Four Expert Tips on Business Value with GAI
Generative AI (GAI) is revolutionizing many aspects of our work life. How can we make the most of GAI’s potential, and where should we set limits? In this article, our AI consultant Jonas Pomoell shares his thoughts on how to best leverage GAI’s capabilities from a business perspective.
For just $5 a month, people can effectively use GAI in their daily lives. It turns out that GAI is not just a fun tool for creating shopping lists or bedtime stories. More relevant use cases are emerging in the corporate world. Here, we share key insights on how GAI can make a real difference in your business.
1. AI Tools Can Be Unpredictable – Set Your Quality Goals
When Jonas Pomoell, AI Lead Consultant at HiQ Finland, explains his passion for generative AI, he doesn’t hesitate to emphasize its indeterministic nature:
“Generative AI is fundamentally flawed and even chaotic. Instead of providing a correct answer, GenAI gives you a random one.”
This means that generative models strive to understand the underlying probability distribution of the data they are trained on, enabling them to create new examples that resemble the training material. When GenAI is faced with a specific question, it always gives an answer, but it doesn’t necessarily have to be correct or optimal—the response is probabilistic.
Effective risk management is therefore crucial when using GenAI. Companies need to clearly define their quality standards and what should happen when answers are incorrect. GenAI can be supplemented with human verification or by automatically dismissing responses that do not meet specific criteria.
“Risk assessment is essential, and it’s something we emphasize at HiQ in every scenario. It’s important to understand the potential outcomes if the model fails to perform. For example, GenAI can serve as a useful tool for creating meta descriptions for e-commerce stores. If the generated text is not up to standard, an employee can manually improve it. The risk here is relatively low, and GenAI has likely saved significant time and effort,” he explains.
Imagine a situation where GAI optimizes 90% of order processing tasks, which is quite an achievement. Sure, the last 10% might require a human touch, but let’s not overlook the massive relief GAI has provided. That’s something to celebrate.
2. Manage Your Expectations
When it comes to successfully implementing GenAI, managing expectations is a must. Overly optimistic expectations can undermine the actual value of GenAI’s capabilities.
Imagine a situation where GenAI optimizes 90% of order processing tasks, which is quite an achievement. Sure, the last 10% might require a human touch, but let’s not overlook the massive relief GAI has provided. That’s something to celebrate, says Jonas Pomoell.
According to him, a common pitfall is poor planning. Companies eager to leverage GenAI can overlook important aspects such as how it fits with the existing IT architecture, what its intended use case is, and defining a clear purpose for it. Moreover, implementing GenAI requires training for employees and potentially changing services. Although GenAI can be transformative, it does not offer an instant solution—it rather reveals emerging development needs.
Timing is another critical aspect. Allocate sufficient time for processing and maintaining source data, which can take up to half of the development time.
“Creating an AI model might seem simple, but data management often exceeds initial estimates. Viewing it as a learning curve is crucial. It requires readiness to delve deep, have discussions, and address potential deviations in the implementation process.”
3. Saving Data = Saving Money
Generative AI demonstrates impressive capabilities in formulating statistics, extracting data, and efficient document management. However, Jonas Pomoell emphasizes the importance of cost management in every GenAI project.
“When planning for GenAI in your business—remember that costs are based on usage. The more you generate, the higher the costs.”
In GenAI, costs are based on tokens, the smallest units of text data processed by large language models (LLMs). Tokens can take various forms—they can represent characters, words, or larger text segments like phrases, depending on the model. Costs can accumulate, with prices around 0.15 cents per 1,000 tokens, a significant amount for industrial players handling massive data volumes.
Therefore, a key component of strategic planning is designing the system to collect data during the process, thereby reducing the need to regenerate the same data.
“During the process, it’s important to remember that generation can be both time-consuming and expensive. Saving the results along with extensive metadata is a smart practice. This method allows you to reevaluate and rework the results, ultimately saving both time and money. Remember, the more you generate, the more it costs,” says Jonas Pomoell.
Choose the cheapest and fastest way to test your ideas during the experimental phase, because you never know if they will work. Once the project is up and running and generating real value, the costs for GenAI become more transparent, and that’s when it’s smart to focus on reducing expenses.
4. Start with Commercial Models
Before diving headfirst into a GenAI project, you need to understand how well-planned strategies can be an enormous cost-saving measure. Initially, Jonas Pomoell recommends validating your hypotheses with commercial models before making major investments.
“Choose the cheapest and fastest way to test your ideas during the experimental phase, because you never know if they will work. Once the project is up and running and generating real value, the costs for GenAI become more transparent, and that’s when it’s smart to focus on reducing expenses.”
Using an in-house model with a clear value proposition often proves to be a sustainable solution. An in-house model, fine-tuned with unique data tailored to a specific use case, becomes highly resilient to replication. For some businesses, owning a model with intellectual property rights can even create an “unfair competitive advantage.”
“Companies that create their own models can establish enormous competitive advantages, with attributes that are extremely challenging for competitors to replicate. This makes GenAI an excellent choice for pioneers aiming to stay ahead of their competition,” concludes Jonas Pomoell.
Key Takeaways
- GAI Tools Are Unpredictable: Define clear indicators for solution quality—this helps determine whether the project is achieving its goals. High-risk cases cannot be automated and require human verification.
- Data Will Keep You Busy: Consider how it fits into your overall architecture, and remember to train your team. Allocate sufficient time for processing and maintaining source data, which can take up to half of the development time.
- Small Cost Streams Can Quickly Become Large Rivers: Generation is a slow and expensive process, so make sure to save results and as much metadata as possible. This enables re-evaluation and reworking of results, saving time and money.
- Start with a Commercial Model: Invest in your own model only when the project is operational and generating tangible value.
Get in touch!
Get in touch!
Choose office or contact HiQ International in Stockholm if you are in doubt.