Opportunities and challenges of artificial intelligence in companies

Using AI for process automation

Increasing productivity through AI

Better working conditions through AI

Challenges in the introduction of AI

Recommendations for action for companies

Conclusion

A key area of application for AI in companies is process automation. This refers to having recurring processes and tasks carried out by technologies in order to relieve the burden on human labour and speed up processes. Traditional robotic process automation (RPA) has already been able to automate rule-based, repetitive tasks - such as data entry, invoice processing or filling out forms. AI is now significantly expanding this automation: by using machine learning and cognitive skills, even more complex processes with unstructured data or decision-making logic can be automated.

According to the Bitkom survey, more than half of companies expect AI to significantly speed up their processes (see bitkom.org). The advantages of AI-supported process automation are obvious: workflows run faster and with fewer errors, resource consumption is reduced and costs are saved (see bitkom.org). For example, an AI system can automatically classify incoming emails and forward them to the relevant departments or enter orders into IT systems independently. This shortens processing times and frees employees from monotonous routine tasks.

Modern solutions combine AI and RPA to create intelligent automation. For example, AI models for document recognition or decision-making can be integrated into RPA processes in order to fully automate end-to-end processes.

In practice, this means that an AI module semantically understands documents and an RPA bot then executes the corresponding subsequent steps in various applications. One application example is the combination of the AI agent platform AIMAX® with the RPA solution EMMA: both systems work together synergistically - EMMA uses the generative AI intelligence of AIMAX® to improve processes, while AIMAX® uses the RPA platform to trigger automated workflows.

Through such approaches to hyperautomation approaches enable companies to automate even complex business processes from start to finish without the need for manual intervention. It is important to analyse and standardise processes in advance to ensure that AI-based automation functions smoothly.

Despite the enormous opportunities, decision-makers should not underestimate the challenges of process automation: It must be clearly defined which processes are suitable and what data the AI needs for them. Cross-departmental collaboration (specialist department and IT) is often necessary to bring the process knowledge together with the technology.

Companies should also start small - e.g. with pilot projects in a defined process - and gradually expand based on the results. This allows them to demonstrate early successes (quick wins) and create acceptance for AI within the company while minimising risks.

In addition to the automation of individual processes, AI contributes to an overall increase in productivity within the company.

On the one hand, AI increases the speed and precision with which tasks can be completed. Routine tasks that used to take a lot of time now run automatically in the background - around the clock and without errors. Employees can work on other tasks in parallel, which increases the overall output volume.

On the other hand, AI enables completely new approaches in areas such as data analysis, decision-making and customer interaction that were previously time-consuming or even impossible.

In the Bitkom study, companies cited numerous expected productivity benefits from AI: from faster and more precise problem analyses and accelerated processes to cost savings and less resource consumption.

AI systems can, for example, analyse huge amounts of data within seconds and recognise patterns where a human would need days - whether in production monitoring, financial controlling or marketing. Such a system even provides expert knowledge that would otherwise not be available by collating information from thousands of sources. As a result, decisions are more informed and in turn made more quickly.

Another aspect is error reduction: automated AI processes make fewer careless mistakes than manual activities (see bitkom.org). Fewer errors mean less rework and quality losses - which also effectively saves time and costs.

In addition, AI can provide proactive support through predictive models (e.g. predictive maintenance in maintenance or predictive analytics in purchasing). Problems are anticipated before they arise and measures can be taken in good time instead of having to react after the fact. This increases the efficiency of processes and ultimately the added value per employee.

Last but not least, AI can also enable completely new products and services - such as personalised recommendations, autonomous service bots or data-driven business models - that open up additional sources of revenue (see bitkom.org).

Companies that utilise AI strategically strengthen their competitiveness and future viability. This connection is also being emphasised politically: Vice Chancellor and Federal Minister of Economics Robert Habeck emphasised at the AI Summit 2024 that broad AI applications are necessary in order to exploit productivity potential for our economy. For individual companies, this means that AI can be a decisive lever for achieving more than the competition with limited resources.

Of course, the actual productivity boost achieved depends on several factors:

• AI projects should be clearly aligned with business goals (e.g. reducing lead time by X per cent or improving sales forecasting accuracy by Y per cent).
• The integration of AI solutions into existing processes must function smoothly - stand-alone solutions bring little productivity if they are not embedded in the overall workflow.
• And employees must be empowered to work with AI in order to realise its full potential (more on this in the next section).

If this succeeds, the productivity gains from AI are considerable: studies assume that AI can noticeably increase economic growth in the long term and significantly boost productivity per employee.

In addition to hard metrics such as process indicators and productivity, AI also has an impact on employees' working conditions. Ideally, the use of AI should mean that employees are relieved of stressful, monotonous or dangerous tasks and can concentrate more on fulfilling activities (see bitkom.org).

In fact, many companies expect precisely this effect: AI-supported automation relieves employees of routine tasks so that they can focus on other, more demanding tasks.

A simple example is the automatic creation of meeting minutes by AI - instead of an employee having to laboriously take notes, they can concentrate fully on the discussion and receive AI-generated minutes afterwards.

Working conditions improved by AI can be seen in various facets:

Activities such as data entry, comparing lists or answering the same enquiries over and over again not only take time, but are also tiring for humans.

When AI takes over these tasks (for example as a chatbot in customer service or as an automated accounting assistant), the mental strain on employees is reduced.

Instead, they can use their working time for creative problem-solving, customer support or strategic tasks, which is often perceived as more fulfilling.

AI can help to better manage workloads, especially at peak times.

In a customer contact centre, for example, an AI system can automatically answer simple enquiries so that employees have to handle fewer calls in queues. This reduces stress and hecticness in everyday working life.

Process automation reduces bottlenecks - deadlines are easier to meet because preparatory work is completed more quickly.

Ideally, this also reduces the need for overtime, which benefits the work-life balance of the workforce.

In industrial environments, AI (in conjunction with robotics) can take over dangerous or physically demanding work.

Be it the use of intelligent robots in production or AI-supported drones in inaccessible areas - humans are less directly exposed to dangerous situations.

AI also reduces the risk of human error that could lead to accidents. Overall, the working environment becomes safer.

By relieving employees of routine tasks, AI enables them to upskill and develop new skills. At the same time, AI enables employees to work with advanced tools, which initially requires training but increases digital expertise in the long term.

For example, a clerk who carries out data analyses with the help of an AI assistant also learns more about data interpretation.

AI can also serve as a virtual coach - for example, by providing personalised learning recommendations or feedback (e.g. an AI-based training system that automatically gives advice on how to create an even better customer experience in customer service after a call).

It is important to carefully organise the use of AI in the working environment. Employees should see the introduction of AI not as a threat, but as a benefit.

This requires transparent communication: what can AI do, what tasks will it take on and how will it change their own area of responsibility?

Decision-makers are well advised to dispel fears and prejudices at an early stage. It often helps to point out specific examples of how AI makes everyday work easier. If employees see that they are relieved of routine work and can take on more interesting projects in return, acceptance increases noticeably.

Co-determination and further training are decisive factors here: the workforce should be involved in the introduction of AI and trained accordingly to create a human-machine team working in partnership.

Last but not least, AI assistance systems can directly support employees in their daily work. Think of intelligent assistants that quickly provide relevant information for complex questions (similar to a search, only more contextualised) or AI systems that suggest new solutions. Such tools provide expert knowledge on demand (see bitkom.org) - even less experienced employees can make informed decisions, which makes their work easier and increases quality.

When AI is perceived as a colleague who has your back and helps you, job satisfaction and motivation improve.

Despite all the opportunities, the challenges of introducing AI in companies should not be neglected. Decision-makers should be aware of the potential hurdles so that they can actively counter them:

AI systems are only as good as the data on which they are trained and used. Many companies are faced with the task of providing sufficient quantities of good quality data.

Data is often distributed in silos, unstructured or in inconsistent form. A data integration and cleansing phase is therefore often required before AI can deliver added value.

Equally important is the issue of data protection - a key concern in Germany in particular. When training AI (especially large language models or cloud AI services), it must be ensured that no sensitive personal data flows out in an uncontrolled manner. Data protection-compliant AI applications may require anonymisation techniques or the use of on-premise solutions, which increases complexity.

The lack of AI expertise is one of the biggest internal hurdles. Data scientists, ML engineers and AI strategists are in demand, but rare on the labour market. At the same time, there is often a lack of internal expertise to initiate and manage AI projects. Companies counter this by providing employees with further training or utilising external partners.

An alternative strategy is no-code or low-code platforms, which also enable non-IT experts to configure AI applications (as in the case of the EMMA platform mentioned above, which allows departmental employees to create automations themselves). Nevertheless, the development of internal AI expertise remains important in the long term in order to be able to act independently and utilise the technologies effectively.

AI projects are change projects. Resistance arises not only among employees who fear for their jobs, but sometimes also among management if the benefits are not clear enough.

A lack of AI strategy or unclear responsibilities within the company can hinder the introduction. It should be defined from the outset who controls the AI initiative, how decisions are made and how results are measured.

Without top management sponsorship and cross-departmental collaboration, AI projects run the risk of becoming siloed.

Ethical guidelines must also be defined in good time: Which applications of AI are desired, where do we as a company draw boundaries (e.g. regarding fully automated decisions without human scrutiny)?

The implementation of AI solutions in existing IT landscapes poses technical challenges. Legacy systems are not always easy to connect to modern AI APIs. Interfaces have to be created, computing capacities provided (sometimes very high for training large models) and answers found to IT security issues.

AI models can be susceptible to bias (distortions) or deliver unexpected results (keyword: hallucinations in generative AI models). Continuous monitoring of AI systems in productive operation is therefore necessary. For critical applications, there must also always be a fallback or backup in case the AI fails.

The legislation surrounding AI is currently undergoing an intensive development process. In particular, the upcoming EU AI Act will set the framework for AI systems in Europe. Many companies are unsure about what regulations and compliance requirements they will face.

Depending on the risk classification of an AI system (e.g. high-risk AI for personnel decisions), strict regulations on transparency, risk analysis and monitoring could apply in future.

Bitkom President Dr Ralf Wintergerst therefore called for politicians to give companies sufficient leeway when implementing the AI Act so as not to stifle innovation.

In practice, this means that companies must keep a close eye on regulatory developments and ideally create internal standards early on that address responsibility and ethics in the use of AI (e.g. setting up an AI ethics council or internal guidelines for AI projects). This will also help to build trust in AI-supported processes with customers and partners.

Given these challenges, it is understandable that 41% of German companies still do not consider AI to be a relevant topic (as of 2024, see bitkom.org). However, the number of AI adopters is falling rapidly (in 2023 it was still 52%, see bitkom.org), while more and more companies are starting to engage with AI. Those companies that proactively tackle the hurdles will be at an advantage - they will create the basis for fully exploiting the opportunities presented by AI.

So how can decision-makers go about realising the opportunities of artificial intelligence and mastering the challenges? Here are some recommendations for starting and expanding AI in the company:

Determine what your company wants to use AI for. Focus on the overarching business objectives - AI is not an end in itself.

Identify areas with high potential (e.g. processes with high manual effort, data-intensive analyses or customer-facing processes) and prioritise them.

Define clear goals for each use case (e.g. "reduce throughput time by 30%" or "increase customer satisfaction by one point").

A written AI strategy creates orientation and commitment.

Start with manageable pilot projects to gain experience. Choose a process or use case that promises tangible benefits but is not business-critical. Test an AI solution there and measure the results.

Learn from successes and failures.

Important: Celebrate quick wins and communicate successes internally to generate enthusiasm. The findings from pilot projects serve as a basis for gradually rolling out AI to other areas.

Put together a team that combines different competences - e.g. technical experts from the business unit, data scientists/analysts, IT experts for integration and change managers. This core team drives the AI initiatives forward.

Clarify responsibilities:

• Who is the product owner of the AI project?
• Who looks after the data?
• Who communicates with the workforce?

If the expertise is lacking internally, bring in external consultants or partner companies, but with the aim of building up your own expertise in the medium to long term.

Create the technical foundations so that AI can work successfully. This includes connecting data sources (keyword: data warehouse or data lake) and ensuring high data quality. Invest in tools for data preparation and management.

If necessary, establish data governance rules so that it is clear which data may be used (data protection!) and how it is maintained.

At the same time, you should check whether the existing IT infrastructure is sufficient - computing-intensive AI applications in particular sometimes require upgrades to hardware or cloud resources.

Involve all stakeholders at an early stage. Inform them openly about planned AI deployments, the expected benefits and possible effects on work processes.

Create acceptance by taking fears seriously and reducing prejudices - e.g. through workshops in which AI systems can be presented and tried out. Offer further training measures so that employees acquire the necessary skills to deal with AI tools (e.g. training in data analysis, training in the operation of new software or basic understanding of AI for everyone).

Promote a learning culture in which employees can make suggestions for improvements in dealing with AI. Employees who have recognised the benefits of AI for themselves will become important ambassadors within the company.

Develop guidelines for the responsible use of AI. For example, stipulate that decisions with far-reaching consequences (hiring decisions, loan commitments, etc.) are not made without human oversight, even if AI is involved.

Ensure that AI decisions are transparent wherever possible - especially if they need to be explainable to customers or supervisory authorities.

Keep an eye on legal developments: Ensure that your AI applications comply with current data protection rules and prepare for the requirements of the EU AI Act.

Document AI models and training data to be accountable.

This forward-looking approach prevents nasty surprises and strengthens confidence in your AI solutions.

AI is a rapidly developing field - nobody has to reinvent the wheel alone. Network with other companies, e.g. in industry associations or at specialist conferences, to learn from experience.

Consider pilot projects with research institutions to gain access to the latest knowledge. Exchanges with start-ups or the use of proven platforms (such as existing AI services for speech or image recognition) can also make it easier to get started.

A make-or-buy assessment is advisable: not every AI solution needs to be developed in-house - the clever use of existing solutions is often the faster route to success (solutions mentioned in the article such as AIMAX® or EMMA are examples of flexible AI tools that companies can use directly instead of programming their own AI systems from scratch).

Define KPIs for AI projects right from the start (e.g. processing time, error rate, customer satisfaction, sales increase) and measure them regularly. This allows you to visualise progress and recognise any need for action at an early stage.

Establish feedback loops: Results from live operation should be fed back into the further development of the AI. It may become apparent that a model needs to be adapted or retrained, or that a certain process step is still stuck - use these findings to iteratively improve your AI applications.

The introduction of AI is not a one-off project, but an ongoing optimisation process.