Should You Use A No-Code AI Platform? Limits and Opportunities

As you can see, some steps of a “traditional” AI project can be automated or made easy using drag and drop tools. From this perspective, I would consider no-code AI platforms as a time-efficient way to speed up prototypes and demo development.

New skillsIt is safe to assume that the rise of no-code AI platforms will also create new skills expectations. In the near future, I would not be surprised if a product manager has to be familiar with at least one no-code AI tool and be knowledgeable about dataset management. I expect to see more and more online training related to these tools.

A specific market

No-code AI is still a growing market. Most of the players seem to have positioned themselves above all on typologies of technologies (NLP, Computer vision, etc.) or specific use cases (CRM Management, …).

In the near future, I expect to see complete tools that make it possible to cover almost all uses, thus avoiding the need to invest in multiple tools and to capitalize on the knowledge.

The industry is composed of both startups and large tech firms developing their own tools. For obvious reasons, it seems that startups are focusing on specific use cases rather than offering several options.

Lock-in strategy & Business model

The most interesting aspect of this industry is how large tech firms are trying to attract new customers to increase their user reach while using a lock-in strategy.

Lock-in strategy:A strategy in which the customer is so dependent on a vendor for products and services that the customer cannot move to another vendor without substantial switching costs, real and/or perceived.

I often wonder if no-code AI startups can survive over the long term if they are not specialized. Indeed, the advantage of large tech firms is the ability to provide customers with a no-brainer approach to staying on that vendor’s platform and roadmap.

Ideally, large tech firms (Google & Microsoft) want companies to be able to use their software development tools and a wider ecosystem of services related to data management.

While some of these solutions are free or based on a subscription model, they might require the intervention of consultants and developers to train users and perform the cloud back end connection engineering.

My opinion after implementing a no-code AI solution

For a couple of months, we have decided to test the effectiveness of a no-code AI platform. In my opinion, the efficiency and usefulness of No-code AI are not a myth.

The biggest benefit for us was when a data scientist could help our Marketing department by evaluating their incomplete ideas thanks to the creation of a quick PoC using no-code tools. Indeed, he/she can advise them to make a no-code PoC and come back to it when the needs are stabilized.

However, some concessions have to be made. Indeed, if you want to produce quickly and without development, you have to be able to lower your expectations.

The success of a project based on Machine Learning through a no-code solution will greatly depend on your functional needs, and when the needs are very specific, it is then necessary to find the right balance between speed of implementation and expectations related to functionalities.

Each solution has a boundary that is intrinsic to the design of the tool. Indeed, these tools are based on models, either the editor has chosen to propose simple models that will be easy to understand and use but in return will lack flexibility because one can only develop within the framework of the model.

On the contrary, other platforms have chosen more elaborate models that allow flexibility comparable to the development of applications by coding. On the other hand, the learning curve and the skills required will be much higher.

I strongly recommend choosing your tool according to the balance you need between ease of use and flexibility . In some cases, it is important to remember that once you have developed an application on a platform, you are linked to that platform for as long as the application is running. In the context of a PoC this is not a problem, but in the context of an application that is expected to last, things can be different.

Indeed, you have to make sure that the required scalability of your project can be achieved using a no-code AI platform. I believe that it is impossible to have a scalable solution using no-code AI platforms for “complex” use cases.

Furthermore, the contractual relationship ( data ownership ) must be carefully examined in terms of cost and reversibility.

The other key element is maintainability . I recommend you determine right from the start if your goal is simply to test the relevance of an idea or to build a long-lasting application. In the first case, it is better to quickly create a disposable PoC using no-code AI solutions if possible. Otherwise, I recommend you use the traditional ML approach and build stabilized and maintainable version.

If your no-code AI supplier can guarantee you scalability (possible in some use cases), I would recommend you to assess the total cost of licensing at scale to ensure long-term maintainability.

The worst thing that can happen is to create a PoC using a “quick and dirty” approach and then go to production trying to scale this same PoC.

LIMITS

In my opinion, the current limits of most (not all) no-code AI platforms are:

The other drawback is related to data. Indeed, you may be familiar with data bias . Depending on the use case, these models can be built and exported by potentially everyone. As such, there is an increased danger of generating biased algorithms.

Obviously, Machine Learning engineers can also create biased solutions….

Shadow IT: a term that refers to Information Technology (IT) applications and infrastructure that are managed and utilized without the knowledge of the enterprise’s IT department.

Data scientists could end up spending as much time fixing their colleagues’ work as they could on their own tasks.

The success of an ML project highly depends on your ability to collect, manage, and maintain a dataset. However, this is usually the work of a data scientist and I am not sure if a “citizen developer” can perform these tasks.

I could have mentioned other potential issues such as deployment, safety, integration with legacy systems, etc. However, I still believe in the usefulness of no-code AI platforms. The success of your project using no-code AI will highly depend on your use case and level of maturity when it comes to data management.

We are merely at the beginning of this trend and I am quite confident that a growing number of companies (especially SMEs) will be tempted to leverage these tools.