So What's An "Intelligent App" Anyways?
Welcome! And let's level set...
What’s an “Intelligent App” and Why Should You Care?
There’s no question that artificial intelligence has become the “it” theme of 2022. From the release of OpenAI’s DALL-E 2 to the public in June to Jasper breaking a long drought in VC funding with its $125M Series A announcement in September, there has been a resurgence of interest in anything “AI/ML” not just from technologists and VCs, but by the general public as well.
Just take a look at RunwayML’s product demo, which quickly racked up 17K likes on Twitter, generating buzz from AI enthusiasts and video creators alike:
We are investors at Madrona, a venture capital firm that has been investing behind these themes for over a decade. We recently held an inaugural “IA Summit,” bringing together leading researchers, investors, operators, and entrepreneurs building the next generation of intelligent applications. We also recently published our second annual IA40 list, which represents the top 40 intelligent application companies.
Which begs the question - what even is an “intelligent application” anyways?
We define intelligent applications to be software services with contextually relevant machine/deep learning models embedded in the application. In other words, these are apps that leverage public and proprietary data sets and foundation models to deliver personalized recommendations, insights, and results to end users. You can contrast intelligent applications with its antithesis: “dumb” (or “not so intelligent”) software that merely performs a task the same way regardless of new or different data. We firmly believe that in the next 5-10 years, all apps will have to be intelligent or just won’t exist.
Enablers vs. Applications
While we use the term “intelligent apps” broadly, we want to distinguish between the class of tools that provide the underlying infrastructure (which we call “Enablers”) and the end-user-facing products (which we call “Applications”).
We classify Enablers between:
Data Infrastructure: Fundamental layer that enables data ingestion, storage, querying & processing, transformation, and data consumption & analysis. Having a unified and holistic view of your data is a precondition to building an intelligent application.
Models: ML models are an expression of an algorithm that analyzes data to find patterns or make predictions. Pre-trained foundation models (sometimes also referred to as Large Language Models or “LLMs”) are trained on a large corpus of multi-modal data and can be descriptive, prescriptive, or predictive.
AI Core: These are the building blocks of AI & ML deployments, including developer tools needed to build and deploy models to production. Sub-categories include AIaaS, autoML, cognitive computing, CVaaS, integrations, and fine-tuning & prompting.
Natural Language/Deep Learning Technology: Leverages computational linguistic techniques to learn from communications data and make predictions about the structure and content of language.
Intelligent applications run the spectrum between those that leverage “in-house” AI/ML models completely trained on their own proprietary data sets to those that use and/or fine-tune “off-the-shelf” foundation models to build generative apps (and everything in between). These intelligent apps span all verticals, including security, sales enablement, finance, life sciences and healthcare, travel and entertainment, and more.
We classify Applications between:
Horizontal Automation Platforms: Enable enterprises across all verticals to leverage AI to automate critical business processes via predictive analytics.
Vertical Applications in AI/ML: AI-integrated software applications that address specific problems within industries and are not always AI first. Many startups in this category design a solution to an industry problem using software and integrate AI & ML to optimize some part of their product.
Generative Apps: Leverages LLMs that are trained on a large corpus of data to produce new and realistic content (images, text, code, video, audio, and more) or even auto-complete tasks (logging new leads or data into your CRM).
Intelligent Application VC Investment Activity
VCs have been aggressively investing in intelligent applications, particularly over the past two years, as shown in the charts below. Despite VC funding decreasing substantially in 2022 across all sectors (as of Q3’22, down 53% YoY according to Crunchbase), interest in AI/ML continues to remain strong. Using Pitchbook data and our above definitions, we calculated $5B of VC funding into Enablers and $26B into Applications as of Q3 2022.
Enablers VC Investment Activity
Applications VC Investment Activity:
So Why Do I Need One More Substack Anyways??
Twice a month, we’ll surface the latest insights and themes in the Intelligent Applications world, as well as a recap of the key private companies that are making waves in this space. We also plan to highlight relevant events (both virtual and in-person), dinners, podcasts, and anything else we think you’d enjoy.
This is a quickly evolving category, and we welcome any and all feedback around the viewpoints and theses expressed in this newsletter (as well as what you would like us to cover in future writeups). And it goes without saying but if you are building the next great intelligent application and want to chat, drop us a line!
Great read! With Generative Apps, what do you test for to identify a product as a "product" vs a "feature"? For example, some of the autocomplete tasks are awesome, but unlikely to be a full-blown product in itself right?