What will it take to build a Generative AI winner?

By Shamillah BankiyaVirginia Pozzato & Zoe Qin

As explored in the first part of our Europe x Generative AI series, we have identified many exciting European companies making breakthroughs in core opportunity spaces across this stack, from B2B applications to LLM Infrastructure & Ops and Foundation Models.

In this piece we’re sharing our view on what we believe it will take to build a winner in this space.

We’ll focus on two major opportunity spaces in the Generative AI stack: B2B applications and LLM Infrastructure & Ops. We’ve honed in on these categories because they are where we see the greatest opportunities to build disruptive businesses.

  • Opportunity 1: B2B applications

Startups in this space are embedding generative AI into their products, serving either a business user or the consumer. These companies present exciting opportunities because they readily tap into large available markets, and can prove their business case to users extremely fast.

The opportunity is large, but it is important to remember that their offering is open to challenge from competitors and incumbents.

Therefore, in order to build an enduring company that stands out in a crowded market, we believe that startups will need to offer:

i) A ‘data moat’
Top companies will need to acquire and maintain a large, unique proprietary data set.

This is vital, because without a data moat incumbents and new entrants can copy any novel features startups release. For example, startups that merely add a ChatGPT API on top of their product will quickly become commoditised.

ii) A B2B software product that can revolutionise workflows

Category winners’ products will fundamentally reimagine the way we work. These companies will offer software that reimagines workflows. They may take time to materialise, but have huge potential.

For example, startups are currently working on using Generative AI to build agents that observe and replicate day-to-day workflows. A clear example here would be reimagining customer relationship management (CRM) systems. A new-age CRM could automatically action next steps after calls, create emails, and even make suggestions of highly relevant opportunities to consider.

Types of winners we see emerging in the application layer…

Our view is that if startups have both a data moat and reimagine workflows for end users, then they are well-placed to win categories globally. This type of Generative AI-based product is strongly differentiated, with the potential to have real impact on the way we live and work in the coming decades.

Such startups really stand out, but we don’t believe they present the only model for global success.

For example, if a company’s product reimagines workflows, but remains reliant on publicly available data models, that company has a good chance at achieving category domination through execution, speed and first-mover advantage. Similarly, if a company leverages a large unique data set to maintain legacy workflows, they could create an IP moat that is also very valuable.

  • Opportunity 2: LLM Infra & Ops

Startups in LLM Infra & Ops are helping firms to build new Generative AI-driven products and features aimed primarily at builders and developers.

Through surveying the market, current demands and fast-moving trends, we think that future LLM Ops category leaders will offer three crucial elements:

i) They will speed up AI software development and delivery

To win in this space, companies will need to demonstrably speed up Generative AI development, and generally make the process of incorporating the technology easier and faster for customers.

ii) They will cut running costs and deliver hard ROI for clients

Winning LLM Ops startups will offer inference optimisation and immediately cut the cost of building Generative AI-driven features for customers. Examples include serverless GPUs.

iii) They will be easy to use and will offer a sense of community

We believe that successful Generative AI LLM Ops companies will abstract a lot of LLM complexity for their customers, making it easier for companies to develop with models. As is typical when creating categories and cultivating user habits, top LLM Ops companies need to be able to create a loyal audience who can both contribute and advocate for their product. This is crucial to building long-term defensiveness, and also creates a community of customers that actively learn from one another — another key element when it comes to gaining trust and traction in the market.

We hope that these insights into what we believe it will take to build a Generative AI winner are helpful for founders navigating a rapidly developing space.

As this exciting market evolves further, we will continue to update and share our thesis — and if you’re a founder innovating in this space with any questions or information to share, please do get in touch. We appreciate the complexity of the challenges faced by even the most exceptional teams behind potential Generative AI rocketships, and we are actively looking to support and jump into this adventure with you.

Please get in touch at shamillah@dawncapital.com, zoe@dawncapital.com, and virginia@dawncapital.com.

Latest from Dawn

26-04-2021

AI 50

Dawn Dawn
Dawn Dawn

to our newsletter

Stay in touch with us