#PowerToTheAIBuilder — AI Will Be Core to Every Software Application
By Veronica Orellana, Brittany Walker and Vivian Cheng
“Suppose you just had a nasty fight with your boyfriend. The algorithm in charge of your sound system will immediately discern your inner emotional turmoil, and based on what it knows about you personally and about human psychology in general, it will play songs tailored to resonate with your gloom and echo your distress.” — 21 Lessons for the 21st Century, Yuval Noah Harrari
This passage by Yuval Noah Harrari came out in 2018, and at the time it sounded incredibly futuristic. The strides AI has made in even the last 6 months, however, have made personalized music based on emotions no longer sound distant, but rather, a near-term reality.
It’s hard to deny the technological shift we’re experiencing from foundational models. They’ve seen rapid growth over the last three years; from OpenAI GPT utilizing 110M parameters in 2018, to Google’s latest Switch C model utilizing 1.6T parameters. There has been unparalleled developer and user adoption for solutions like Stable Diffusion, Dall-E and ChatGPT, all of which surpassed 1M users in less than two weeks.
The application layer has exploded with startups offering solutions for marketers, sales reps, creators, coders, software designers, lawyers, nurses, interior designers, photographers, etc. AI will be the future co-pilot of every professional. Companies across every vertical, be it sales and marketing, supply chain, healthcare, insurance, etc., are all starting to reap the benefits of AI. Even our parents have created their AI generated selfies using Lensa, and used Jasper to help fine-tune their emails.
What Incumbent App Layer Companies Think About AI
We’ve spent the past several weeks at CRV debating the generative AI market. Will value be captured by startups or incumbents? Do incumbents even care to implement this technology? How do you build differentiation?
At CRV we’ve invested in many generational application layer companies like Airtable, Iterable, Zendesk, HubSpot, Drift and more.
We’ve also been actively investing in AI focused companies that are revolutionizing these spaces, and looking to back more!
We chatted with several product and AI leaders within our portfolio (thanks to those who shared their time with us!) and wanted to share their perspective on generative AI.
We covered a lot of topics, but a couple of main themes emerged from our discussions:
- Generative AI is here to stay and will play a role in the future. It was clear from all of our conversations that product leaders within application layer companies are all thinking about ways to integrate generative AI into their products. We’ve already seen companies like Canva, Figma, Zapier and Adobe release generative features, and we anticipate that many more will follow. The question remains on whether AI will become core to their overall product strategy or if their venture into AI will stop with these thin layers/plug-ins.
2. The barrier for incumbent companies to adopt AI will not be a technological challenge, but rather an organizational one. Unlike the move from on-prem to cloud, which required huge tech investments, open source LLMs have made it easy for companies to incorporate generative AI into their product. The bigger challenge for larger organizations will be one of prioritization and organizational structure. As the team at one CRV backed company pointed out, “companies that were not AI-centric from the beginning will likely face the innovator’s dilemma of how much do you sacrifice existing business vs. becoming a leader in the next generation.” Org structure will likely play a role in this as well — do you have a separate AI team, or is AI a key pillar across all product and engineering teams? The more intertwined AI is across an organization, the higher likelihood that incumbents will release core AI-native features.
3. Domain knowledge, a workflow-centric approach and good distribution are critical to success. As a product lead on the AI team at HubSpot pointed out, “startups and incumbents get equal access to these models, so execution, good go to market and strong workflows will set you apart.” Another way to build differentiation is to verticalize since it requires a more unique data set and strong domain knowledge. As startups continue to pop up, incumbents have the advantage of sitting on lots of customer data and a platform through which they can more easily distribute their offering.
4. Proprietary data leads to hyper personalization, and the more personalized, the stronger your moat. When speaking with Iterable, it was clear that their customers want personalized models. Getting, “closer to the semantic understanding of each vertical, like media or entertainment, is key to being more nuanced than just marketing language.” Going a step further than that, Iterable imagines a world in which AI will optimize communications for each individual consumer, suggesting hyper-personalized and rich messaging and media. This optimization strategy is only as good as the data to train it on — the more users you have, the more feedback data you collect. We are already seeing Jasper use Cerebras’ Andromeda AI supercomputer to create personalized models for enterprise customers that self optimize based on usage data and past performance.
Foundational Models Are Supercharging All Forms of AI
Besides generative AI, we are excited about foundational models that are supercharging our abilities in search, RPA, audio, computer vision, image recognition and more.
In “sleepier” industries like manufacturing, supply chain, financial services and healthcare, there are tons of repetitive and manual tasks that still get done on spreadsheets or even pen and paper. We anticipate that many of these industries will significantly benefit from AI capabilities that are not necessarily generative.
In healthcare, we believe RPA, computer vision and speech-to-text capabilities will be big needle movers. We’re already seeing companies like ImVitro and Alife apply computer vision to the IVF fertility imaging process to speed up efficiency and reduce human error. In dental, companies like Overjet are applying AI to help with imaging analysis and early dental disease detection. To combat Electronic Health Record (EHR) fatigue, which has become a large problem in the medical industry, companies like Deepscribe are leveraging speech to text capabilities and automation to reduce the amount of note taking and manual input that doctors must do on a daily basis. We are also seeing several players emerge in the customer support and billing segment of healthcare (Outbound AI, Birch AI, MajorBoost), historically a super manual and time intensive process,
In manufacturing, we anticipate computer vision, image recognition and speech continuing to play large roles. As nearshoring continues to rise, computer vision will be critical in helping companies utilize their facilities 24/7. Camera vision and sensors can be used in warehouses to maintain inventory status, alert managers when a production line is down, assist in quality control, repair operations and more. With the rise of robots in manufacturing firms, we anticipate seeing more AI embedded into these platforms as well. On the workflow side, companies like Datch are enabling frontline workers to use voice as a means to capture information and automate industrial processes.
In consumer, AI has the potential to be the next paradigm shift particularly in how we search and consume information. For the first time, Google’s core business is vulnerable as ChatGPT, and startups like Hebbia and You reinvent what it means to index and surface the “right” information and provide context. On the private data layer, Rewind.ai, Supernormal and others are cataloging all of our digital and voice communication in the workplace to surface and store new data that hasn’t been captured previously. Delving deeper into the co-pilot for X thesis — we anticipate that AI will cut cost curves for many consumer services. Companies like Speak will transform the way we learn languages. We’re looking for more exciting AI applications in fields like interior design, education, music creation, renovation and many more — the potential is unlimited.
We are also excited about companies leveraging powerful semantic search to surface information across multiple sources and platforms. Companies like Viable and Enterpret allow product management teams to search and synthesize customer feedback from support tickets, social media accounts, surveys, etc. to better understand customer pain points and inform product roadmaps. In the legal space, companies like Fileread, Harvey AI, Lexion are enabling lawyers to supercharge their search by asking specific questions, surfacing relevant information, summarizing clause language, suggesting redlines, etc.
As McKinsey’s State of AI in 2022 report points out, all industries will see meaningful impact from both revenue generation and cost reduction. Unsurprisingly, the less digital industries will likely see stronger cost reductions compared to more digital industries like sales and marketing. We are eager to see which markets create the largest AI-native companies.
What Areas Are We Excited About at the App Layer
Despite incumbents adopting this technology, we still believe there is a large opportunity for startups to capture value. At the application layer, there are a couple of areas that we are particularly excited about at CRV:
- Vertical Solutions
Owning a targeted customer base and use case means having a deep understanding of their workflow and integrating with relevant platforms. This tends to result in high product engagement, strong data feedback loops and overall stickier products. Verticalized data sets (healthcare, insurance, legal, construction, etc) tend to be harder to get access to. Thus, there is significant advantage in being able to fine-tune a model using this specific data set, and building a data moat over time through product engagement.
Companies like HighArc are automating workflows for the construction industry, Flux AI is automating the development and design of hardware products, Birch AI, which automates the complexity of call centers for the healthcare industry and Cradle Bio which is auto-designing better proteins.
2. Net New Mediums & Workflows
Generative AI has enabled the creation of new workflows and new mediums through which we can distribute and create content — synthetic video (Synthesia, Rephrase AI, Tavus), synthetic audio (Resemble AI, Well Said Labs, Sanas, Rime), synthetic music (Harmonai, Musico), etc. These technologies are changing the way that marketing, sales, media companies, gaming companies and even learning and development teams create content. In this specific category, we believe that startups will capture the vast majority of the value, since no incumbent is well positioned.
As companies continue to become more global, being able to translate content into any language in real time is highly valuable. As is the ability to generate videos and audio from a short snippet instead of spendings tons of money on an entire video production. We’re eager to see how quickly these technologies become mainstream.
We believe that we’re in the early innings of these new processes. Entire workflows will be disrupted, as will careers. We are not entirely sure what these new forms will necessarily look like, but are excited for the ways in which AI will change our behaviors.
3. Functional Platform Approaches
We are equally excited about companies that are targeting a specific functional group with an AI-native approach in mind. Building new workflows and integrations that are enabled and supercharged by AI is an exciting way to displace incumbents. As customers interact and put more of their data into a product, AI-native companies will capture this value, create personalized output for each customer, and build a data moat over time. We love what companies like RunwayML are doing in the video creation space, Regie in the sales world, Monterey in the product management space and Modyfi for designers.
What Makes an AI Company Stand Out
We’ve outlined below our investment framework for the application layer. We believe that defensibility can be created at the product level with innovative workflows and data moats. Teams building at the application layer will need to deeply understand the core workflows of their users and have a strong nose for product and distribution strategy. For average contract values (ACVs) to grow, products must become multiplayer and signs of mission criticality will have to point to ROI on cost cutting or revenue generation.
Although generative AI is already consensus, we believe we’re still in the early innings of this platform shift. In the next five years, we expect that almost every software application will incorporate AI. As Yuval Noah Harrari stated, “the job market of 2050 might well be characterized by human-AI cooperation rather than competition.”
We’re super excited to partner with the next generational application layer companies in this market. If you’re building in this space, our team would love to hear from you! Please fill out this form and we will be in touch.