AI’s transformative potential and its challenges

Generative AI, with other AI disciplines, is rapidly revolutionizing how we do business and our work, unlocking opportunities for unprecedented efficiency gains, innovation, enhanced customer service, streamlined operations, and more insightful data-driven decision-making. Still, it can be hard to identify the correct use cases where to invest that will bring real tangible benefits. As with any other IT project, also AI initiatives can fail – and some of them will. Investing or evaluating products that have AI features can be expensive, and the AI features can be black boxes with narrow use cases, lack of transparency and customization capabilities. These can lead to lower value and lack of trust and might not fit to your business goals and needs. When investing in AI enriched products look for transparency, flexibility, and customization capabilities to be able to match your goals and possible future needs.

Understanding the limits of general-purpose AI tools

Web based solutions such as Microsoft Copilot or ChatGPT are great for simple conversational use cases, but they have limits in what data they can access and how they are used or integrated to your business processes, in addition to the data security and privacy concerns. Hallucination is also one concern when using these generalized solutions. How should you then get started with AI-solutions that provide value and integrate to your business processes?

The value of a strong AI-partner

It can be hard to understand what AI can do and what it cannot. Here a good partner is valuable – that can assist in your AI journey in both business and technical perspectives:

• Understanding your business, assisting in identifying, validating, and prioritizing the potential opportunities
• Understanding ethical aspects of AI such as bias, transparency, privacy, security and accountability including understanding of applicable laws and regulations like EU AI Act
• Technical competence to help to understand the AI technology, options, trade-offs and build the AI solutions.

Proof before scale: The smart way to validate AI

Solution demos and proof of concepts are a great way to prove the initial AI solution value. Not only that the opportunity is technically doable but making visible that you have the right data and that it can be used for the intended purpose. They are excellent to ensure stakeholder buy-in by demonstrating the potential of AI within the organization. From business perspective they validate the business case, for example that the solution has potential to increase sales, increase process quality, reduce costs or improve customer satisfaction. In addition, it should be visible that the value of the solution can be improved with more and better data, examples, and more time to tune the solution such as process complexity and LLM prompts.

One way to start your AI journey is to build short-lived or minimum viable product (MVP) solutions. These should be quick to build, have low costs but still use your real data and provide value. These can build trust, expand AI knowledge and act as the foundation for your next steps in your AI journey.

Case example: Improving product data with AI in 2 days

We have a very good example of this from a recent customer case, where we executed product name enrichment and translations for hundreds of thousands of products. The enrichment was not a simple task because of the domain specific terminology, custom abbreviations and spelling variations. But by correctly creating the AI-process and instructing LLM the results were excellent improving the product findability with uniform spelling, wording conventions and translations. And all of this was done in less than two days.

Data quality is critical for AI success

Examples like the above are great – not only to prove the AI value – but to improve data quality which plays very important role in AI processes. AI success hinges on data quality to be able to provide value and reliable results. Make sure that your data governance practices are in place to ensure data quality, data security and data availability for the AI solutions.

You can read more of this from our article The meaning of data quality for augmented AI initiatives.

Build a strategic AI portfolio

When you move forward, AI portfolio and literacy start to play more important role. Like more traditional investment areas, AI initiatives should be collected to a portfolio and prioritized correctly. Value, effort, and risks are topics to consider when prioritizing AI initiatives and low effort, low risk and high value are the ones where to start. However, consider starting with lower pace and then scale up when you have more experiences and proven values. Previous experiences are also important matter to consider when prioritizing AI initiatives. There is always potential that in areas which you have not explored yet, the value might not be what you can otherwise expect.

Change management in the age of AI

AI literacy is how you share, educate, and build the AI-culture in your organization. AI poses many questions in employees especially about job displacement and good AI literacy practices play big role in how your employees react to AI initiatives. Share your AI goals, how they affect current roles, what possibilities they offer in the future and educate your employees to use AI in everyday tasks. Two-way communication is important to be able to understand employee concerns and to get feedback to be able to improve the AI literacy practices. Be a proactive AI-leader to gain the trust.