7 Steps to Successfully Implement AI
A Practical Guide for CEOs
What’s exciting as a private individual can feel overwhelming as a corporate leader
It’s been a remarkable week, not only because of the U.S. presidential election or NVIDIA surpassing Apple to become the world’s most valuable company, though that was part of it. Personally, it’s been a week filled with fascinating experiences.
I attended three conferences, each focused on artificial intelligence. The first explored the implications of the European AI Act from a legal standpoint. The second showcased traditional machine learning and Generative AI solutions implemented by an IT vendor. At the third, I had the opportunity to participate as a panelist, discussing the application of AI in corporate settings.
The biggest takeaway for me from these three different discussions was realizing the vast gap between the fantasy world of ever-evolving, astonishing AI solutions flooding the internet and the everyday corporate reality.
When we look at the countless new tools as private individuals, the enormous array of options we encounter daily is indeed highly motivating, uplifting, and even entertaining. Each tool is intriguing to try out, and experimenting with them is enjoyable. Personally, I love creating beautiful visuals from text in an instant with napkin.ai, composing music with Suno, generating images with Photoshop, or producing podcasts with NotebookLM. Today’s popular generative AI tools have happily amassed hundreds of millions of monthly users.
Source: Andreessen Horowitz
But are these tools really suitable for answering a business leader’s question: “How should I approach AI implementation?”
No, or only in a very limited way. The vast majority of these tools—at least in their well-known, free versions—are most useful for private individuals or small, one-person businesses, offering simple, modular solutions for smaller business challenges.
However, applying them in a corporate environment raises a host of questions that, if left unaddressed, could lead to major setbacks.
Here are the key questions we face daily:
Which tool should I choose from the many available?
There are dozens of large language models (LLMs) alone. The most well-known is ChatGPT, but there are also numerous other viable options on the market, many of which outperform it in specific areas. This analysis compares various LLMs.
Source: https://mindsdb.com/
And this is just the tip of the iceberg when it comes to language models. Beyond them, there are thousands of tools for image generation, video creation, music composition, and podcast production. And we haven't even touched on non-generative AI solutions…
Here’s an incredibly detailed and useful compilation of the most important AI solutions for 2024. It’s overwhelming. As the Germans say, "Qual der Wahl," or the agony of choice. Check it out here.
Source: https://mad.firstmark.com
Returning to the frequently asked question, I believe that the question itself is unfortunately misguided. It’s tempting to grab onto a tool and quickly produce a flashy, understandable demo or Proof of Concept. But is this the right approach? Does the company even know what problem it truly wants to solve? Is it spending money wisely? Is this problem really the most critical to address? Is the solution sustainable? Worthwhile? Is there a cheaper, better option? Usually, these questions remain unanswered.
How should I even begin with AI?
This is a much better question, though the answer is far from simple. Many companies have been using AI solutions for years, possibly without even realizing it. Does the company use Microsoft 365? It likely already incorporates AI through features like Copilot. With a premium subscription, Teams can transcribe meetings, and Office has become quite “smart.” Does it use SAP? Then it likely leverages AI. Are there elevators in the office? Those are often AI-controlled. Is there a logistics fleet with routes and driver schedules optimized by computer algorithms? Behind all of these are AI-powered solutions.
Most large companies today, by using modern management systems (ERP, logistics, CRM, ticketing, monitoring, predictive financial forecasting), inherently gain built-in and continuously advancing AI capabilities.
The question today often centers on how to use the rapidly advancing generative AI solutions that became widespread with the breakthrough of ChatGPT two years ago.
The good news is that for a company that keeps its core management systems, IT infrastructure, business processes, and regulatory compliance frameworks up-to-date, adopting generative AI solutions actively and with a quick ROI is not a huge, insurmountable task.
However, for companies lacking or missing these foundations—and there are many in this situation—beginning with building the basics is essential. The journey is similar for both types of companies, but those that are less mature will need to spend much more time establishing a solid foundation.
Establishing the Foundations: A Path to AI Implementation and Development
Below, I present a simple framework designed to help any business leader systematically review and consider:
how thoroughly their company has examined and developed the foundational requirements for AI implementation,
what steps are necessary to solidify these foundations,
and, based on this, in which areas and how AI can support the achievement of their goals.
An overview of the framework and its main components is illustrated in the diagram below.
Source: Horváth
A have merged some of the above framework components to get a simpler path that includes key steps for every business leader in AI implementation
1. Setting Clear Goals
The first step is to clarify why we want to engage with AI. Jumping in solely due to the hype is not advisable. Understanding the specific objectives—whether improving efficiency, enhancing customer experience, driving innovation, or achieving competitive advantage—helps shape a meaningful AI strategy.
Source: Gartner
As illustrated in the diagram, generative AI currently sits at the peak of Gartner’s hype cycle—the “peak of inflated expectations.” There’s a lot of pressure to join the movement, but starting projects merely to keep up with others isn’t wise.
Beyond the superficial goal of “keeping up with the competition,” there are more substantive objectives that AI adoption can fulfill. Broad AI implementation can mobilize and transform the entire company, essentially preparing it for the future. This future will likely see the widespread adoption of AI solutions, bringing a world where humans and AI agents work together in hybrid teams, supporting and strengthening one another. Preparing for this shift cannot start too soon.
Additional, straightforward goals for AI adoption include increasing efficiency, enhancing employee satisfaction, and boosting retention. There are numerous benchmarks and examples supporting these benefits, but the key is to explore them systematically, as covered in the next section.
In many companies, the first critical step—clarifying objectives—is often overlooked. Without this, it’s difficult to get started effectively, so don’t skip it. To define goals, it can be beneficial to bring in external support, as an independent facilitator can help surface these objectives more easily within the leadership team.
2. Identifying Use Cases and Added Value
Once we’ve established our goals with AI solutions, it’s essential to systematically consider which areas and solutions could be beneficial.
Is the goal to transform the company by involving as many employees as possible? In that case, it may be wise to start with foundational training, followed by a general tool that everyone can use easily, such as a company GPT chatbot to help interpret internal policies or streamline onboarding.
If IT efficiency is the goal, a tool that automatically processes IT tickets or creates documentation for deployed IT applications could be ideal.
If the goal is to increase customer retention, a tool that predicts churn might be our solution.
The range of potential use cases and solutions is nearly endless and unique to each company. The key is to evaluate them based on a consistent set of criteria. Most evaluations consider both the potential value created by the use case and its feasibility. Each of these factors can have multiple sub-criteria, which can even be customized at the company level.
The diagram below shows Gartner’s evaluation model through a Contact Center example.
Source: Gartner
In an average company, dozens of realistic, high-value use cases can be identified. These can be internal (HR, procurement, finance, IT) or customer-facing, market services (such as customer service chatbots, voice assistants, etc.).
It’s essential to prioritize these use cases, as it’s impractical to launch too many simultaneously. Resources are always limited, so it’s wise to start with one or two use cases that are easiest to implement and create the most value. This approach also helps promote and spread AI initiatives within the company.
To determine the business value of each use case, it’s also necessary to consider the expected costs of developing, implementing, and operating each AI application (from a TCO, or Total Cost of Ownership perspective). This can be challenging, as AI solution pricing is often complex and can be surprisingly high.
While many popular AI applications have free versions, they are generally unsuitable for serious corporate use. Key issues include:
Data Privacy Concerns: Free versions often store data in the cloud and may use it for learning purposes, raising privacy issues.
Limited Capabilities: Free versions have constraints, such as limited text input length, image resolution, or audio/video duration.
Lack of Centralized User Management: Free versions don’t offer secure, centralized user management.
Cybersecurity Risks: Allowing employees to register on external services with corporate email—and possibly even passwords—can pose security risks.
Non-Integration with Corporate IT Systems: Free systems aren’t managed or operated by corporate IT, which can compromise control and security.
Most companies struggle to identify AI use cases because they lack visibility into what today’s available tools can actually achieve, how much they cost, or how to bring their own leaders together for a collaborative brainstorming session. Often, time constraints, lack of motivation, or the absence of an effective methodology make it challenging to run a productive use case definition workshop. In such situations, seeking external support can be highly beneficial.
3. AI Governance, Risk Management, and Organizational Processes
In many companies, it remains unclear who is responsible for overseeing the implementation, ongoing supervision, and development of AI. Is it the CEO’s role? The CIO’s? Or a team’s?
Who decides which AI solutions will be adopted? Who ensures the proper handling of training and test data and addresses potential biases? Who is accountable for documenting AI applications and ensuring transparency for customers? Who monitors AI systems, given that they evolve over time and may develop issues even if they initially perform flawlessly? What risks do AI systems entail, and how should they be managed?
These are critical questions that must be addressed by establishing the right organizational structures and regulatory processes.
For organizations that already have ISO quality management, IT governance, and compliance standards in place, adapting these frameworks for AI can help ensure safe, compliant, and effective AI operations. For those lacking these foundations, now is the time to develop them.
It’s wise to initially define AI implementation as a significant company-wide project or program, with sponsorship and leadership positioned at the highest levels. An AI initiative is most credible and impactful when driven by the Board or CEO, setting an example for the entire company.
Broad employee involvement is also beneficial. Engaging 5-15% of the workforce in pilot testing and early development can enhance user experience and support wider adoption across the company.
While projects eventually conclude, AI systems continue to grow—in capabilities, scale, and reach. It’s crucial to establish permanent organizational units and processes to manage these systems effectively over the long term.
In many companies, basic business and IT processes are either undefined, have low maturity, lack consistency, or are poorly regulated. Integrating the necessary regulations for AI into such a low-maturity environment is challenging. However, it’s possible to start by establishing regulations specifically for AI processes, even if there are no existing company-wide standards.
4. Training and Engaging Employees
For AI implementation to have any chance of success, it’s essential to develop employees’ AI literacy—their ability to use AI tools effectively.
If employees don’t know how to handle AI tools properly, understand each tool’s purpose, or create effective instructions (prompts), the results will be poor, leading to frustration and a negative perception of AI as a whole.
Training should be tailored for different target groups (leaders, general users, IT specialists) with a focus relevant to each group’s role.
In addition to AI literacy training, it’s also beneficial to involve a broad range of employees in testing and experimenting with applications during pilot projects. This approach helps foster acceptance of the technology and alleviate any concerns.
A common mistake is that companies treat AI implementation as just another IT project, with IT solely in control, while neglecting to adequately prepare users and address essential change management tasks.
5. Establishing AI Technology Foundations
Due to the high computational demands of AI models, most solutions are cloud-based. While smaller models can now run on local devices, the rapid innovation cycles and constantly evolving AI services make cloud-based operation almost unavoidable.
The company must have the capability to agilely test new solutions and, through well-structured processes, integrate them into its workflows and IT operations.
This requires creating a user-friendly yet secure environment, supported by a robust IT infrastructure and technological foundations.
In many companies, the necessary IT foundations are lacking. The application landscape is overly heterogeneous, user management is inconsistent, and there are no well-established processes for IT development, implementation, or operations. Additionally, continuous monitoring and cost control of cloud-based applications are often insufficient. Without these fundamentals, launching AI projects can be risky, as costs can quickly spiral out of control.
6. Data and Algorithms
AI applications require data—data for training models, testing, and validation. Data is also generated during model use, which can be recycled to further improve the model.
If a company’s data is insufficient in quantity or quality, the applicability of AI models will be limited. Building a company-specific GPT model is futile if the documents fed into it contain inaccurate or inadequate information, as the resulting answers will be unreliable.
Every company has the responsibility to establish and maintain effective data governance, ensuring the structure and quality of master data and valuable business data.
7. Ensuring Legal Compliance
Implementing AI systems requires adherence to numerous existing and upcoming regulations, many of which are extensive and complex.
Here is a list of the most important regulations mandatory for companies operating in the European Union:
AI Act: The AI Act is the world’s first comprehensive legislation for regulating artificial intelligence, effective from July 12, 2024. Its goal is to enhance the safety and transparency of AI systems. Link: EU AI Act
GDPR (General Data Protection Regulation): GDPR is the EU’s primary data protection regulation, covering the protection of personal data, including data processed by AI systems. Link: GDPR
NIS2 Directive (Directive on Security of Network and Information Systems): The NIS2 Directive aims to strengthen cybersecurity within the EU, particularly to protect critical infrastructure, impacting the security of AI systems as well. Link: NIS2
DORA (Digital Operational Resilience Act): DORA’s goal is to enhance digital operational resilience in the financial sector, including AI systems used in financial services. Link: DORA
ePrivacy Regulation: The ePrivacy Regulation aims to protect privacy in electronic communications, closely linked to GDPR and AI applications. Link: ePrivacy
Cybersecurity Act: This law strengthens the EU cybersecurity agency (ENISA) and introduces cybersecurity certifications, which may be relevant for the security of AI systems. Link: Cybersecurity Act
Data Governance Act: The Data Governance Act aims to improve the management of public and personal data, which plays a key role in AI data handling. Link: Data Governance Act
Artificial Intelligence Liability Directive: This proposed directive aims to clarify liability for damages caused by AI systems, ensuring user protection. Link: AI Liability Directive
Globally, many other AI-related regulations are in effect, but it’s clear that the EU leads in regulatory coverage—at least in terms of quantity. What this means for the EU’s competitiveness is another matter (not very promising). For further insights, I recommend Mario Draghi’s EU Competitiveness Report, a thought-provoking read.
Compliance with regulations is a responsibility for every company. The AI landscape is increasingly characterized by stricter regulations, as policymakers have recognized that AI tools are advancing at lightning speed and may pose numerous, previously unforeseen challenges. Currently, regulators are struggling to keep up with these developments. As a business leader, it is essential to monitor all relevant regulations continually and ensure compliance, which, unfortunately, requires increasingly substantial resources.
Expected Impacts of AI Implementation
In an earlier article, I mentioned that, according to a Gartner survey, 59% of CEOs consider AI the most important technology for the next three years. Significantly so.
Source: Gartner
It’s no coincidence that successful AI implementations hold substantial potential for improving efficiency and increasing EBIT.
The introduction of artificial intelligence brings numerous advantages that are particularly attractive for business leaders, as they provide a direct competitive edge by enhancing efficiency and optimizing operations.
A series of interesting, useful AI-related statistics can be found at this link.
Here, I summarize the most important efficiency-related benefits:
Expected Impacts of AI Implementation
Widespread Adoption and Rapid Growth: According to McKinsey, 50-60% of companies are already using AI to automate various business processes, which can increase productivity by 10-15% annually. IBM’s survey indicates that 35% of companies actively use AI, while another 42% are considering implementation, primarily to improve cost-effectiveness and customer experience.
Cost Reduction and Long-Term ROI: The AI market value reached $136.55 billion in 2023, with a projected annual growth rate of 40%. This growth offers companies the opportunity to optimize costs, especially through automating repetitive tasks, resulting in significant long-term returns.
Decision-Making Support and Faster Responsiveness: Generative AI allows companies to analyze large datasets quickly, facilitating faster and more informed decision-making. According to Forrester research, AI-driven decision-making boosts executive confidence and can speed up decision processes by 20-30%, especially in a rapidly changing market environment.
Improved Customer Experience: AI-powered chatbots and virtual assistants enable more efficient customer service by providing faster, more accurate responses to user inquiries. Strategy Analytics reports that 41% of customers positively view the quick service provided by AI solutions, enhancing loyalty and satisfaction.
Future Growth Potential: AI’s global economic impact is massive: PwC forecasts that AI will contribute $15.7 trillion to the global economy by 2030. This growth will be driven primarily by AI-supported new technologies, more efficient business processes, and the development of innovative products.
Let’s Dare to Dream Small! Take That First, Concrete Step…
To any business leader who hasn’t yet ventured into the world of AI, my advice is clear: start the journey. There’s simply no alternative. AI will play a critical role in all future innovations and key technologies. We need to prepare our people and our company to live with these tools—and, once we use them, to maximize their potential.
You don’t need to think in terms of all-encompassing, complex solutions right away. As a first step, a leadership training session or user education program will do. Then, gradually, follow the path outlined above.
If you’re still unsure where to begin or would appreciate someone to accompany you on the journey, I’d like to hear your comments or questions!