Implementation of AI: Obstacles and Recommendations
Studies paint a picture of how things are going with the implementation of AI in companies. Obstacles are often a need for more human resources and a lack of support from management.
Artificial intelligence (AI) is becoming increasingly present in our (professional) everyday life. Who would have thought 20 years ago that there would one day be software to help those responsible for sales make statements about events likely to occur?
AI-based data analyses, such as those possible with predictive analytics, help to create forecasts and derive recommendations for action. For example, how likely will a specific customer turn his back on the company (churn prediction)? Or how likely is a customer to buy a particular product or accept a specific price?
To answer these and other questions in sales, AI in predictive analytics software can be essential support.
But AI is also used in other areas. In the blog post “Where artificial intelligence already supports sales today”, you will find seven application areas.
The biggest Obstacles to the Use of AI According to Bitkom
There is plenty to say about all the advantages artificial intelligence brings to companies. The range is wide: efficiency and productivity gains, taking over tedious routine tasks, automated processes, reduced errors, and lower costs.
But until you reach that point and AI shows the intended effects, you must first implement it. A perhaps banal realisation, but practice shows this is often associated with significant hurdles. Due to the diverse business models of companies, there are many different and individual, situation-dependent hurdles, although survey results or studies allow a certain objectification of the difficulties.
For example, in a survey of 606 companies with 20 or more employees from all sectors in Germany, the digital association Bitkom concluded that AI is only making slow progress in companies.
The biggest obstacles to the use of AI in companies are currently:
– Lack of human resources (62 per cent)
– Lack of data for AI use (62 per cent)
– Lack of financial resources (50 per cent)
– Uncertainty due to legal hurdles (49 per cent)
– Lack of technical know-how (48 per cent)
– Lack of time (46 per cent)
Around a third cite a lack of acceptance among employees and a general lack of trust in AI. 22 per cent still need use cases for AI in the company. It is also noteworthy that in only 14 per cent of the companies, the management drives the topic of AI forward – which means that it is by no means a “top priority” everywhere.
Access to Relevant Data and Compliance with Data Governance as Barriers
The study “AI Ambitions 2022” by the market research institute Vanson Bourne on behalf of Fivetran, a provider of modern data integration, also identified hurdles and obstacles. 550 IT executives and data scientists from the USA, Great Britain, Ireland, France and Germany participated in the online survey. The central statement of the study: Despite ambitions and willingness to invest, companies are only partially successful in using AI within the framework of decision-making processes.
41 per cent of the companies surveyed still see the potential for improvement in the implementation of AI. Accordingly, they are realigning their IT budgets. The investment projects in Germany focus on using new technologies for data integration, security and data governance, AI and machine learning, and developing AI specialists and data scientists.
According to the study, there are these hurdles in the implementation of AI:
– Data integration
Around three-quarters of the companies surveyed can only partially access relevant data for AI systems, workloads and models for machine learning. 73 per cent have difficulties with the ETL process (Extract, Transform, Load) and can only partially translate the insights gained from data into recommendations for action.
– Financial losses
Many AI projects are started but then not completed. These projects cost companies a lot of money. Respondents estimate that, on average, five per cent of global annual revenue is lost due to immature machine-learning models based on unclean or faulty data.
– Ineffective use of personnel
Companies only use the potential of data scientists and data engineers to a partial extent. That is, even though these employees play a crucial role in implementing AI. Instead, these data experts spend most of their time preparing the data.
– Compliance with data governance
According to the study, 90 per cent of respondents see a need for improvement in compliance with data governance, i.e. the legal requirements for handling data. In Germany, this assessment is more positive. Here, only 18 per cent see little or no need for improvements.
Data are the Basis of Artificial Intelligence
The study shows: In most companies (92 per cent), the basis for using artificial intelligence has long been laid. These companies collect and use data from their operational systems and use them for machine learning models. Despite this existing technical infrastructure, however, more than half are still in the early stages when using AI, or they use it to a moderate extent. Only 14 per cent rely on AI-supported processes for decision-making. Remarkably, 90 per cent of companies rely on manual data processes rather than automation using machine learning and AI.
George Fraser, CEO of Fivetran, emphasises: “What the survey makes clear is that companies have a huge amount of catching up when transferring and accessing data. However, a successful AI programme needs a solid data foundation, which usually starts with a cloud data warehouse or data lake. Data analytics teams that rely on a modern data stack here can take full advantage of their data and realise true ROI in AI and data science.”
Implementation of AI mostly Fails due to Internal Organisation
Other expert reports, such as that of the CBS International Business School entitled “Artificial Intelligence in Small and Medium-Sized Enterprises”, name three areas in which there are predominantly hurdles in implementing AI: technology, organisation and people.
According to the report, technological hurdles include:
● Difficulty in accessing external AI expertise.
● Risk of misuse of technology, data protection and digital ethics.
Organisational hurdles include:
● Lack of support from senior management
● Lack of understanding of the strategic benefits
● Lack of database
● Poor technical infrastructure
● Lack of compatibility with processes and organisational structures.
In the case of the human factor, these barriers exist:
● Internal resistance from the workforce
● Concerns from customers
On the technical side, many see the increasing autonomy of AI-based systems as a problem. That raises the question of personal or third-party responsibility. To avoid taking a security risk here, a legal framework is recommended that must be adhered to. The same applies to digital ethics, especially to the issue of discrimination through AI.
A significant hurdle in implementing AI in companies is the need for more expertise in this field within the companies themselves. One solution is the engagement of external consultancies. The authors of the CBS International Business School study recommend approaching these firms with a concrete problem. It is essential to check their specialisations and start with a Minimum Viable Product (MVP). The advantage: The problem is manageable, it is easier to keep track of things, and you can quickly make progress.
CALCULATE NOW THE ROI OF QYMATIX PREDICTIVE SALES SOFTWARE
Management must Support the Implementation of AI in a Visionary Way
The implementation of AI is more than just day-to-day operational business. Such a step, which often touches companies in their DNA, is strategic and lays the foundations for a company’s future. Therefore, this step should be positioned at the executive level and focused accordingly. Without advocacy and support, without active management commitment to the implementation of AI, it will not succeed.
The prerequisite for this commitment from management is their conviction that AI brings a strategic advantage to the company. Those who are not convinced of the added value of this step will neither develop a vision for it nor release budgets for it. To ensure a better understanding of this added value and to follow innovations in this field, the use of so-called innovation scouts can be helpful. They visit special AI events or trade fairs and ask about new technologies there. In principle, they are something like “translators” for the management of the companies.
You can counter the fears that ultimately drive employees during the implementation of AI (for example, job loss) with well-organised change management. The following measures characterise that:
● Intensive communication
● Creating an understanding of the possibilities and functioning of AI
Open discussion of risks
Space for fears and concerns
● Further training opportunities
● All process participants and future users are involved early during the development phase.
The following applies to all implementation projects: companies should examine which problems can be solved using AI.
That requires a fundamental understanding of the technology, an AI strategy that fits the company and a team that can implement this strategy. The next step is to implement the chosen path in practice. Initially, a small use case is sufficient to gather experience. The AI application can be expanded with the knowledge gained, and you can achieve the first added value.
