WHAT TYPE OF ARTIFICIAL INTELLIGENCE DOES MY COMPANY NEED? HOW CAN I CHOOSE A GOOD AI PROVIDER?

Digital transformation should be top of the agenda for any company working towards Industry 4.0 and there is no doubt that Artificial Intelligence plays an essential role in this metamorphosis.

There is currently a wide array of Artificial Intelligence solutions and providers. However, it is very important to remember that not all companies need either the same type of Artificial Intelligence or the same type of provider.

Through this post, we hope to provide a practical guide for companies that have decided to apply Artificial Intelligence but do not know which type of solution or provider to pick.

When we are not sure about something, the first thing we usually do is look for information on it on Google and in this case, obviously, the top search results will be Big Tech sites (Google, Amazon, Microsoft, etc.). However, Big Tech’s AI solutions are not always the most appropriate for all companies. If you dig a little deeper into your search, another type of provider will emerge such as young start-ups, consultancies, etc.

This creates even more uncertainty, so we must reset and ask some basic questions:

What is my access to the Internet?

Many companies, above all in the industrial sector, have major restrictions on internet access. For example, it is possible to work on an assembly line and not have access to the web.

What type of company am I?

Am I a new company where my product/service is based on AI? Am I an existing company to which I wish to apply AI to strengthen my product/service? Am I an SME that does not have the same resources as a major company, although I wish to implant Artificial Intelligence?

~

What about my data privacy?

What are my company’s restrictions in terms of extracting data and passing it on to third parties for different uses? 

After asking these basic questions, it is important to define three key aspects for your AI solution.

1. WHAT TYPE OF ARTIFICIAL INTELLIGENCE DO I NEED?

PREFABRICATED/PRETRAINED ARTIFICIAL INTELLIGENCE

These are AI solutions for specific tasks, usually generic and recurring. For example, Artificial Intelligence for face and voice recognition, chatbots, automatic translators, etc. This involves training AI, a specific model that can be sold to many different customers because they have the same need. We might say that this is like selling a boxed ready-to-use product.

ARTIFICIAL INTELLIGENCE FOR SPECIFIC TASKS

Artificial Intelligence solutions developed by suppliers who train specific AI for your task.

2. WHAT TYPES OF LICENCES ARE THERE?

PERPETUAL

The customer buys the right to use the Artificial Intelligence indefinitely (no expiry date). It is purchased in a single payment.

SUBSCRIPTION

The customer purchases the right to use the AI for a limited time at a much lower price than the perpetual licence. The periods are usually monthly, quarterly, annual, etc. with an option to renew the subscription automatically.

PAY AS YOU GO

The customer pays for their real use of the AI, usually based on the number of predictions made by the Artificial Intelligence or on the computation and storage resources required by the AI.

3. WHICH PROVIDER MODEL WOULD WORK FOR ME?

CLOUD BASED

Solutions mainly developed by Big Tech. They are usually pay-as-you-go. What is the usual disadvantage of these Artificial Intelligence solutions?

The task for which the AI has been trained does not match my needs and therefore does not perform as required. Data privacy might be another disadvantage of Cloud Based systems. In other words, this AI must extract your company’s data from Big Tech servers and in many companies and organisations, the privacy policy makes this impossible.

ON PREMISES

Artificial Intelligence that runs directly in our system. This mainly helps us get round the privacy issues of the Cloud Based Solutions because you run the AI in your company and so there is no out-going data.

The disadvantage of this type of Artificial Intelligence solution for companies can be higher implementation costs and that the company’s systems require more adaptation, so the return is usually not seen until later, in the long term.

PRE-BUILT SOLUTIONS

Pre-trained Artificial Intelligence for very specific use cases. These are usually developed by pre-existing suppliers (such as artificial vision or machine translation) that provide their know-how in a generic Artificial Intelligence product that is pre-trained.

EMBEDDED SOLUTIONS

These are final products that have built-in Artificial Intelligence.

HYBRIDS

Artificial Intelligence solutions for companies that are not sure about their business model and decide on a hybrid model. One example would be consultancies that offer packs of consultant hours or supplier tools and platforms that cover many types of use cases.

At this point of knowledge of the heterogeneity of the Artificial Intelligence market, it is possible that you have had to rule out what is usually the cheapest option, using Big Tech, because its generic solutions do not adapt correctly to the specific task that your company wishes to automate. There are so many specific tasks that (for the time being), it is impossible to cover them entirely with a single type of Artificial Intelligence.

In this case, if you know that you need Artificial Intelligence trained specifically for your task, companies such as Neuraptic and its ENAIA Artificial Intelligence and Machine Learning platform are your solution.