2023 Predictions & Startup Opportunities
Examining opportunities across ecommerce, applied AI, search, gaming, and the future of work
Many holiday dinner table conversations consisted of questions such as, “Will ChatGPT start doing students’ homework?” or “look at my Lensa picture.” This new AI era has clearly reached the mass consumer with ChatGPT, but so with some toy-like applications such as Lensa. Going into 2023, I’m excited about startups that are leveraging new technology to create wonderful user experiences for both consumers and enterprises.
Throughout the next few weeks, I plan to dive deeper into these topics, as well as highlight my favorite articles I read each week. One of my 2023 goals is to post on this Substack more frequently.
Video commerce is about to explode:
With the rise in video consumption, I suspect consumers will increasingly shop and discover products through video commerce apps. Collectables and beauty have taken off (two categories where consumers value additional product details), but I expect video-based shopping and discovery through mobile apps to expand into other categories such as food, travel, and anything with a high-cost purchasing decision. Many of these applications will leverage a paradigm shift in search, which I examine below. Video commerce will continue to grow in live, long-form, and short-form content. Last year we saw livestream commerce companies WhatNot gain traction. Amazon launched a TikTok-like shopping feed that supports video-based commerce. At Madrona we led, video shopping app, Trendio’s first round of capital.
Generative AI is changing the way we interact with the internet:
A new generation of AI application companies, leveraging foundation models, are emerging. In 2023 we will see the arrival of GPT-4, the costs to train models will become cheaper, and models will become multi-modal, creating opportunity for new applications across consumer and enterprise. During the past 10 years, AI innovation has been primarily at the infrastructure layer. But in 2023 we will see value emerge at the application layer. For example, At Madrona we invested in Runway a few months ago. Marketing teams in enterprises are already utilizing Runway for their visual content creation and video editing.
AI products are becoming readily embraced by everyday consumers. Lensa may look like a toy and Dall-E may simply generate square images, but they have sparked family dinner table conversations between people that have never heard of a foundation model. These fun and easy applications promise to create massive consumer adoption and shifts in technology.
As AI becomes increasingly democratized and underlying models potentially commoditized, applications will need to differentiate based on winning critical problem spaces and amazing UX/UI. Founders will need to find a differentiated problem and solution, but these models will enable them to build faster and create magical user experiences. There are opportunities for synthetic media across image, video, data, voice, audio, and text. While this is exciting, it’s also important to note that this technology can produce toxic content including hate speech and biases. The technology is confidently wrong, so it is paramount that AI applications keep humans in the loop. They can help improve human productivity, but they do not replace human judgement, empathy, or creativity.
Here are a variety of ideas for foundation model applications that may emerge. These are in no specific order. Note: If you are building a generative AI application or working with foundation models, I’d love to chat!
Gaming content: Generative AI assets for games. Games involve thousands of different types of creative asset combinations in complex product pipelines, including textures, 2D art, 3D art, music, sound effects, animations, characters, and cinematics. Foundation models will be able to take text prompts to generate assets in games, unlocking a new level of user creativity.
Mobile gaming: Foundation models may unlock further adoption of mobile gaming. 72% of Roblox sessions happen on mobile devices, yet creation tools are primarily utilized on the PC, limiting the amount of Roblox creators. With AI tools, users will more easily generate and edit assets in games on their phone, serving a flywheel from player to creator. Roblox studio is too technical for most players. However, with text prompts to generate assets in games, new mobile-first gaming platforms may emerge.
Vertically specific design tools: Creatives across advertising, architecture, interior design, and fashion art are starting to leverage Midjourney for AI assisted process work. Each of these design verticals have different workflows and content needs. Unlocking the opportunity for vertical specific generative content can assist the designer in their ideation process. Poly for example generates color palates and customizable resolutions for textures that are valuable for interior designers and architects. UX/UI will be an important differentiator for these applications.
Ecommerce: Today, people use Honey to find a discount code or deal for a product, but the next era of computing will consist of natural language interfaces that allow people to tell their computers what they want directly, rather than doing it by hand. With ChatGPT, humans ask it a question, and it answers it for you, they don’t have to search the web to find the answer. Similarly, we may not have to search product pages to shop and make a purchase. Natural language interfaces may be able to find the best product, discount code, and even make a purchase for us online. For example, you could type “go to Nordstrom’s website and find me a dress to wear to a wedding under $200.” Or ask your personal AI assistant “remember this dress and let me know when it’s on sale.”
Hospitality: Models will be able to book trips, hotels, or restaurant reservations for us. For example, you could ask it, “where should I eat dinner Friday night in Seattle.” Based on what it has learned about you, it will make recommendations for your preferences and even book the reservation.
Workflow: Models will be able to create excel models for us. For example, I could give it sales data and ask it to create a cohort analysis from the data, saving a business analyst or investor hours of time.
Data: Generative AI will be used to help data workers ask data anything and get questions instantly. Seek.ai is already doing this with an intelligent data layer and conversational search engine creating an experience that is more productive. Microsoft CEO, Satya Nadella predicts that as much as 10% of all data could be AI generated within the next few years.
Legal, insurance, documentation: GPT for legal documentation, insurance claims, or any data-heavy manual task. For example, Lexion uses foundation models to help legal professionals create documents and related busy work.
Visual communication: People spend time communicating in Discord and Slack. This communication is heavily text focused, but as visual creations become easier, could generative AI images unlock stronger relationships and discussions between people online? Will more people communicate through images instead of words? I don’t know what this looks like as a business model, but I’m waiting for generative emojis. 😊
Search is becoming intelligent:
Generative AI is positioned to reinvent online search engines and digital assistants like Alexa. It is cheaper and faster to create high quality content with the help of generative AI, resulting in the volume of digital content positioned to explode. With the launch of ChatGPT, the way people retrieve information (often digital content) is changing. Search can be defined as the act of looking for something and synthesizing a conclusion from it. This encompasses where the user goes to discover and retrieve that information (often Google).
People rediscover and retrieve information across enterprise and consumer applications – almost every application involves search. Advances in ML unlock new data types and support search agents with contextual intelligence. Recent datasets like LAION are enabling multi-modal models across images and that can help developer teams make use of their visual data.
Large language models bring reasoning and context to search. Searching thus becomes more conversational, allowing users to iteratively search. Search will become more personalized with context, as these models take previous searches into account.
Consumer interest in TikTok increased during the past few years because of its use of AI for personalized content recommendations. 40% of Gen Z users prefer using TikTok and Instagram for internet search over Google. There’s a great Stratechery article on how Google may be doomed as consumers turn to other forms of information retrieval. We will likely also see infrastructure companies flourish as well as developer ecosystems to redesign search. The combination of reasoning and context from large language models may change the following search experiences:
Consumer public internet search (Google)
Enterprise application search (Salesforce)
Email & workflow search (Outlook/Office)
Content editing tools search (Descript)
Social media search (Instagram)
Chat forums and messaging search (Slack)
Ecommerce search (Amazon)
Traditional brands will adopt community-focused consumer behavior from Web3:
Leading digital brands will go physical and physical brands will go digital. Regardless of the underlying technology, consumer businesses can leverage the consumer behavior in Web3 around community, digital identity, and ownership. Customers want to connect their digital and physical identity, experience, and community. Brands need to build deeper relationships with their customers through new memberships, co-creation, and ways to engage their customer. Gen Z and Gen Alpha buy digital fashion in Roblox. Consumers are demanding interoperability across digital and physical experiences. Nike’s Swoosh is a perfect example. There is an opportunity for a platform that helps brands embrace this.
Marketplaces for people to re-skill:
During the Covid-19 pandemic, 40% of Americans switched jobs, prioritizing salaries, benefits, and work-life balance. Companies like General Assembly retrained workers for jobs in tech, but with massive tech layoffs and hiring freezes, an opportunity may exist for new labor marketplaces that help people find their next job.
Despite less opportunities in tech, major industries such as hospitality, healthcare, retail, and education are facing labor shortages. Many of these sectors operate with legacy tech. This creates an opportunity for new software and marketplaces that help employees connect with hiring managers and train for new job opportunities.
Additionally, Gen Z is entering the workforce, and their expectations for work represent a paradigm shift. They are more entrepreneurial, want to make an impact, and expect flexibility. They may work two part-time jobs or have a side hustle. Work and life become more fluid in Gen Z’s ideal work setting. 42% of Gen Z want to combine their passion and social life with their work. We will continue seeing the rise of the creator economy and gig-economy and software needed to fit their work process.
Additional reading:
Microsoft Bets Big on the Creator of Chat GPT in Race to Dominate A.I.: The Microsoft-OpenAI partnership is a strong indication for where generative AI technology is headed. My colleague, Matt McIlwain, was quoted in this article.
AI and the Big Five: In this Stratechery article, Ben Thompson forecasts the impact of AI on the big five in context of Clayton Christensen's The Innovator's Dilemma. Disruptive technologies bring to a market a very different value proposition than had been available previously. Generally, disruptive technologies initially underperform established products in mainstream markets.
General Catalyst, Spark in Talks to Back OpenAI Rival: Investor interest in OpenAI, Adept, and Anthropic contrast the broader pullback in startup valuations. There is true enthusiasm for foundation models and the applications being built on top.
What White Lotus, Glass Onion, and Triangle of Sadness imply about culture: Michelle Wiles explains how the definition of luxury is changing in 2023 is shifting as shown in White Lotas.
The Hottest Gen Z Gadget is a 20-year-old Digital Camera: Gen Z-ers are turning to digital cameras because they appear to be more authentic.
What Gen Z Got for Christmas in 2022: Casey Lewis highlighted a key TikTok trend, Gen Z is showing what they got for Christmas, and it is surprisingly all the same. Despite valuing individuality, the generation is still choosing to follow specific trends.
How We Learned to Be Lonely: Many people have continued to spend time alone despite the pandemic winding down. The Atlantic explains we are in the midst of a loneliness crisis.
Pickleball popularity exploded last year, with more than 36 million playing the sport: Pickleball has become a compelling easy to learn sport for adults. Will we see the Nike or Arc’teryx of Pickleball?
Why Beauty is the New ‘It-Bag” for Luxury E-Commerce: With continuing Zoom meetings, WFH, TikTok trends, beauty continues to be a growing category with luxury apparel e-commerce companies, Moda Operandi and Farfetch entering the category.