This is a guest post by Dr Stephen Anning, Visiting Researcher in the Department of Web Science at the University of Southampton and online tutor for the MA in Artificial Intelligence.
This blog post shares insights from student discussions in the ‘Introduction to AI’ module on the University of Southampton’s online MA in Artificial Intelligence, focusing on how we move from AI hype to real world application.
Designed for non STEM students, the course explores not just how AI works, but how it can be applied responsibly and effectively in organisational contexts, bridging the gap between ambitious ideas and operational reality.
Understanding the gap between AI hype and real-world application
In this discussion, we transitioned from the technical foundations of data into the realm of imagination and creativity, exploring how AI fits into society and the workplace. Our discussion touched upon the critical balance between "blue-sky" potential and the grounded reality of business cases, ethics, and human-centricity.
This is the start of where we ground our assertions about AI in the realities of how the underlying technologies work. This is where we earn our credibility.
The cost of AI innovation and challenges for mid-sized organisations
We began by addressing a common frustration: the gap between claims about what AI can achieve and what is actually feasible. As was rightly pointed out, while we can imagine automating complex feeds, the reality often involves significant costs, both in terms of expensive developer talent and high licensing fees for tools like Co-pilot. The reward is not always greater than the cost.
We identified a "three-tier" challenge in the market. Large organisations have the scale to build secure "sandboxes" for their data, while individuals can easily experiment with open tools.
However, mid-sized companies - the cornerstone of our economy - often struggle. They are too large to risk data leakage but not large enough to easily afford the bespoke support structures required for secure, high-value AI implementation.
What is augmented intelligence and how does human machine teaming work
A key theme that emerged was the shift from Artificial Intelligence to Augmented Intelligence. We spoke about a compelling example of using AI to mark GCSE practice papers. By allowing AI to handle the 95% of mundane, descriptive tasks, humans (in this case, parents or teachers) can focus their energy on the 5% where the student needs help.
This delegation of tasks is the essence of Human-Machine Teaming. AI is exceptionally good at getting us to "good enough" or handling high-volume, repetitive tasks. As I shared from my own experience, however, attempting to use AI for master's level marking often fails at tasks requiring deep critical analysis and non-linear logic.
Remember how LLMs are trained: they are trained using question-answer techniques; they are not necessarily trained to perform non-linear analytical tasks. Our goal is to find the "sweet spot" where AI enhances human productivity rather than just chasing marginal efficiencies.
Why AI ethics and bias are critical for business and society
We had a profound discussion on bias and ethics, sparked by an observation on how data often reflects a male-centric world. We must recognise that AI doesn't just reflect existing biases, it magnifies them. Whether with medical data or hate-speech algorithms, these systems are only as good as the data they are fed and how well that data reflects the real-world scenario it seeks to model.
We also discussed commercial ethics. Using AI to set warranty thresholds, for instance, has massive financial implications. If a study is riddled with unaddressed biases, it isn't just an "ethical" failure, it’s a bad business decision.
I want to reiterate that we don't pursue responsible AI just to pursue a moral good; we do it because it’s the framework required to deliver valuable, defensible, and operationally effective products.
Key principles for building responsible and effective AI systems
As we move forward there are three principles to keep in mind:
Solve the right problem: As with our discussion regarding healthcare, don't just automate a task because you can. If a consultant is overwhelmed, perhaps they need a resource management system rather than an AI note-taker that might strip away their "thinking time."
Explainability and transparency: Avoid "black-box" solutions. Your proposals should demonstrate how the system arrives at its conclusions. In my own research on hostile narratives, I showed how we use AI to extract specific evidence for plot points, allowing a human (like a jury) to verify the logic. Explainability forces you to think about how you build an algorithm so it provides meaningful output to the users.
Test and Evaluation (T&E): Critical. If you make a claim that your AI can do X, your implementation plan must include a rigorous T&E workstream. You need to prove it works, identify where it fails, and show how you will mitigate those errors.
Find out more about the MA in Artificial Intelligence
AI is rapidly moving from experimentation to real world application across industries. Understanding how to implement these technologies effectively, responsibly, and with clear business value is becoming an essential skill for professionals in a wide range of fields.
The University of Southampton’s online MA in Artificial Intelligence is a conversion course designed for people from non STEM backgrounds. It’s aimed at those who want to understand not just how AI works, but how it can be applied in practical, ethical, and operational contexts. You don’t need prior coding experience to take part.
Explore the course
