ChatGPT and other art generators like DALL-E 2, Stable Diffusion and Midjourney are becoming famous for their capabilities.
But, generating output that matches the desired result is still a challenge.
For example, it may not produce content that aligns with truth, insight, reliability, and originality goals. Moreover, it lacks common sense and fundamental knowledge of the world, which can lead to flawed and nonsensical content.
Enter prompt engineering:
The art of crafting the correct input to get the output you want from generative AI. With prompt engineering, you can steer the outcome towards your desired direction.
Whether building a startup from scratch or developing products on LLM Generative AI like Jasper, AI Prompt Engineering can help you achieve your goals. Using trial and error and keeping a few key things in mind, you can create prompts that generate content that aligns with your business goals.
Check out the collage below to see how we started with a simple prompt and iterated until we got closer to what we had in mind. See how details were being modified, added or removed with each iteration, all with improving text in the prompt. Seeing how prompt engineering can lead to such beautiful results is fascinating.
But where did the spell-casting metaphor for AI prompt engineering come from? It was first introduced by GPT-3 researcher and OpenAI co-founder Greg Brockman, who likened generating text with AI models to casting spells. The metaphor has gained traction in the AI community, with some finding it helpful while others have criticized it as potentially harmful.
If you want to hear more about this debate, check out this snippet from the Changelog podcast episode, where the hosts discuss the pros and cons of using the spell-casting metaphor in AI.
Expanding the spell-casting metaphor—here’s the key takeaway:
Writing prompts, unlike writing regular code. No API reference or programming language specification will let you predict precisely what will happen.
Instead, you have to experiment: try different fragments of prompts and see what works. Then, as you get a feel for these fragments, you can explore what happens when you combine them.
Over time, you will develop an intuition for what works. Then, you’ll build your collection of fragments and patterns and exchange those with other people.
The weird thing about this process is that no one can truly understand exactly how each fragment works—not even the creators of the models.
It's important to note that while the spell-casting metaphor can be a helpful way to conceptualize generating prompts with AI models, it should not be taken too literally.
With a better understanding of how prompts work, we can now explore how they can be used to generate unique and impressive outputs. So, let's dive deeper into what a prompt is and explore some examples.
What is a Prompt?
According to Simon Willison:
Communicating clearly with AI systems is as hard as communicating with others. You need to ensure that the AI system understands the context and language used, just like figuring out what someone already knows or doesn't know when talking to them.
To be a great author of prompts, you need to be meticulous. You constantly run experiments, take detailed notes, and determine which components are necessary for the prompt to work and which are just a waste of tokens.
Simon also suggests that you must be a bit of a scientist when approaching prompt engineering. Figuring out what works and what doesn't when dealing with the world's most complicated black box system is a considerable challenge. And it's essential to resist superstitious thinking, as humans and LLMs are pattern-matching machines. You don't want to learn the wrong lessons from a successful prompt entirely.
So, if you want to craft effective prompts for AI generators, take Simon's advice and approach it systematically and scientifically.
Key elements to include in a prompt for AI art generation:
Subject: represented by nouns suggests to the AI system what scene to generate.
Description: implies additional information related to the subject, such as adjectives (stunning, lovely), background description, etc.
Style: indicates the theme of the image. It can include artist names (Picasso, etc.) or custom styles such as fantasy, detailed, modern, contemporary, etc.
Graphics: stands for computer graphics engine type that enforces the effectiveness of the image. Such keywords include unreal engine, 3D rendering, and octane render.
Quality: indicates the image quality, such as 4K, 8K, or HD.
These fundamental elements help the AI model generate specific outputs and lead to awe-inspiring works of art and writing.
Here's an example of a prompt sequence used to generate the banner image for this essay:
Orange color rose with water, photo
Orange color rose made of flames, photograph
Orange color rose made of flames and water in a balanced composition, photograph
A balanced and artistic composition with a rose of orange color in the middle of the frame. Rose flower looks floating and showcases the velvety texture and rich color of its soft petals
Realistic fire, colorful smoke, and water fuse like a beautiful orange-colored rose. Rose flowers showcase the velvety texture and rich color of soft petals and Flames. brightly lit, photorealistic
Realistic fire, colorful smoke, and water fuse like a beautiful orange-colored rose. Rose flowers showcase the velvety texture and rich color of soft petals and Flames. brightly lit, photorealistic, 4k
Realistic fire, colorful smoke, and water fuse in the middle of the frame taking the shape of a brightly lit and beautiful orange-colored rose. Rose flowers showcase the velvety texture and rich color of their soft petals, and Flames, and smoke is visible in the frame. 4K, dramatic lighting, brightly lit, photorealistic
Clear and effective prompts are essential for applications that integrate generative AI models into user-facing products.
With the rise of AI-first startups, such as Jasper and Copy.ai, there is a growing demand for text generation services that can automate content creation at scale. The following section will explore how AI-first startups disrupt the text generation industry and drive innovation in this space.
From Jasper to Copy.ai: How ChatGPT's API is Disrupting the Text Generation Startups
Large language models (LLMs) are making big waves in the AI industry.
Tech giants like OpenAI, Microsoft, Google, and Meta (formerly Facebook) lead the charge in developing next-gen models. These LLMs are integrated into various products and services such as Bing Search, Bing Chat, Microsoft Teams, Google's Workspace and more.
The real game-changer, though, was ChatGPT. It exploded onto the scene and snagged a million users in just four days!
But here's the exciting part— this trend means following:
This is a massive opportunity for startups and we’ll witness the emergence of AI-first startups like copy.ai and Jasper
New industries prioritize automation over human involvement, paving the way for the next billion-dollar valuation startups.
It's not just us saying this - according to Sacra.com report, Jasper and Copy.ai are two prime examples of startups that have leveraged GPT-3 to achieve significant growth. As you can see from the graph below, they raked in a combined $83 million in ARR in 2022 alone!
Building features vs. apps is crucial for startups working on generative AI products.
These products take various forms, such as desktop apps, mobile apps, Figma plugins, Chrome extensions, and even Discord bots, and it's easy when integrated with tools users already work with. However, the challenge is determining which of these will become standalone companies and which will be absorbed by incumbents like Microsoft and Google
Let's look at 5 AI-first startups using ChatGPT's API to disrupt the market:
Jasper.ai - Jasper started as a copywriting tool for Facebook and Google ads but quickly gained traction by targeting its pre-existing community of market. As a result, the As a result, the company overgrew, reaching $42.5M in ARR within its first 12 months, and is projected to surpass $75M in 2022.
Copy.ai - Copy.ai built five MVPs on GPT-3, with CEO Paul Yacoubian driving initial growth through viral "building in public" tweets. It grew from $2M ARR to $10M in 14 months and is projected to reach ~$11.6M ARR by the end of 2022.
Typeface - Former Adobe CTO's startup, Typeface, has raised $65 million to develop an AI-powered content marketing platform that can help businesses create personalized content at scale. The platform is built on OpenAI's GPT-3.5 and a customized version of Stable Diffusion 2.0, and it can generate personalized blogs, Instagram posts, and websites for companies. It uses a company's existing content, brand logos, and other visual assets to train its model and create content tailored to each enterprise's brand voice and audience.
Robin.ai - Robin.ai is a generative AI startup that helps businesses automate their writing workflows. With ChatGPT's API, Robin.ai can create custom writing models that mimic your brand's voice and style.
Shortwave - Shortwave is an AI-powered email app that uses ChatGPT's API to generate intelligent summaries of your emails. With Shortwave, you can save time and quickly understand the essential points of each email.
In addition to text-generating startups, many AI-based image-generation startups are gaining popularity and are focused on building personalized images for marketers and design matching their brand and design language.
In conclusion, the AI industry is evolving rapidly, and LLMs are at the forefront of this revolution. Startups that are quick to adapt and harness the potential of LLMs have a chance to create groundbreaking products and disrupt the market. However, it's essential to remember the intense competition in this space, with big players like Google and Microsoft vying for market share.
To stay ahead of the curve, startups must focus on building products that won't become obsolete in a rapidly changing technological landscape.
The following section will explore how to create cutting-edge products and keep up with the ever-evolving AI landscape.
Building AI-Native SaaS products that won't be obsolete
The tech industry is always on the move, especially AI.
This week, OpenAI dropped GPT-4 on us, proving that even the most advanced AI technology never stops evolving.
It's wild to think that what we thought was cutting-edge just a few years ago is now obsolete. But that's just how quickly things move in this space. As a startup founder, keeping up with the latest trends and ensuring your products won't become outdated in just a few months is essential.
But this rapid pace of AI innovation also presents an excellent opportunity for AI-native startups to create products that will stand the test of time.
So, what's the secret to creating AI-first native products that won't be obsolete in a few months or years?
Here are the 7 key considerations to keep in mind.
Collect proprietary customer data and fine-tune the generic models.
Focus on creating a great UX that instils a sense of trust and reliability.
Develop a strong distribution strategy and go-to-market plan.
Build a community of practitioners or partners to support your product.
Ensure breadth and depth of integrations into customers' systems and workflows.
Focus on the end-to-end use case for your target persona.
Understand trade-offs, choose the right tech stack, and adapt to changing underlying models to make data your moat.
Here's what I mean by the breadth and depth of integrations: More than thin layers around LLM APIs are needed to gain a competitive edge in AI-first apps. To win, you need deep integrations and optimized workflows that solve real problems with the scalability and efficiency that wasn't possible before LLMs.
For example:
Copy.ai sees AI-powered content as a wedge into building a HubSpot-like growth automation platform that hooks into your software stack and automates decision-making with a human in the loop. Rather than just writing Facebook ads, Copy.ai's platform would run, and A/B test them, start/pause campaigns as needed, bid for impressions on its own, and modify the campaign text and creative to get better results, creating a powerful moat around its product.
Jasper, on the other hand, is aiming to use their fine-tuned models to build a unified AI experience widget that's deeply embedded into all enterprise workflows like sales & marketing, support, HR, legal, and finance, similar to how Grammarly ($13B) hooks into any app where users interact with the text. Jasper's Chrome extension is the first step towards this, and it expects 95% of future usage to happen inside other apps rather than its standalone web app, creating a powerful moat around its product.
In conclusion, the advent of AI-first apps revolutionizes how we work and collaborate.
The possibilities are vast and expanding from note-taking and copyrighting to end-to-end SDR automation, code generation, customer support automation, medical/health assistants, and education.
While Microsoft, Google, Meta and Adobe have the advantage of bundling AI as a feature for near-free in their apps, non-AI native incumbents risk stapling AI on as an afterthought rather than weaving AI into the foundation of their products.
This disadvantages them compared to AI-native apps when building a cohesive, end-to-end AI product experience.
The key takeaway: Startups must understand the trade-offs of product development, choose the right tech stack, and adapt to changing underlying models to make UX and data their moat. Deep integrations and optimized workflows that solve real problems with scalability and efficiency are necessary to gain a competitive edge in AI-first apps.
As we move forward, it's important to challenge assumptions and embrace the transformative power of AI-first apps.
With new interfaces, CRMs, tax prep copilots, and research assistants all fair game, the possibilities for innovation are limitless.
The question is, which companies will rise to the challenge and build the next generation of AI-native products?
On Twitter 🐦
Ai Startups and Products worth a look 👁️
TypingMind - A user interface improvement for ChatGPT
Ellie - A Powerful Al email assistant
Deep Agency - An Al Photo Studio & Modelling Agency
AvatarAi.me - The platform to create your perfect profile photo
Unstudio.ai - A tool to generate stunning visuals for your D2C products
Recommended Reading 📚
A guide to getting started with PLG
Why Building Your Startup's Brand From Day 1 Matters
Report on Jasper
The inside story of how ChatGPT was built
The AI Scaling Hypothesis
In conclusion, the rise of AI-first startups presents both challenges and opportunities for SaaS companies.
The real winners will be the ones who can create seamless integrations and optimized workflows, solve real problems with scalable efficiency, and ultimately build a strong moat around their product.
As Andrew Ng said, "AI is the new electricity," and those who can harness its power will be the ones who shape the future of our digital world.
And that's it for this week! If you enjoyed this article, subscribe to our newsletter for more insights and tips on growth stories. And follow me on Twitter
Subscribed
Remember to share this post with anyone who might find it valuable. Thanks for reading, and we'll see you next time on Growthstore.xyz!
Until next time 👋
- Sri
PS: Midjourney and ChatGPT generated some of the images and content in this essay.
Love Jasper (formerly Jarvis). 3 years and counting and I'm still learning the best ways to get the best outcomes.
Great read, Sri. I was trying out Tome after seeing their GPT 4 launch video and was mind blown with what it could do. As a marketer, I feel the landscape has just expanded and more importantly, there is going to be wave of democratization of content.