How startups can accelerate your generative AI journey

For many organisations, enterprise software from established software companies provides an effective route into generative AI. These platforms are tried and tested, and offer robust, versatile solutions that are well suited to many enterprise applications.

However, for organisations looking to further enhance their generative AI offerings, startups can provide novel and differentiated products that focus on innovative and more targeted use cases. When used in the right way, these startup products can accelerate an organisation’s generative AI journey, and provide a competitive edge against the wider market.

In this blog, we’ll take a closer look at the generative AI startup ecosystem, and discuss when and how startups may be a good option for your organisation.

The generative AI startup landscape

Since generative AI reached mainstream attention in late 2022, we’ve seen a rapid increase in the number of startups focusing on generative AI – with approximately 5 times the Venture Capital (VC) invested in 2023 into generative AI startups compared to 2022 according to analysis by Dealroom.

Broadly speaking, we see four types of generative AI startup that are relevant to commercial organisations:

• Model developers: A handful of well-funded startups led the way with some of the more powerful early Large Language Models (LLMs). Recently, many more startups have moved into this space and are developing additional LLMs and other generative AI models, often with a specific flavour. These include smaller, more efficient, more open, and higher performing models for more targeted use cases.
• Infrastructure and deployment: Productionising generative AI requires a lot more than just AI models, and for organisations looking to develop their own generative AI products, there are a range of startups that provide targeted solutions for different layers of a generative AI stack. These range from infrastructure optimisation to user monitoring, performance evaluation, security, privacy, and more.
• Horizontal applications: Many organisations face similar challenges regardless of industry, and there are novel startups targeting these horizontal applications using generative AI. These include knowledge management, enterprise search, internal communications, support services, and more.
• Vertical applications: Finally, there are also startups that apply generative AI to specific industry verticals, ranging from retail and supply chain management to financial services, tax, legal, defence, security, energy, and more.

From our analysis of Pitchbook and CB Insights data, the majority of existing funding (approximately 65%, or £18.6 billion) currently focuses on the first two bullets i.e. i) model development, and ii) infrastructure and deployment, although this data is skewed by a handful of larger funding rounds that reflect in part the significant cost of developing cutting edge generative AI models. Approximately 30% of funding, or £5.5 billion, is focused on horizontal cross-industry applications, whereas only 5%, or approximately £1 billion, is currently focusing on vertical, more industry-specific startups.

Whilst these figures are approximate, they provide some indication of the level of activity across different segments of the generative AI startup ecosystem.

Why partner with startups?

Startups provide an additional mechanism for organisations to integrate generative AI quickly into both internal and external activities.

Compared to internal product development, startups can provide ready-to-go products that do not require large upfront development and on-going maintenance costs, allowing organisations to test and deploy generative AI quicker, cheaper and with fewer specialist skills.

Compared to more established players, startups often provide more specialist and niche products which can allow organisations to explore a greater number of use cases using novel approaches, new technologies, and industry-specific features.

From our experience, startup products are also often well-designed with a clear focus on customer and user value, and increasingly provide a wide range of enterprise-grade features out-the-box, including security, privacy, integration and scaling. This is illustrated through many well-known startup products that are already used widely across industry.

What are some disadvantages and risks of generative AI startups?

Whilst startups provide targeted solutions for organisations, they are not always the best option.

Some of the more obvious risks stem from their size and funding. Startups are inherently more fragile, and can be acquired, pivot their product, or worst-case scenario, cease to exist altogether. This can render startup products unavailable, or only available via integration into other software platforms.

There are other risks too, such as a proliferation of fragmented software tools. If not careful, organisations can end up using a large number of separate products that solve individual tasks or functions, but require users to switch between many unconnected applications.

Moreover, there are other reasons why organisations may choose to build their own software solutions or purchase from larger vendors. Organisations may want to prioritise learning from experimentation and building, or have greater control over the final product. They may also want better integration with existing software, or greater visibility over the backend systems and algorithms.

How can organisations navigate these opportunities and risks?

This all points towards a complex set of decisions.

Whilst generative AI startups do provide additional opportunities for organisations, they also have disadvantages and risks that must be weighed against other options. When mapped against the large number of potential enterprise use cases for generative AI, this creates a lot of complexity.

At Deloitte, we navigate these challenges in a few ways. We have specific frameworks for managing build vs. buy decisions and for performing commercial due diligence, including for generative AI startups. We also have a specialist Startup Ventures team that focuses specifically on the startup ecosystem, including scanning, relationship management, testing and scaling. This team helps other teams both inside the firm and in client organisations to identify the right opportunities, and to navigate some of the challenges and idiosyncrasies of partnering with startups.

Startups are not always the right solution, but when managed effectively, they can be a great option for organisations. With the pace of development remaining high, they provide organisations with an additional route to test and integrate generative AI quickly into commercial use cases, and to explore more novel, challenging or specific use cases that would otherwise be difficult to tackle using other existing solutions.

If you would like to learn more about the range of startup-specific services that Deloitte offers, contact our startup team.