If you spend a fortune on Generative AI, you miss the mark

Do you need billions to implement Generative AI in a way that adds value to your business? No, not if you ask us. On the contrary, you’ll get very far with just a couple of million and get more than started for less than half a million. Though the market is booming with articles about companies that have spent billions implementing Generative AI, we believe that both your money and your employees' time are better spent modestly and beautifully than grandiosely and expensively. Actually, most of the work consists in ensuring technical integration in terms of security and data protection, educating your employees, and setting frameworks for the generation of ideas on the use of Generative AI across the company.

The AI hype wave we are currently in the middle of has been characterised by two things over the past several months:

• A fundamental realisation among business leaders and employees that Generative AI will fundamentally change our daily lives and work processes.
• A dominating discussion about where exactly Generative AI will make a difference and change reality.

In addition, there is a high demand for use cases and examples of strategic anchoring. A demand that we have observed is typically met with the following two incorrect responses from major providers in the market:

1. You need to structure and clean your data so that you can fine-tune one of the major Generative AI models on the market – because (insert the name of the provider yourself) has done so with proclaimed success.
2. Here you have an orchestrator/a switchboard to manage all your many use cases and switch between the models you use.

Does it sound technical? It does because it is a fundamental attempt to solve a strategic deficit by means of engineering: We do what we can technically because we don’t know what is needed.

A problem that stems from the fact that the supplier market for Generative AI solutions naturally has been and still is affected by exactly the same thing as their customers – they have few specific use cases apart from open chatbots and none that they have seen scaled to hundreds or thousands of employees.

Which means they don’t have a clear idea about what the world will look like with Generative AI in just six months.

So, what do you do to avoid being caught with your pants down?

AI technology is evolving at lightning speed, and that’s why you can easily end up spending unnecessary time and money if you jump uncritically into the hype wave.

Let's paint a specific picture.

If you choose to go with the solution of fine-tuning one of the major Generative AI models on the market – such as a GPT-4 model from OpenAI – you may well expect to spend at least 3 million USD. That is the amount OpenAI itself requires you to be willing to invest at a minimum before they will even discuss fine-tuning with you.

Then six months pass with you using that solution, and a new, better, or faster model appears on the market. What do you do now?

Because wouldn't it be rather extreme to fine-tune again, isn't it?

Fine-tuning a model makes it more proprietary and specific to your company, thereby ensuring you a stronger Generative AI for the specific needs of your company. But here follows the problem – what are your needs?

That's where you need to start. And you do that by focusing on ideas for implementation and use rather than spending millions on technical features.

So, you should turn your attention to your employees when you start working with and implementing AI technology. They have already run ahead in many companies, which means they have an overview of the landscape and often have made valuable experiences with what the challenges are for the company.

In addition, employees should be educated to understand what the technology can do and be encouraged to use it where they can see it makes sense. And following that, the management’s task is to establish a good framework for developing and exchanging ideas.

You can, for example, do this by establishing a track with workshops where Generative AI is put into play as an automation opportunity on unstructured data. Quite recognisable in style with a classic change project, where the crucial difference is that the decision about the change has been made for you.

Concurrently, it’s important that you ...

• Discuss IT security
• Consider methods of application and automation designs
• Consider risk profiles for the use of the technology in different departments and business areas.

Only then does your company get to the AI technical considerations that are crucial for you to make the decisions that fit your needs. And here the answers are much simpler than they appear.

Most solutions are easy and inexpensive to access in the cloud and consumption-based, so the work of integrating them into your reality isn’t complex, but at most a bit cumbersome. A bit like adjusting the car using the infotainment system rather than starting to tinker with the engine.

This also means that we unequivocally believe that you can get started with Generative AI without having to spend a fortune – neither in resources nor employee hours.