Artificial intelligence can identify patterns and insights in vast datasets, aiding in investment decision-making, trend spotting, and strategic planning. To truly leverage the power of generative AI, take a good look at your data first. 

"I can hardly point to any area that isn't impacted by it. AI and algorithms are everywhere. It's truly astonishing," said Allan Peter Engsig-Karup, Research Engineer Director, SimCorp. 

With a background as an associate professor in Scientific Computing at Denmark Technological University (DTU), he isn't one to give simplistic or one-sided answers.  

However, on the topic of AI's potential, he is decidedly firm. 

"I have done research across several industrial sectors, and AI affects them all. The impact is profound," says Engsig-Karup. 

He is not the only one with this perspective on the potential of AI, especially generative AI.  

A recent study[1] from market intelligence firm Gartner places generative AI at the peak of inflated expectations in its 2023 Hype Cycle for Emerging Technologies.

Major tech companies like Microsoft, Google, Meta, AWS, and IBM have jumped into the arena[2], developing generative AI solutions tailored for the financial sector, as it’s a sector interested in new technology, especially for processing vast number of data sources or to perform number crunching analytics.

There's a risk that if you overinvest in the wrong initiatives or do not understand well enough the pros and cons of emerging technology, the returns will be minimal.

Allan Peter Engsig-Karup
Research Engineer Director, SimCorp

Having spent years examining intricate issues in computational mathematics and modern scientific computing paradigms, and linking these developments to engineering applications, Engsig-Karup is no stranger to complexity.  

His dual roles in academia and a global software company offer him unique insights into maximizing the potential of emerging technologies. 

"It turns out many companies struggle to navigate the AI landscape. There's a risk that if you overinvest in the wrong initiatives or do not understand well enough the pros and cons of emerging technology, the returns will be minimal," he observes. "In my view, it's crucial to invest in operationalizing AI as well as building expertise in your workforce that can be utilized within business domains. This way, you can better discern where AI will be most beneficial," explains Engsig-Karup. 

At SimCorp, he plays a pivotal role in crafting and executing use cases where technology can make a tangible difference, helping both employees and clients alike. 

"AI excels in areas like automation and compliance, and it's adept at discerning patterns in data. That's its core strength," he emphasizes. "The goal should be creating lasting value. There's been a buzz around ChatGPT for a while, but without a deep understanding of its capabilities, its potential remains untapped. The trick is to identify cases where large language model powered technologies such as ChatGPT is the right tool for previously insurmountable challenges," states Engsig-Karup. 

What is good data? 

The amount of data is growing everywhere, and digitization is at the top of all agendas.  

Once you have data, you can do machine learning on top of it that is specific to the problem you want to solve, but the quality of the data is crucial for creating useful and valid results. 

Therefore, there’s a few considerations you must make when selecting datasets: 

  1. Size and diversity: Ensure that the dataset is large enough to capture nuances and variations within the domain. 
  2. Quality and accuracy: Make sure the dataset is well-curated and representative for the field. Align capabilities with specific business goals. 
  3. Bias: Identify and mitigate bias in the data. If the dataset has a certain bias all output will be skewed. 
  4. Expertise: Draw upon expert domain-specific knowledge when creating the dataset. This will improve the relevance and quality of the later results. 
  5. Test and validate: Separate a section of the dataset that can be used for testing and validation of the results from the rest of the dataset. This way you can test and validate if your generative AI is able to generalize and handle new data in a good way. 
  6. Governance: bring the solutions to deployment with minimal gaps between experimentations and productization utilizing machine learning operations, securing proper governance related to maintenance, monitoring, cost-control, auditing, etc.

In the financial realm, generative AI is making headway as an instrument for risk assessment, market forecasts, and portfolio management, drawing insights from intricate financial data. 

Regarding his vision for generative AI's future, the Research Engineer Director anticipates its consistent growth, propelled by a broader digital transformation agenda. 

"Consider ChatGPT, which has emerged swiftly, heralding a revolution. It embodies generative AI's power through multi-modal learning abilities, achieving feats previously considered unattainable and on an unprecedented scale," he elucidates. 

However, realizing this vision requires a foundational shift. Unlocking generative AI's potential hinges on data. 

"To harness this technology, one must gather data or establish structured workflows prior to AI deployment. With the right data in hand, one can then apply machine learning tailored to specific problems or tasks," he notes. 

High-quality data is paramount. 

"It's of utmost importance. An increasing body of evidence suggests that significant advancements can be achieved by refining data quality. This holds true irrespective of the problem at hand," he continues. 

"Because data-driven technology is not flawless. Poor data quality leads to flawed outputs. Garbage in, garbage out. With robust data and the right expertise, outcomes can be exceptional," Engsig-Karup concludes. 

Poor data quality leads to flawed outputs. Garbage in, garbage out. With robust data and the right expertise, outcomes can be exceptional.

Allan Peter Engsig-Karup
Research Engineer Director, SimCorp

For him, AI's most commendable applications work seamlessly in the background being invisible to the user, solving challenges without drawing attention to themselves. For instance, car GPS systems. Users implicitly trust these systems to accurately chart routes and predict estimated time of arrivals, all thanks to data accrued over years through digital workflows. 

"Often, you're oblivious to AI's role in the solutions you use daily, but you trust in their efficacy. At SimCorp, we design workflows that resonate with users, overlaying them with pertinent recommendations. It's a subtle yet impactful use of AI," he adds. 

Learn more about SimCorp’s use of automation and machine learning here

4 key questions to ask before implementing generative AI to your business:

The amount of data is growing everywhere, and digitization is at the top of all agendas.

Once you have data, you can do machine learning on top of it that is specific to the problem you want to solve, but the quality of the data is crucial for creating useful and valid results.

Therefore, there’s a few considerations you must make when selecting datasets:

  1. Value proposition: How do AI solutions benefit solving the task at hand? Or a potential user of a service? 
  2. Data privacy and security: How does the solution ensure the privacy and security of data? 
  3. Customization and adaptability: Can the solution be customized to align with the specific business needs and processes? 
  4. Explainability and bias mitigation: What measures are in place to identify and mitigate potential biases in the generated content and decisions?