Daiki logo blue
Picture of Paolo Campigotto
Paolo Campigotto
Senior Data Scientist

Doing AI Right: How to Successfully Start Your First AI Project

Given the complexity of AI projects, it's important to have a strategic approach when embarking on your first AI project to ensure that you're set up for success

AI is a powerful tool that can provide invaluable help to businesses in basically any industry. However, it is not a silver bullet that can straightforwardly solve any problem. Furthermore, AI power and wide applicability come at a cost.

AI projects are indeed often more experimental and complex than regular IT or software development projects, with a higher chance of failure, not to mention the ethical risks of AI deployment. In this post, we will cover how to start an AI project from a data science perspective. 

Modern AI techniques are inductive reasoning systems that learn from data the information needed to solve the problem at hand. As data is the main source of knowledge, data collection is one of the natural first steps of any AI project. However, at Daiki, we do not recommend data collection as the absolute first step. Instead, we start an AI project by transferring knowledge of the problem domain and business context to the AI experts. 

Domain knowledge transfer and data collection

With a deep understanding of the problem domain, AI experts can contribute to identifying what data should be collected to feed the AI system in the later stages. This is a key point: AI experts should be involved already in the data collection phase to enable the collection of relevant data that is representative of the problem domain. 

As a matter of fact, poor-quality data resulting from an inappropriate collection process may frustrate even a technically flawless application of AI on said data. To put it differently, better data beats better algorithms. Nowadays, issues affecting the data collection process like selection bias are well-known, and the experience of people familiar with analyzing data scientifically certainly helps prevent these issues. 

Furthermore, ethical issues related to data collection, storage, and usage should be taken into account from the very beginning of the AI project. Data engineers, data scientists, and legal experts should collaborate to ensure data collection and usage in compliance with privacy laws and regulations, protecting sensitive information and maintaining data confidentiality when needed.

An attentive reader will have already noticed that several professional roles have been mentioned so far.  These include domain experts like business owners, project managers, or process engineers, AI experts like data/AI scientists, data engineers managing the infrastructure for data collection and storage, and legal experts for the legal issues related to data collection, storage, and accessibility. 

AI indeed demands a wide range of skills going beyond traditional software or data analytics projects. It is key to bring the right people together and assign them precise roles and responsibilities, as collaboration among people with different expertise and backgrounds often turns out to not be a piece of cake.

Do you really need AI to solve the problem?

Weird as it may seem, the second step in starting an AI project is to understand whether AI is actually the best approach to solve the problem at hand. 

Before developing a sophisticated AI algorithm, it is advisable to identify and test simple heuristic solutions for your problem. 

Start with simple heuristics

For example, if you are aiming at a recommendation system for your online shop, a user A that has visualized a product B may simply be recommended products visualized by other users that have also shown interest in B. Alternatively, products visualized by users with a demographic profile similar to that of B may be recommended. 

It is worthwhile to test these two simple heuristic solutions before embarking on the development of sophisticated AI algorithms. If heuristic solutions do not yield satisfactory results, you now have a well-founded reason to resort to the more sophisticated techniques. 

Not only that, the performance of heuristic solutions also offers a baseline for the success of AI techniques. The heuristic solutions should implement simple business logic, requiring a limited amount of data that is often available even before starting the AI project. 

When this is the case, Daiki suggests designing and deploying heuristic solutions as the very first step of the AI project. If you are convinced that AI can bring the most out of your company, often heuristic solutions will get you to a good starting point.

Define performance metrics

At the beginning of any AI project, it is also crucial to define the performance metrics measuring the success of the project. Getting back to our example, are our recommendations considered successful if the user just visualizes them and/or saves them in her cart, thus increasing the time spent in our online shop? Or are recommendations only successful if the user buys one of the recommended products?

Defining the performance metrics involves deciding what is tracked in the AI project. A good rule of thumb is to start tracking relevant metrics as early as possible. If something proves to be important in the future, it is advisable to have preemptively collected historical data about it.

Furthermore, defining the performance metrics is preparatory for setting the objective of the AI algorithm,  i.e.,  which metric(s) the AI algorithm will optimize. Based on the objective, different design choices for tailoring the AI solution to the problem can be made, e.g. the definition of a custom loss function.  

Ethical considerations

In addition to domain transfer knowledge, design of heuristic baselines, selection of suitable performance metrics, and deployment of a data collection process compliant with data privacy, the initialization of an AI project should also investigate the ethical implications and risks of the AI system under development. 

Many AI systems are not just purely technical systems, but rather socio-technical systems. Their outcomes may affect a wide range of stakeholders, including, of course, the final users. An example is provided by algorithms deciding the granting of loans to banks’ customers. 

The ethical use of AI requires the system’s outputs and decisions to be rational and fair, and requires transparent explanations supporting these decisions. Ethical issues are especially critical when sensitive data like personal information are used. It is therefore crucial to identify all the stakeholders affected by the AI system and the ethical risks of the system’s decisions. This is clearly not an issue that can be solved at a purely technical level. 

Once again, close collaboration between different professional figures including legal and ethical specialists, AI specialists, and business experts is required. Furthermore, with the coming AI regulations like the EU AI Act, ethical risks will pose not only reputational and financial risks but also legal risks for your company.

Daiki was built to support the proper kickoff and further development of AI projects, ensuring that AI is used in line with regulations, bringing together key stakeholders, and providing the best practices and methodologies to unlock the full AI potential for your company. 

Check out our whitepaper to learn how to develop a successful AI strategy

Related articles

Given the complexity of AI projects, it's important to have a strategic approach when embarking on your first AI project to ensure that you're set up for success
The United Nations General Assembly adopted a resolution advocating for trustworthy AI systems, offering a good basis for continued international policy on AI governance
Human-Computer Interaction (HCI) is a crucial lens that informs the technology we build, the way we build it, and whom to consider in the process. When we discuss responsible AI, then, we can see the crucial role of HCI in the development and adoption of AI.
Daiki Logo

Apply today

Daiki Logo

Book Demo Today

Daiki Logo

Partner with us

Daiki Logo

Join our waitlist for responsible AI development

Daiki Logo

Join our waitlist for responsible AI development

Daiki Logo

Join our waitlist for responsible AI development

Daiki Logo

Join waitlist for responsible AI development