Our Thesis on AI

* The cover image for this blog post was kindly created by ChatGPT. We apologize for the spelling mistake on its behalf, but we weren't able to correct it at this time...

Much has been written about the breakthroughs in Artificial Intelligence caused by the launch of the Generative AI revolution by OpenAI. The profound implications for society today and in coming decades are impossible to ignore. This is an “iPhone moment for mobile”, “the Web moment for the Internet” type of event.

At 2048 Ventures we are a thesis-driven pre-seed and seed firm. Our style is to spend a lot of time thinking about the future and form both macro (space focused) and micro (opportunity focused) ideas. By being thesis driven, we are better equipped to sift through the tens of thousands of startup ideas that land in our queue every year.

Similarly with AI, we have not been rushing to invest into the infrastructure layer or a slew of copy cat generative AI startups. Instead we spent months thinking about what long term defensible AI-powered business are and formulating a clear lens on what we are excited about.

We landed on two focus areas which we believe have the highest potential.

1. Unique Data Capture through Existing Workflow

Data is the new oil. This has been said many, many times, and is specifically true now in the age of AI. Purely algorithmic AI, without a proprietary data set, has no moat and is not defensible long term. The moat comes from proprietary data that then can be fed into proprietary ML models to create a competitive advantage.

So how does one obtain proprietary data?

Our answer is through creating a valuable business workflow which generates proprietary data as an exhaust of performing the workflow. These types of businesses aren’t thought of typically as an AI business, but in fact have the best advantage. Let's look at concrete examples.

Soona, a 2048 Ventures portfolio company, is the largest provider of fast casual photo and video for e-commerce brands. The bread and butter of soona’s business is virtual shoots where brands send soona their products and get back fantastic looking photo and video assets. Soona’s workflow has caused the company to accumulate millions of proprietary photos and videos. Now, soona has launched an AI offering where they leverage their data set and learnings to accelerate the creation of branded assets and make them even cheaper.

Another example is Impilo.health, which is quickly becoming the largest provider of logistics and data APIs for remote patient monitoring. Impilo allows health systems to procure and provision glucometers, heart monitors, etc. to be used by patients in their homes. Next, Impilo streams the data from the patient devices into the Impilo cloud and makes it available to healthcare providers in real time. As of today, the company has no AI/ML offering, but it is rapidly accumulating exceptionally valuable datasets which will lend itself well to algorithmic analysis.

A very different and exciting example of the interplay between the workflow, proprietary data and ML can be increasingly seen in biotech companies. This is the category we are incredibly bullish on and allocate about 30% of our Fund II. Many biotech problems, whether a drug target discovery or a formulation, are essentially problems of combinatorial search through a large set of genetic information. ML models take data from the lab and narrow it down to a set of possible solutions. That in turn gets verified in the lab and fed back into the ML model and so on.

The feedback loops between ML and lab data are awesome because they cause the generation of progressively more valuable proprietary data and proprietary models. Examples of companies from our portfolio include Coding.bio - a platform for CAR-T, Persist.ai - a platform for slow release drug formulation and Caravel.bio - a platform for enzymes for sustainable catalysis.

2. Reinvention of Workflow through AI

While the first category uses existing workflows to generate proprietary data, the second category we are excited about uses AI to fully re-imagine the workflows.

If we take a step back and review the last few months of generative AI, it’s clear that what initially seemed like an opportunity has been captured by incumbents. For example, Canva and Adobe launched their AI offerings and shutdown the opportunity for startups that were narrowly focused on algorithmic image generation without owning the distribution. Broadly, incumbents that have distribution can push out incremental AI innovation into their existing workflows, blocking the opportunity for startups.

This means that for a startup to win, the focus needs to be on winning the distribution game and this can only happen by fully disrupting and fully re-imagining the existing workflows. Specifically, this means that instead of plugging AI into an existing UX, the startup must re-invent and make the entire UX better.

This is exactly what one of our recent investments Ripple is doing for e-commerce marketing. Ripple is on a mission to re-imagine e-commerce marketing workflows by replacing agencies and complex back and forth with AI. Instead of using drafts, templates, image libraries, etc., Ripple starts by analyzing branded content and then allows e-commerce brands to quickly create fantastic looking emails just through a few prompts. Ripple’s early customers love the speed and the ease of use, as well as the fact that Ripple seamlessly integrates with Klaviyo. Ripples vision is to bring the same ease of creation and automation to all other channels of e-commerce marketing.

Another portfolio company Psyrin develops digital biomarkers for serious mental illness - they're starting with integration in assessment workflows, aiding clinical decision making. But as they accumulate proprietary data, they will move to reinvent the intake and monitoring workflow.

In the past few months, the 2048 Ventures team met with many startups working on new AI tools for lawyers and accountants. Many of these startups have been trying to wedge on top of existing tools and overlay AI on top of existing workflows. We are somewhat skeptical about this approach especially in services industries like legal and accounting where vendors thrive and make more money on information asymmetry and inefficiency.

We feel that re-inventing workflows is essential in order to disrupt and to wedge into these legacy situations. We think that, perhaps, a fully vertically integrated approach has a better chance. We imagine a next generation legal firm that is mostly powered by AI, but still has a human in the loop and a human as a UX - both to make sure mistakes are not made and, more importantly, ensure trust and piece of mind are preserved. What remains to be seen is if such an approach would yield a truly venture scale outcome.

To sum up, we are incredibly bullish and optimistic about AI opportunities for years to come.

If you are building a startup that is focused on workflows with strong data capture / system of record or a startup re-inventing existing workflows through AI please reach out at 2048.vc/pitch.

‍