Skip to main content

Successful AI projects: Best Practices for planning and implementation

What should companies do to reap the benefits of artificial intelligence (AI) as quickly as possible? Expectations are often huge. Some companies want to use AI to give their business model superpowers overnight. In other businesses, managers don’t feel digitally mature enough and unnecessarily delay their start with AI. Both attitudes hinder the sustainable success of artificial intelligence. 

We explain how you can prepare an AI project for optimal success and how to make sure your investment pays off.

Misconceptions slow down digital progress  

Especially in small and medium-sized businesses, various misconceptions persist that discourage decision-makers from tackling AI projects and slow down the expansion of AI in Germany. 

Leaders of small and medium-sized enterprises, for example, often believe that AI projects are too expensive. It is true that anyone who has an AI application developed in-house from scratch has to invest large amounts of money and many months to years of time.  And while the know-how is tediously built up in-house, technological developments on the market might easily overtake the project’s progress.

For corporations that are already well advanced in their digital transformation and have their own AI expertise, in-house development can be worthwhile – especially if the system is to process sensitive data, it will represent the company’s core competence, and/or no suitable application is available on the market. 

In most cases, however, in-house development is not necessary to benefit from AI. By now, there is standardized software (AI as a Service) for many use cases, companies can configure according to their requirements. While this does not provide them with unique applications that give them a competitive advantage that is difficult to catch up with, they do benefit quickly and still leave many hesitant market participants behind. 

And last but not least, the AI application concept determines the extent of value companies generate: the technology only offers the potential for competitive advantages. One and the same AI software can lead to efficiency leaps at one company, while it does not pay off for a long time at another. The key is to identify the right area of application and integrate it intelligently into the company’s own processes.  

AI as a service may not be the ideal solution for every organization, but it is a solution that allows many companies to keep up with technological developments, even if they lack the IT infrastructure for in-house development. 

Working with external service providers removes another concern: the belief that one’s own organization must already be largely digitally transformed or that huge data sets must exist in order to make effective use of artificial intelligence. Neither is true as a general rule. Of course, AI requires a certain digital infrastructure, but in many cases AI applications can be deployed for clearly defined use cases. It is not necessary to have a fully interconnected company. Bringing in AI consultants with real-world experience can help find the best possible case and avoid unnecessarily limiting the company’s success.  

Let semantha convince you!

Roadmap for successful AI projects

There are several approaches companies can take to set up AI projects. Which approach is optimal depends on the level of digital transformation, goals, and business organization. However, the key milestones of these roadmaps are always the same. 

  1. Analyze the status quo
  2. Define AI objectives
  3. Develop AI strategy
  4. AI introduction

1. Analysis of the status quo

First, senior management should take an honest inventory of their own company’s digital maturity. How pronounced is digitalization? Which departments are suitable for an AI pilot project from a technical perspective? In which departments are significant improvements important and valuable from a business perspective?

Even before individual use cases are considered, managing directors should get an overview of the data available in the company and the quality of this data. If not already done, a data strategy should be worked out and data collection systematized. The existing data base shows where AI deployment is possible. 

2. Define AI objectives

In the second step, senior management should agree on their AI goals: should the current business model be optimized, should new markets be developed, or should an entirely new business model be developed? 

Answers help to create on overview of all potential use cases. Companies should clearly quantify their benefit expectations from AI. If managers want to improve the efficiency of a process, key figures must be defined for a before-and-after comparison.  

It is important that all stakeholders approach the AI project with realistic expectations. They must be aware that the IT requirements are different from those of traditional systems, that initially more personnel effort and maintenance or training will be required before work is made easier and the added value exceeds previous methods.

3. AI strategy 

Companies start with a pilot project but they should always embed their first project in an AI strategy that is in harmony with the global corporate strategy. 

What is the overarching goal of all AI applications planned for the future? Companies should define which processes, products and business areas they want to optimize or develop with the help of AI,  which technologies and methods they plan to use and how success will be measured. Which are the financial and personnel requirements necessary for implementation? What organizational changes will have to be made successively? All of this is important data that sets a framework for further action. 

By now at the latest, companies should also have decided whether they want to implement their strategy with external software or develop solutions themselves.

4. AI introduction 

Finally, the company is ready to start a pilot project. For this, critical points in the value chain are particularly suitable. Where there has been a bottleneck up to now, optimization will likely have a high impact on productivity and the positive effects of AI will become apparent quickly. And this is all the more important because there is usually still a lot of skepticism among employees. Quick wins motivate and encourage changes in the corporate culture. 

However, the success of AI projects depends to a large extent on how well employees accept the new technology. Are they willing to educate themselves, get training, learn new processes and collaborate with AI? Project managers should announce the coming changes early on and thus give departments and teams the opportunity to prepare and participate in shaping the change. Having senior management’s support is essential. The CEO should communicate clearly to employees that AI projects are of major importance for the entire company. 

When selecting an initial use case, it often makes sense for companies to seek external advice. When AI projects are unsuccessfully discontinued it is often due to concepts that have been created in the ivory tower. External AI experts bring practical experience to the table and can steer the process with the right questions and a methodical approach to help companies avoid the biggest mistakes.

One of these mistakes: view the introduction of AI solely as an IT project. The business department affected by the new software is only brought in late or is not allowed to give feedback until the application is ready for use. This is no basis for success. Employees know the hurdles in their work processes much better than IT managers, who see things purely through the lens of technical feasibility.

Therefore, project teams should ideally consist of business unit managers who are open to new digital technologies and IT managers, with the affected department taking the lead. IT can be given veto power focusing on feasibility aspects. If an internal innovation department is in place, it makes sense to anchor the AI pilot project here – but having this kind of in-house expertise is not a prerequisite for a successful start with artificial intelligence.

Book a demo.

How companies maximize their success with AI projects  

While introducing AI starts with a project, unlike a traditional project, it is never truly complete. It is not enough to train an AI application once and then have it magically deliver the desired results. Monitoring and training costs are high, especially at the beginning. Employees not only have to train the system, but also check if the database is as suitable as assumed and updated it regularly. Results should be reviewed at intervals and use cases adjusted or expanded if necessary. 

It is important to nurture this mindset with all stakeholders early on to avoid disappointment after going live and to give the project a realistic chance of success. Often, it is only after go-live that apparent how organizational structures need to be changed and which new skills employees must develop. An agile mindset is needed when it comes to handling of technology, roles and processes. Companies must improve in iterations over several months until AI unleashes its full potential in the corporate value chain.

Customers

We have written extraordinary success stories with some of our clients. You can read them here: Success stories.