How to turn GenAI into your secret career weapon
Are you ready to move up in your organization? Make GenAI your new best friend – and watch your career soar to new heights as you learn to harness new skills....
Artificial Intelligence
The right AI for the job: Generative vs. legacy – when to use each
If GenAI is a hammer, everything seems like a nail – but it isn’t appropriate for every task. Here’s a guide to help you decide when generative AI is most useful and when it should be avoided.
Since its launch in late 2022, generative AI (GenAI) has often been seen as the ultimate solution for many challenges. However, this isn’t always the case. If GenAI is the hammer, not everything is a nail. To make informed decisions about when to use GenAI and when to rely on traditional AI or other digital tools, it’s crucial to understand the differences between these technologies and their respective strengths and limitations.
Understanding the differences
Legacy AI, or rule-based AI, has been around for decades and operates on deterministic algorithms designed for specific tasks. This means that for any given input, it provides a consistent, predictable output using preset rules. This characteristic is crucial for tasks where accuracy and reliability are paramount.
Legacy AI also excels in environments where the cost of errors is high, such as finance and healthcare. Moreover, the decisions made by rule-based AI are transparent, repeatable, and explainable, which is essential for tasks heavy in compliance and regulation.
GenAI, in contrast, uses vast datasets to learn patterns and generate new content or data based on those patterns. This probabilistic approach allows it to produce varied outputs for similar inputs, making it highly suitable for creative tasks.
The generative aspect is what sets it apart, enabling it to create text, images, audio, and video. This capability makes GenAI a powerful tool for a wide range of applications, from content creation to customer interaction.
“The e-commerce platform Shopify has introduced Shopify Magic, an AI tool that generates compelling product descriptions in seconds, saving merchants time.”
When to use GenAI
GenAI is particularly effective for tasks that are creative and exploratory. Its ability to generate diverse and unique outputs makes it ideal for brainstorming sessions to come up with “new” ideas.
It can help marketing teams develop campaigns by generating catchy slogans or designing visually appealing graphics. Similarly, in content creation, GenAI can assist writers by producing article drafts, summarizing long documents, or creating engaging social media posts.
For example, the ecommerce platform Shopify has introduced Shopify Magic, an AI tool that generates compelling product descriptions in seconds, saving merchants time.
In customer service, the technology can handle general, open-ended queries to customer inquiries, draft polite and professional emails, and even engage in natural-sounding conversations. This makes it a valuable tool for businesses looking to enhance their customer support without significantly increasing operational costs.
Klarna’s AI assistant dealt with two-thirds of customer-service chats in its first month, doing work equivalent to 700 humans at the Swedish fintech company. The AI reduced repeat inquiries by 25% and cut resolution times from 11 to under two minutes. Available in 23 markets and 35 languages, it’s expected to boost Klarna’s profits by $40m this year.
GenAI also has applications in programming. It can assist developers by generating code snippets, suggesting improvements, and even debugging existing code. This capability helps streamline the development process, allowing programmers to focus on the more complex and strategic parts of their projects.
Moreover, GenAI serves as an excellent tool for learning and development. Language learners, for instance, can benefit from conversational practice, where the AI provides immediate feedback and suggestions. This can help improve fluency quickly.
“GenAI is inefficient in power usage, consuming vast computing power and generating significant heat.”
When to avoid GenAI
Despite its impressive capabilities, this wildly popular tool has limitations – which make it unsuitable for certain tasks. In scenarios where accuracy is critical, such as in financial forecasting or medical diagnosis, GenAI’s probabilistic nature can be a drawback.
So, when given an input or prompt, it doesn’t produce a single, fixed response. Instead, it calculates the probabilities of various possible continuations based on its training data, and then generates a response that can vary each time the same prompt is provided.
The potential for generating incorrect or misleading information, often referred to as “hallucinations”, makes it risky for high-stakes tasks. In such cases, the deterministic and precise nature of rule-based AI is more appropriate.
In regulated industries where compliance and explainability are crucial, traditional AI’s transparent decision-making process is invaluable. For example, in legal services or regulatory compliance, the ability to trace and justify each decision is essential. Rule-based AI provides the necessary clarity and consistency, ensuring that decisions are statistically accurate, defensible, and repeatable.
Additionally, GenAI is inefficient in power usage, consuming vast computing power and generating significant heat. Cooling data centers for these models requires large water consumption. Experts predict AI demand will increase water withdrawal to 4.2–6.6 billion cubic meters by 2027, equal to about half the UK’s annual consumption, making them unsuitable for sustainability-focused projects, or in tasks where energy consumption needs to be minimized.
By understanding the strengths and limitations of both generative and rule-based AI, organizations can make informed decisions about which technology to employ, ensuring they achieve the best possible outcomes for their specific needs.
Making the right choice
To determine whether GenAI is suitable for a particular task in a professional setting, organizations should start by clearly defining their objectives or use cases. Rather than focusing on the technology itself, it’s essential to consider the specific requirements and constraints of the task. A decision tree can help guide this process:
- Objective analysis: What is the primary goal of the task? If it involves creativity, idea generation, or handling general queries, GenAI might be the right choice. If the task requires precision, accuracy, and explainability, rule-based AI may be more appropriate.
- Error tolerance: Consider the cost of errors. If the task can tolerate some level of inaccuracy without significant consequences, GenAI can be useful. However, if the cost of getting something wrong is high, such as in financial transactions or medical diagnoses, rule-based AI is safer.
- Task specificity: Assess whether the task is general or specific. GenAI performs well with broad, open-ended tasks, while rule-based AI excels in narrow, well-defined ones.
- Ethical considerations: In areas that are prone to misinformation and bias, the application of GenAI should be considered with caution. While it may help to have GenAI’s input, this is an area where human judgement is preferred over any given AI technology.
- Regulatory requirements: Examine any regulatory or compliance requirements. If the task involves strict regulations and the need for explainability, rule-based AI is the better option.
- Resource efficiency: Evaluate the importance of resource efficiency. If sustainability and power consumption are critical factors, consider the vast computational demands of GenAI.
Ultimately, GenAI offers organizations exceptional capabilities for creative, general-purpose, and open-ended tasks where some errors are acceptable, making it a valuable tool for many organizations. However, its probabilistic nature and potential for inaccuracies mean it should be used cautiously in high-stakes or regulated environments where precision and explainability are essential, or where resource efficiency is important.
By understanding the strengths and limitations of both generative and rule-based AI, organizations can make informed decisions about which technology to employ, ensuring they achieve the best possible outcomes for their specific needs.
Often, this involves using both legacy AI and GenAI together in the same business context. Instead of viewing them as a one-size-fits-all solution, they should be seen as distinct tools optimized for different purposes. Understanding the specific business problem to be solved is essential for selecting the most appropriate tool for the task.
All views expressed herein are those of the author and have been specifically developed and published in accordance with the principles of academic freedom. As such, such views are not necessarily held or endorsed by TONOMUS or its affiliates.
Authors
Related
How are regulators shaping the use of corporate data in the AI era?
Today, advancements in artificial intelligence (AI), particularly large language models (LLMs), are changing the value proposition by potentially transforming mountains of unstructured data into actionable insights. ...
What’s happening with GenAI in China?
Global leaders should know where China stands with generative AI and what strategic actions make sense with an eye to the future. ...
AI compliance is a strategy problem
A patchwork of emerging AI rules is raising companies’ compliance risk. Ready? Here’s how to think strategically about new risks now. ...
Learn Brain Circuits
Join us for daily exercises focusing on issues from team building to developing an actionable sustainability plan to personal development. Go on - they only take five minutes.
Explore Leadership
What makes a great leader? Do you need charisma? How do you inspire your team? Our experts offer actionable insights through first-person narratives, behind-the-scenes interviews and The Help Desk.
Join Membership
Log in here to join in the conversation with the I by IMD community. Your subscription grants you access to the quarterly magazine plus daily articles, videos, podcasts and learning exercises.