Table of Contents
- The Value of Evolutionary AI Implementation in Finance
- Exaggerated KPIs and the Transition from the “AI Hype” Cycle to “AI Winter”
- How to Approach AI Implementation Realistically?
The Value of Evolutionary AI Implementation in Finance
In the first article of this series, we demonstrated that generative artificial intelligence is not just a passing market trend. It’s a technology that demands a mature business and ethical approach, not just because of the new obligations introduced by the EU Artificial Intelligence Act, but above all due to technology providers’ responsibility toward their clients and the duty organizations have to the end users of AI-powered solutions.
We also emphasized the potential consequences of inflating expectations toward the success of GenAI-driven implementations, citing statistics from some of the industry audits holding a promise that clients can expect a 40–50% productivity increase. A much more realistic forecast is 10–20%, which is still a solid outcome.
Bringing this result into reality may work to the software provider’s disadvantage, as their competitors often promise greater benefits. That’s why full transparency from the very beginning of gathering business requirements and at every stage of collaborating with the client is so important, so that banks have a realistic overview of the expected return on investment. Of course, it’s far better to be upfront than to have to explain unmet promises and unfulfilled agreements later. That’s why, among other reasons, we decided to address the topic of managing expectations regarding GenAI implementations in the finance and banking sectors.
Exaggerated KPIs and the Transition from the “AI Hype” Cycle to “AI Winter”
Stephen Brobst, Chief Technology Officer at Ab Initio Software, points out that when it comes to technology adoption, we are currently in the third wave of interest in artificial intelligence. So why did the first two waves end? Because of overblown promises. The market tends to follow a cycle that begins with “AI hype”, only to be followed by an “AI winter”—a phase when funding dries up and interest in implementation wanes, as companies move from initial excitement to disillusionment after unrealistic expectations go unmet.
Don’t overpromise, just overdeliver. In my opinion, this is a much better strategy. Ab Initio is an engineering company, so we don’t focus on false advertising and aren’t prone to exaggerating expectations. Delivering proof of value and proof of concept before executing a large project is far more preferable.
Based on the experience of Ailleron experts, to successfully complete a Proof of Concept (PoC), technology providers—in alignment with their clients—should focus on a solution scope that clearly reduces the time employees spend on specific tasks. It is crucial that end users experience tangible benefits at this stage, so they begin to rely on and trust the solution enough to consider broader adoption. Otherwise, PoC testing risks being perceived as an extra burden rather than a tool that facilitates daily work.
How to Approach AI Implementation Realistically?
The scenario of a prolonged “AI winter” is an extreme, unlikely case. Brobst reassures that the technology market may face a cold front, where clients will realize that providers promised them too much, leading to a decline from the peak of AI interest due to inflated expectations about the results that could be achieved.
Nevertheless, numerous success stories and evidence from industry leaders already confirm that optimizing entire business processes with AI delivers significant value. Additionally, there is ample proof that end users genuinely benefit from these tools, finding them to be a substantial aid. While software vendors and consultants will likely continue exaggerating forecasts of potential gains, we at least now know how to navigate inflated market expectations cautiously.
In the third article of this series, we’ll take a step back from the GenAI hype and explore the broader opportunities that artificial intelligence and machine learning bring to banking.
In our next article, you’ll find out more about:
- Less obvious applications and the role of traditional ML
- Use cases going beyond text generation
- Prompt engineering, fine-tuning, and retrieval-augmented generation (RAG)
- Choosing the right technological solution for a specific business case
Do you need expert support to define the right AI deployment strategy for your business?
Learn more
