Posted by 
AI adoption is rising—especially Agentic AI—but most companies automate low-impact tasks instead of solving core business problems. Let’s look into the potential across software, healthcare, and finance.
SwissCognitive Guest Blogger: Terryel Hu – “The Missed Potential of AI: Finding Problems Worth Solving With Agentic AI”
Agentic AI has already been initiated across virtually every industry. There is no doubt that agentic AI—whether it is a model from Google, OpenAI, or Microsoft—can outperform the typical human on routine tasks. While Nvidia uses the phrase “digital human,” its capability would be not much different to traditional automation. Companies are eager to showcase their AI adoption. What they’re actually doing is automating the parts of their business that have perhaps the least impact. From digital receptionists to chatbot-powered call centers, the use of AI today mostly optimizes activities that don’t impact the core business.
Let’s take a closer look at how industries are deploying AI and how they have drifted from business problems.
Software Engineering
Software engineering is perhaps one industry where the impact of AI has been felt most. The idea of replacing coding tasks may even be controversial among engineers themselves. But it shouldn’t be surprising. Facebook didn’t emerge as a billion-dollar company because someone wrote better code. While Mark Zuckerberg and his initial team were talented coders, that was not the core reason he founded Facebook. It came from identifying a simple but powerful human need—connection. The original Facebook team had technical talent, but their real value was solving a social coordination problem in a digital space. It didn’t start with a programming language. It started with a purpose.
Modern software teams often forget that aspect. Many software engineers have quickly deployed AI to fast-track their tasks. Agile and Scrum have turned into checklists. Many are quick to adopt AI to improve code quality—so much so that clean code has become the end goal, rather than business performance. If AI is being used to shave off hours in version control or deployment pipelines, that’s fine, but that’s not transformation.
One can automate parts of the code while the other manages quality. But why must software engineers focus exclusively on writing code and managing infrastructure? We’re not short of code. We’re short of courage in choosing the problems worth solving. Software engineers are an interesting group. Some have gone on to found companies, solve infrastructure issues, and develop systems to improve operations. They have demonstrated to the world they can tackle other complex challenges. So why limit themselves?
Healthcare
Let us now turn to another popular industry—healthcare. Imagine walking into a clinic where you’re greeted by a digital receptionist. A screen asks you to fill in your symptoms and medical history. A chatbot guides you through your check-in. Everything is seamless. Everything is automated.
That’s the type of automation that would excite innovators. The head of strategy would report this return on investment to the CEO. They decide that more is better. While this is certainly a sign of innovation, it can also lead leaders into a peculiar trap: the belief that agentic AI must be used to replace activities.
Many will applaud the cost savings in a hospital without considering whether it improved patient care. This is the paradox of digital transformation in healthcare. Automating a receptionist is considered a success, but it does nothing to address the core business: accurate diagnoses, specialist consultations, and patient care.
If the goal is quality care, then why bring in AI as the ultimate solution? If the aim was simply to automate a questionnaire at the front desk, a survey tool could have achieved it just as easily. In fact, other solutions can do that job and are far more cost-effective. If a survey form was the requirement, there are already plenty available internally. The marketing team likely has the licenses and skills to set up such surveys. Implementing a chatbot and a survey creator both offer a potential solution.
Too often, the decision to invest in, develop, and deploy AI is based on visibility—not value. It feels modern. It looks good. But it rarely changes the outcomes that matter—like patient recovery. There’s nothing inherently wrong with automating repetitive tasks, but such investments don’t address the core problems. They demand deeper integration of AI into medical knowledge, not just administrative tasks. That’s where the real value lies. And that’s where attention should go.
Finance
In the sense of core problems, the financial services industry is no different in its approach to AI. Banks, in particular, are among the earliest adopters of digital infrastructure. But when it comes to AI, their focus has been narrow. Ask any bank executive about AI, and they’ll talk about chatbots, automated credit risk models, or contact centers. These are useful, but the benefits from AI are marginal compared to existing solutions. An analyst can already complete modelling tasks in Excel. Contact centers can be outsourced.
The areas with larger transactions remain practically untouched: institutional banking, mergers and acquisitions, and portfolio structuring. AI could be used to simulate complex investment scenarios, stress test portfolios under hundreds of assumptions, or identify acquisition targets through large-scale pattern matching. These applications would directly enhance the bank’s competitive edge. But instead, we see AI applied to the call center. It’s considered an easy win. It reduces headcount. It improves NPS scores. Though it tries to improve activities that don’t differentiate the business meaningfully.
AI should be used to amplify a company’s strengths and yet it’s applied to patch inefficient activities. Banks were not built to run call centers. They were built to create financial value. If AI doesn’t serve that core, it’s just another tool lost in the back office.
The Solution? Start with the Problem
Too many AI initiatives start with what’s easiest—not what’s essential. It’s easy to get excited by AI-generated emails or faster note-taking apps. These are driven by the belief that doing things faster is always better. They focus on surface-level processes—admin, customer service, documentation—not because those are the most impactful, but because they’re perceived as low-risk and easier to automate.
What would happen if we flipped this approach, and began from the core of the business?
It’s time to ask: What are we really here to do? For a logistics firm, it might be to ensure on-time delivery while minimizing uncertainty. For a university, it might be improving student educational outcomes while ensuring they are job-ready. AI strategy should begin from the core problems faced by a business (or government), rather than going straight to automating tasks.
Instead of retrofitting AI to old workflows, start with the core business and build solutions around it. Deciding which problems are worth solving still requires judgment. One of the powerful, and yet controversial, aspects of AI is that it can make judgement calls. AI is excellent at organizing data. When you enter a prompt, for example, it can organize volumes of data into digestible information and can even present a coherent answer. It’s increasingly capable of drawing inferences and producing explanations. What it can’t do, as yet, is identify which problems are worth solving in the first place.
AI excels at declarative knowledge— facts, summaries, and correlations. But most of the decisions that move businesses forward are not declarative. They involve group dynamics, trade-offs, and execution. Even the pioneers behind the top AI tools will have explored scenarios, tested assumptions, and generated creative solutions. That part remains a human capability. AI might tell you how to optimize a logistics network but only if someone prompts that question in the first place.
About the Author:
Terryel Hu is a founder, advisor, and innovator recognized by Thinkers360 as a Top 100 Leader in Management. He is dedicated to helping businesses bridge capability gaps, and raise awareness about AI strategy.
Der Beitrag The Missed Potential of AI: Finding Problems Worth Solving With Agentic AI erschien zuerst auf SwissCognitive | AI Ventures, Advisory & Research.
