‍

In our earlier article, we explored the usefulness of AI /GPT for company research and how we have incorporated those features into ListAlpha.

What we did not cover, was the second order question of: “well, if I can perform all of this analysis at runtime using GPT, what will happen to the people and companies that make a living from doing this manually?”

Today, we wanted to dive deeper into this topic - how will AI/GPT propagate through the PE software industry and what are the most vulnerable parts of the value chain?

‍

‍

1. The lesson from stock photography

‍

‍

Stock photography can offer an interesting insight into the impact of generative-AI on a given industry, largely because widespread adoption has arrived earlier.

‍Midjourney (a photo generating software) was released in July of 2022 (several months before GPT 3.5) and since has been on a geometrically upward path of self improvement to a point where now it is possible to create any photo or illustration, no matter how niche, strange or bizarre, within seconds.

This change happened so rapidly (in under 12 months), that most industry participants have fully internalised the corresponding implication:

• For Consumers, this is an extremely useful and powerful tool. We can now generate renderings on demand for any project, post or book within seconds and at a fraction of a cost.
• For Designers and photographers, this is very annoying / bordering on depressing, as a large part of their skillset has been made obsolete. They now need to move into adjacent fields or learn how to use new AI generative tools (e.g., Adobe FIrefly)
• Finally, there is one category of stakeholders who are extremely not entertained by this sudden technological breakthrough and that is stock photography companies and their shareholders.

Imagine you are Shutterstock and you are sitting on a multimillion collection of stock photo images, which are now nothing less but slightly inferior (and much more expensive) version of what can be generated on the fly using a GPU. The very core tenet of your value proposition was the depth of the catalog, which has now been trumped by an infinite imagination of generative models.

This is a tough place to be as a business. Many of the existing players will try to "product" their way out by offering generative AI tools, however that is a difficult thing to execute well.

‍

‍

2. So what went wrong?

‍

‍

Looking backwards, these businesses were built on a very large collection of images, which at their core, were outputs of low skilled cognitive work.

On its own, a few stock images don't represent any significant value. However when aggregated together into a searchable database, a lot of value can be derived from this collection for consumers. The likes of Shutterstock would pay $15 per hour to a Bulgarian graphic designer (who had the skills, but lacked the distribution platform) and then resell their work hundreds of times to a global audience of consumers.

This was a great business model that gave rise to many companies (Vecteezy, Freepik, Storyblocks, PixaBay, etc.) until now. The problem is that when the marginal cost of performing the low value cognitive work goes from [x] to zero, the value of a collection of that work collapses as well.

We believe that this dynamic will be a recurring pattern in the coming years as generative-AI's cognitive abilities expand from image generation to every single other industry. Video footage (Motion array), sound collections (Epidemic Sound - a multibillion dollar EQT deal) are likely to be next.

The collateral damage, and the amount of equity value at risk here, is truly astonishing.

‍

‍

3. Impact on Existing Software Players

‍

‍

Bringing this concept to private markets, you may ask yourself:

- “Which businesses depends on a large database of commodity information?" ...or, "which business relies on a large number of low skilled offshore workers to produce insights for them?"

Well, you are literally describing every one of the logos below:

If an investor is able to generate a great description of a target business entirely out of GPTs memory (for free), why should they pay a $50k subscription fee for Pitchbook? If GPT is able to thoroughly explain the impact of cloud computing in the Retail Industry, why should one need that Gartner subscription or that report from Forrester?

Yes, the company platforms offer more (e.g, financials, cap table, previous rounds, etc.), however those data points are not worth the c.$50k per year access fee on their own. The main value in a company database is the ability to get to a concise company profile, quickly and consistently (for every target). The cost of offering that service, has (or very soon will) gone to zero, which marks the question - so what will be left of these companies?

‍

‍

4. Who will survive?

‍

The impact on a given PE software player will depend on (a) the starting position of the business, and (b) the amount of action / repositioning that they will take in the face of this coming threat. Our prediction is that in light of the GPT threat the following companies will do better:

• Data businesses that rely relatively less on low skilled workers for their content creation
• Companies that focus on harder / more sophisticated analysis. (E.g., the quality of work coming out of Gartner was always questionable, even in the pre-GPT era. In contrast, highly quantitative equity research content will be harder for AI's to reproduce)
• Companies built on proprietary data sets which are not available online (on the contrary -  importing Company’s House data through API is no longer a moat)
• Using other people's content (e.g., AlphaSense which relies on bank's equity research reports) rather than generating your own
• Being a lower cost players such as Crunchbase (at a $5 price point) you are a lot more protected as a business from revenue erosion than a super premium database like BoardEx or Pitchbook

‍

‍

5. Implications for future software design

‍

‍

The impact of these trends will be profound, as a result we at ListAlpha are taking a significantly different approach to our architecture in order to future proof it for the trends ahead:

• Build out heavy user side logic → when all of the information on target companies and people is readily available through GPTs, it is the user side interfaces and customisations that allow for you to differentiate. Centralised databases with public information are a thing of the past
• Rendering judgement and assessment in realtime is key. I personally believe that we will see to Co-pilotization of many industries, private equity included.
• Integrate with as many APIs and data sources as possible. The AI models will be able to parse them easily, hence there no constraints
• The moat is in UI, user data and how you use the user data to train the model and judgement

‍

‍