How to survive OpenAI as a Business

Have you ever spent months building a product only to find that a big player like OpenAI suddenly offers an off-the-shelf solution that does the same thing? Or perhaps you've watched as technological advances lower the barriers to entry, making it easier for new competitors to flood the market you thought you had cornered.

In today's fast-paced technology AI landscape, being disrupted by giants or new entrants is a looming threat for many businesses. The rapid evolution of AI and low-code platforms means that what was once a unique offering can quickly become commonplace. Companies that fail to anticipate these changes could find their hard work overshadowed overnight.

This scenario highlights the urgent need to develop strategies to protect your business from such disruption. It's not just about creating something innovative; it's about building resilience against the tides of technological change and competitive pressure.

In this blog, we'll explore key strategies that SaaS companies can implement to gain a competitive edge and survive the long game. We'll look at how seamless integrations, strong networks, customer data, technological innovation and financial strength can help you future-proof your business in an ever-evolving marketplace.

Outline:

1. Changing AI market dynamics.
   • The rise of contextual memory
   • The Assistant API and Custom GPT
   • Autonomous Control and Computer Use (Anthropic)
2. Strategies to gain competitive advantage
   • 1 Seamless Integration: Building robust ecosystems
   • 2 Network Effects: Adding value through community
   • 3 Data Advantage: Leveraging customer insights
   • 4 Technological Superiority: Leading through innovation
   • 5 Financial Strength: Scale with capital
   • Entry and exit barriers: Securing market position
   • Changing market dynamics with AI and low-code platforms
3. Wrap-up.

But before we go any further, let's take a step back and look at how the market dynamics in AI are changing and disrupting entire industries (which admittedly came out of nowhere).

I can think of just a few recent AI examples that I would like to share with you:

#1 The increase in contextual memory.

If you have never heard the term "context memory" before, in a nutshell it is the amount of text you can pass in a conversation to an LLM like OpenAI GPT.

For example, if you ask OpenAI to summarise some text, the context window is, in very simplified terms, the number of words in your text divided by three, which is called a "token" (again, very simplified).

In the "early" days, the context length was very limited, only a few thousand tokens (I remember 4096 tokens from OpenAI). For some scenarios this was quite sufficient, but it was certainly limiting.

And where there is a limit, there is always an opportunity for companies that try to fill the gap and make more possible than what the so-called foundation models like OpenAI GPT could offer with the limited token size.

Example:

Matching CVs with jobs was a classic example of this scenario.

With 4096 tokens, you could squeeze a CV into this context window, but not the job description. (Nowadays, it's a no-brainer.) So companies started looking for a solution to fill this gap by automating and breaking up the long text into chunks and matching or embedding those chunks, etc. This worked pretty well because it was an improvement over traditional matching algorithms.

But then OpenAI came along with news like this - an increase in the context window. And not just OpenAI, of course; they all followed.

https://venturebeat.com/ai/openai-launches-experimental-gpt-4o-long-output-model-with-16x-token-capacity/

So what happened:

If these companies relied solely on splitting the context window, most of their "business model", if we can call it that, evaporated faster than they had anticipated.

The user could now simply stuff a CV with 20 years' experience with any job description text from a website and get a comprehensive report of hits and misses.

Game over.

#2 The Assistant API and custom GPTs

LLMs are great for getting answers to general knowledge that the model has been trained on, but in the real world we don't just want answers to general knowledge, we want answers to our very specific situations for our custom data, whether for personal use cases or for business.

Even after the context window was expanded, there were and still are limits to how many documents an LLM can read at once.

So many companies (until today) have taken advantage of this high demand for using AI to get answers and apply AI to their own custom data, and have started to provide the so-called RAG (Retrieval-Augmented Generation) pipelines for almost everything.

Chat with your PDF
Chat with your documents
Chat with your website...
And so on.

The use cases may be completely different, but the underlying technology is more or less the same: the large amount of data that needs to be used to answer questions (whether it's text, PDF or even media) is embedded and stored in a vector database for quick retrieval at a later time and answered by the LLM. This technology remains the holy grail of enterprise AI.

And then OpenAI came up with this news at the 2023 developer conference, reported here and elsewhere, and dropped another bombshell with Custom GPT and the Assistant API:

https://www.wired.com/story/how-to-use-chatgpt-create-custom-gpt-openai/

If you have heard of the Assistant API and Custom GPT - in a nutshell, both (the Assistant for developers and the Custom GPT for non-technical users) can tailor their own GPT along with custom data and custom prompts, and deploy this new custom brain along with their custom prompt and data in a few clicks.

Both did exactly what many companies have built their business cases around and sold to their customers for a few dollars.

Now anyone can build their own custom GPT or use the Assistant API with some wrapper code to build the same with more control over embedding, etc.

Example:

For example, if you are a developer and want to get specific answers to a bunch of documents related to a product you are developing, you could simply upload those documents into a Custom GPT and start asking questions in a Custom GPT that you could also share with your team.

Of course, the applications of Custom GPTs go far beyond this; if you are creative, the sky is the limit. Today, there are tons of products from other vendors that try to solve the same problem, most notably Notebook LM.

The impact on the industry

With these advances, companies that did more than just stitch together a few frameworks to build such a RAG pipeline for their customers and had a strong UX or tailored their product for very specific use cases (like legal) were not yet driven out of business because they had more to offer than the off-the-shelf solution from OpenAI & Co.

However, competition has increased significantly due to the lowered barriers to entry for competitors.

#3 Autonomous Control and Computer Use (Anthropic)

Computer Use was another bombshell just released, but this time from Anthropic, to address another very general problem: controlling user interfaces in the OS or a browser using AI by taking screenshots and letting the LLM navigate towards a given goal. That's what AI agents do.

You can read more about computer use in my previous article.

Computer Use is still in its infancy, but it is certainly not going to stop at its current stage - it is only a matter of time before this technology evolves and makes not only companies, but also regular jobs obsolete.

Does all this sound very gloomy? It does not have to be if you stay informed and take the right countermeasures.

In the next section, I will cover a few strategies that help companies build a sustainable competitive advantage to survive against industry giants and fierce competition.

#1 Seamless integration: Building Robust Ecosystems

A significant advantage for SaaS companies is their ability to integrate seamlessly with a wide range of other applications and platforms. This deep integration creates a "sticky" ecosystem that makes it difficult for customers to switch to something else.

For example, Make.com , a leading automation platform, connects effortlessly with numerous third-party applications. By integrating with marketing tools, customer support systems, and analytics services, Salesforce becomes an integral part of a company's daily operations that is difficult to replace.

Transferring data and reconfiguring systems can be both challenging and costly, resulting in significant switching costs.

As a result, customers are more likely to stick with a service that's deeply embedded in their workflow, even if they encounter occasional problems.

#2 Network effects: Adding value through community

SaaS providers also benefit from network effects by building communities of partners, affiliates and users. The more people who use the platform, the more valuable it becomes to everyone involved.

Take Facebook, Microsoft Teams or WhatsApp for example. Their rapid growth is partly due to their large user bases and integrations. As more companies adopt Teams for their communication and collaboration needs, the value of the platform increases, encouraging even more companies to join.

These expanding networks not only make the platform more useful, but also create a strong competitive barrier**.

New entrants find it difficult to build a similar network without significant time and investment.

#3 Data advantage: Leveraging customer insights

In today's data-driven world, data is incredibly valuable. SaaS companies that collect and analyse large amounts of customer data can provide personalised experiences and insights that are hard for others to match.

For example, DeepL, an AI-powered translation SaaS, uses user data (keeping privacy at the side for now) to continuously train its models to improve translation, outpacing any new entrants who solely use the plain foundation models from OpenAI & Co.

Over time, the unique data these companies collect becomes an important competitive advantage.

Competitors can't easily replicate the insights that come from years of data collection and model refinement.

#4 Technological superiority: Leading with innovation

SaaS companies that invest heavily in technology and/or user experience can leapfrog competitors with superior products and services.

Ongoing investment in research and development leads to continuous innovation that sets a company apart.

However, it's important to remember that technological advantages can fade as competitors catch up, so constant innovation is key.

#5 Financial strength: Scale with capital

Having substantial financial resources allows SaaS companies to invest in cutting-edge technology and infrastructure that may be out of reach for smaller competitors.

Companies such as Google and Amazon are pouring billions into research and cloud infrastructure. Their financial muscle allows them to develop and offer services on a scale that smaller companies can't easily match.

This financial advantage creates high barriers to entry for new competitors, as the initial investment required to compete effectively is enormous**.

Entry and exit barriers: Securing market position

High entry and exit barriers help incumbent SaaS providers fend off new competition and reduce customer churn. These barriers can be technological, financial, or related to customer data and integration complexity.

Take enterprise resource planning (ERP) systems such as SAP as an example. These systems require significant time and resources to implement. Once they're in place, the complexity and cost of switching providers discourages customers from leaving.

While having a first mover advantage is advantageous, maintaining it requires continuous innovation and the delivery of consistent value.

Companies need to ensure that their barriers are sustainable and not just based on being first to market.

Changing market dynamics: breaking down barriers with AI and low-code platforms

Traditionally, entering the software industry has been difficult due to the high cost and expertise required for software development. However, advances in AI and the rise of low-code and no-code platforms have significantly lowered these barriers.

These technologies enable the rapid development of Minimum Viable Products (MVPs) with minimal coding skills, allowing new players to compete more effectively.

Platforms such as Replit, Make.com, OutSystems and many others offer low-code solutions that help companies develop applications quickly, reducing time and costs.

Many have even gone a step further, using AI to automate the stitching together of UI blocks using natural language, so that the user's only task is to specify the product features and UI.

In addition, AI code generation tools like GitHub Copilot are making it easier than ever for citizen developers to build software faster.

While quality can still be an issue for startups, these advances are greatly reducing the barriers to entry that once required heavy investments in software and developer expertise.

Conclusion

In the age of AI and low-code platforms, launching a SaaS startup has become easier, attracting many innovators to build and launch new applications quickly. But this ease of entry isn't just for innovators - it also opens the door to more competitors.

That's why it's critical to evaluate how easy it is for others to replicate your offering and how difficult it is for customers to change providers**.
By leveraging seamless integrations, building strong networks, leveraging customer data, leading with technology and having the financial strength to invest in innovation, SaaS companies can maintain a competitive edge in this dynamic landscape.

Building a sustainable business requires thinking beyond the initial excitement of creating a cool app and focusing on long-term strategies that ensure lasting success.