This is more than an abstract question. It's a real-world issue affecting everyone, from engineers to business leaders to decision-makers. It's about understanding that knowledge and certainty can be elusive and that adopting a more flexible mindset - one rooted in humility - can lead to better outcomes.

The Gettier Problem: A New Lens on What It Means to "Know" Something

At first glance, the concept of knowledge seems straightforward: to know something, we need three things—a justified true belief (JTB). For centuries, philosophers have agreed that to know something, you need:

1. Justification: You have good reasons or evidence for your belief.

2. Truth: The belief you hold must align with reality.

3. Belief: You must actually hold that thought to be true.

This framework is familiar to many of us, whether we realize it or not. In business, for example, we justify our decisions with data, and when those decisions lead to successful outcomes, we assume we knew the right answer all along. But in 1963, philosopher Edmund Gettier introduced a challenge to this view of knowledge, revealing that even when all three conditions are met, we might still be wrong.

Gettier showed that it's possible to hold a justified true belief and still not truly know something because of hidden factors or coincidences that distort our understanding. These are Gettier-like situations, where we appear to be right but only by accident. This forces us to reconsider what it means to really "know" something and introduces a critical lesson for anyone making decisions in complex environments.

Real-World Application: Gettier-like Situations in Decision-Making

Gettier's insight is not just an academic exercise - it plays out in real-life situations, particularly in decision-making roles. Imagine this scenario:

You've just overseen a major product launch. Your team analyzed the market, gathered data, and made a strategic decision to emphasize a specific feature, believing it would drive new customer signups. After the launch, there was a sharp spike in new users. At first glance, it looks like your decision was spot-on. But then, you discover something unexpected: a big competitor experienced a service outage at the same time, driving users to your product by coincidence.

In such a case, your belief that the feature would drive signups was justified. It aligned with the outcome, making it true. But the real reason for the success was external, something you didn't predict. This is a Gettier-like situation - you were right, but not for the reasons you thought. These moments remind us that even well-founded decisions can be influenced by unseen factors, and they highlight the importance of staying open to new information and feedback.

The Pitfalls of Certainty: Why "Strong Opinions, Loosely Held" Matters

When faced with uncertainty, many people turn to the idea of "strong opinions, loosely held". This concept suggests that we should form strong, well-reasoned opinions but be willing to change them when new evidence emerges. It's a sound philosophy, but it can also be misunderstood.

Some take this phrase to mean that opinions can be formed quickly and dropped or shifted at the first sign of doubt. However, this approach can lead to superficial decision-making, where opinions are adopted without deep thought and changed without sufficient reason. In reality, strong opinions should be built on solid evidence, reflection, and analysis—what we might call justified true beliefs.

At the same time, holding opinions loosely means recognizing that even strong foundations can be flawed. The possibility of Gettier-like situations means we should remain humble, open to new data, and willing to adjust our thinking when necessary. This is the balance that "strong opinions, loosely held" seeks to strike: confidence without stubbornness, decisiveness without rigidity.

Navigating Uncertainty: Building Humility into Decision-Making

In both work and life, decision-making often involves acting with incomplete information. Knowing everything upfront is impossible, but we can still make better choices by embracing epistemic humility, recognizing that our knowledge is always limited and that we could be wrong.

Here's how to apply this mindset:

• Start with a Strong Foundation (JTB): Justified true beliefs offer a practical guide for forming strong opinions. Before reaching a conclusion, ensure it's built on solid evidence, whether you're deciding on a business strategy, a technical solution, or a personal choice. But remember that, as Gettier demonstrated, even this foundation isn't foolproof.

• Stay Open to Contradictions (Gettier): Once you've formed an opinion, remain open to new evidence or perspectives that might challenge it. Be ready to revisit your beliefs if circumstances change or if new data emerges that reveals hidden variables.

• Encourage Open Dialogue (Strong Opinions, Loosely Held): Collaboration and feedback are vital to avoiding blind spots. Inviting others to challenge your assumptions and provide alternative viewpoints creates a richer decision-making environment. This is also where the "loosely held" part of the equation comes into play: Strong opinions are necessary to drive action, but openness to change ensures they're flexible enough to adapt when needed.

This approach isn't just for leaders - it's for anyone making decisions, whether in engineering, product management, or daily life. Epistemic humility helps us navigate the unknown by keeping us open to learning and improvement, even when we think we're certain.

Conclusion: Leading with Humility and Confidence

The Gettier problem and the concept of "strong opinions, loosely held" teach us an important lesson about the nature of knowledge and decision-making. In a complex and unpredictable world, even the most well-reasoned decisions can be affected by factors beyond our control. By building a solid foundation for our beliefs, staying open to new evidence, and fostering open dialogue, we can navigate uncertainty with both humility and confidence.

Whether you're leading a team, writing code, or simply making everyday decisions, recognizing the limits of your knowledge is a strength, not a weakness. It allows you to adapt, learn, and grow - qualities that make you a better decision-maker, a better leader, and a better collaborator.

In my journey with generative AI, I've seen firsthand how our words can shape AI responses in surprising and delightful ways. It's as if we're all becoming storytellers, crafting unique experiences through interactions with these language models. The technical barriers that once limited this field have fallen away, opening up a playground of possibilities for anyone with curiosity and ideas.

But what does this mean for us, whether we're tech enthusiasts, business professionals, or simply curious individuals? It's about finding new ways to express our ideas and approach problems. Interestingly, even if you've never written a script in your life, thinking like a screenwriter might be the key to unlocking the full potential of these AI interactions. Let's explore how this new frontier changes the game and why it's worth getting excited about.

The Evolution of AI Personalities

Remember when chatbots were nothing more than glorified FAQ systems? I do. They were clunky, often frustrating, and about as personable as a vending machine. But even then, some visionaries saw the potential for something more.

Years ago, I stumbled upon an article that fascinated me. It described how some companies were hiring novelists to craft personalities for their chatbots. Can you imagine writers creating backstories, quirks, and habits for digital entities? In 2019, the idea gained traction, and tangible examples emerged, such as TD Bank Group's chatbot, Clari. The bank took an innovative approach by infusing Clari with a "friendly" and "approachable" personality crafted by TV screenwriters [2]. This marked a shift from just answering common user questions to creating a whole new customer experience. This approach was groundbreaking then, hinting at the potential for more engaging AI interactions.

Back then, the idea seemed almost whimsical - a creative luxury few could afford, but it planted a seed of possibility in my mind. Today, this concept is no longer a distant dream. It's a reality where the creative and technical barriers have all but disappeared. The ChatGPT moment far surpassed those early experiments. Large Language Models (LLMs) are not just chatbots with pre-programmed responses or carefully crafted personalities. This technology has simplified the process of creating valuable and engaging AI. We all get to be the screenwriters now.

The Art of Prompting

Think about your favorite movie character for a moment. What makes them memorable? Is it their witty one-liners? Their unique perspective on the world? Their backstory and how they, therefore, view and approach challenges? Now, imagine you could create a character like that, not for a film, but for an AI interaction with a specific purpose.

That's essentially what we're doing when we craft prompts for LLMs. We're not just asking questions or giving instructions – we're setting the stage, defining the character, and initiating a dialogue. It's screenwriting but with an interactive twist because we cannot predict the whole movie, just how it starts.

When I first read through Claude's system prompts, I was struck by how much they reminded me of character notes for a script. "Claude is very smart and intellectually curious. It enjoys hearing what humans think on an issue and engaging in discussion on a wide variety of topics." These are more than just technical guidelines - they're elements of character development. Just as a screenwriter or novelist builds a character's persona through small, deliberate choices, these prompts shape how Claude responds to the myriad situations it encounters. This ensures Claude remains consistent, reliable, and engaging, much like a well-written character in an ongoing series.

In this sense, crafting prompts is a lot like writing a script. We provide the AI with a framework - a set of behaviors, tones, and preferences that guide its responses. But because this script is interactive, we, as users, are also part of it. Our prompts shape the AI's responses, and those responses, in turn, influence the prompts that follow. It's a dynamic, creative dance between humans and machines.

Want to brainstorm with an AI that thinks like your favorite detective? Craft a prompt that sets that tone. Need a writing partner with a flair for the dramatic? You can create that experience. The AI becomes a canvas for our creativity, responding and adapting to the scenarios we conceive. Who says technical FAQs always need to be boring? Maybe, instead of (politely) telling people to RTFM [3], I could point them to a sarcastically pre-prompted LLM doing the bidding for me. The possibilities are limitless, as some people (and many instruction fine-tuned LLMs) like to say.

Embracing the Screenwriter's Mindset

The art of crafting AI interactions can be much like screenwriting. It's about creating characters, setting scenes, and guiding narratives - all through the power of your prompts. You don't need a Hollywood background to excel at this; your creativity and willingness to experiment are far more valuable. Let's dive into how you can apply a screenwriter's approach to your AI interactions, enhancing their depth, relevance, and impact.

Start with a Vision

Every great script begins with a vision. Before you type a single word, ask yourself: What kind of interaction do I want to create? Are you looking for a brainstorming session with a witty AI companion? A deep dive into complex topics with a wise guide? Or perhaps a playful exchange with a charmingly confused time traveler from the past?

Defining this vision is your first act. It sets the tone, guides your prompts, and lays the foundation for a rich, engaging interaction. Remember, you're not just inputting a query; you're creating a whole experience.

Layer in Nuance

Now, let's add some depth to our character. In screenwriting, the most memorable characters are multi-dimensional. The same principle applies here. Consider tone, context, and perspective. Is your AI detective world-weary and cynical or enthusiastic and idealistic? Is your philosopher from Ancient Greece or the distant future?

These layers of nuance can transform a mundane query into a fascinating dialogue. Try asking the same question to different 'characters' and watch how the responses evolve. It's like directing different actors to play the same role - each brings something unique to the performance.

Quick Iterations: Your Rehearsal Stage

Here's where our screenwriting analogy takes an improvisational turn. Unlike a fixed script, your AI interaction is a dynamic, evolving performance. Don't get caught up in crafting the perfect prompt. Instead, think of each interaction as a rehearsal.

Start with a rough idea, test it out, and refine it based on the AI's response. Did the tone come across as intended? Was the context clear? Did you get the information you needed? Each iteration is a chance to hone your prompt, tighten your dialogue, and shape the interaction.

The Director's Cut

As you become more comfortable with this process, you'll find yourself naturally slipping into the director's chair. You'll learn to cut what doesn't work, emphasize what does, and shape the experience to your needs.

Sometimes, the most interesting moments come from unexpected plot twists. Feel free to throw in a surprise element or challenge. Ask your AI detective to solve a case... on Mars. Request a philosophy lesson from Aristotle... about TikTok. These creative leaps can lead to fascinating insights and novel ideas.

Breaking the Fourth Wall

Finally, remember that you're not just the screenwriter and director - you're also part of the cast. Feel free to break the fourth wall. Ask the AI about its role in your interaction. Discuss the nature of your dialogue. This meta-approach can lead to more tailored responses and a deeper understanding of how to craft effective prompts.

Conclusion: Your Stage Awaits

As we wrap up our journey from the early days of clunky chatbots to the rich, interactive world of generative AI, it's clear that we're standing on the brink of a new era in human-machine interaction. The stage is set, the spotlight is on, and the script? Well, that's up to you.

The democratization of AI means that this creative power is no longer limited to tech giants or AI specialists. It's in your hands. Whether you're a business professional looking to innovate, a writer seeking inspiration, or simply a curious individual exploring new possibilities, the AI stage is yours to direct.

So, embrace your inner screenwriter. Start with a vision, layer in nuance, embrace the unexpected, iterate quickly, and don't be afraid to break the fourth wall. Remember, in this new world of AI interaction, you're not just the writer – you're the director, the actor, and the audience all at once.

As you step into this role, ask yourself: What kind of stories will you create? What worlds will you build? What characters will you bring to life? The possibilities are as limitless as your imagination.

Now, lights, camera, action! Your AI adventure awaits.

References

[1] Anthropic System Prompts

[2] How TV Screenwriters Are Giving TD’s New Chatbot an Almost Human Charm

[3] RTFM - Wikipedia

This article explores how randomness and controlled "hallucinations" in AI can drive significant business value. From efficient algorithms like CVM to creative outputs in generative AI, embracing imperfections can lead to innovation and strategic advantages. AI's "good enough" solutions save time and inspire new ideas, enhancing human creativity and decision-making. By leveraging AI's randomness, businesses can uncover hidden opportunities and achieve faster, more effective outcomes. Sometimes, it's not about perfect accuracy but using AI's quirks to our advantage.

As I work extensively with generative AI, I'm sometimes challenged with balancing precision and innovation. Recently, a breakthrough in counting algorithms, leveraging randomness for efficient solutions, sparked a thought: Could similar principles apply to AI, especially for managing hallucinations? This article explores how principles of randomness in computational systems, like counting algorithms, can be applied to generative AI to create substantial business value even when perfect accuracy isn't achieved.

To illustrate this, consider a recent article [1] about counting words in large data sets or texts like Hamlet. Imagine trying to count the unique words in Shakespeare's play. Traditional methods might struggle with efficiency, such as scanning the entire text and storing each word. Instead, computer scientists developed a new algorithm using randomness to make an educated guess, significantly reducing required resources. This concept of "good enough" accuracy through controlled randomness can be incredibly powerful.

The Power of Randomness in Algorithms

The new CVM algorithm, developed by and named after its creators (Chakraborty, Variyam, and Meel), efficiently uses randomness to estimate the number of unique elements in a data stream. By progressively filtering and keeping track of a small number of elements, this algorithm achieves accuracy with minimal memory usage. While not perfectly accurate, it provides sufficient precision for many practical purposes. This efficiency through controlled randomness is a valuable lesson for generative AI, where probabilistic methods are used to generate content.

The CVM algorithm's strength lies in its simplicity and effectiveness in handling large datasets with limited resources. By embracing the inherent uncertainty of randomness, the algorithm solves a longstanding problem in computer science with a novel approach. This innovation highlights the potential of probabilistic methods to tackle complex problems efficiently, suggesting broader applicability beyond counting algorithms.

Generative AI and Hallucinations

In generative AI, hallucinations - outputs that are factually incorrect or unfaithful to the source material - are common. These hallucinations stem from the probabilistic nature of Large Language Models (LLMs), which, like the CVM algorithm, use randomness to generate content. While this can lead to inaccuracies, it also enables creativity and contextual richness that deterministic models might miss.

Hallucinations in LLMs can be broadly categorized into intrinsic and extrinsic types. Intrinsic hallucinations occur when the generated output contradicts the source content. For example, if an AI describes a story where "The cat chased the mouse", and instead generates, "The mouse chased the cat", it introduces an intrinsic hallucination. Extrinsic hallucinations introduce information not present in the source material. For instance, if the AI generates, "The cat and mouse stopped for a tea party" when no such event was mentioned in the original story, it would be an extrinsic hallucination.

While hallucinations are often viewed as flaws, they are a byproduct of LLMs' design. As Andrej Karpathy, a prominent figure in AI, suggests, LLMs are "dream machines" weaving words together based on their training data and prompts. These models generate text by predicting the next word in a sequence based on probabilities, resulting in creative and contextually rich outputs, although sometimes factually inaccurate.

Karpathy's view emphasizes that hallucinations are not merely bugs but an integral part of how LLMs function. This perspective encourages us to see beyond the flaws and recognize the potential benefits of LLMs' creative outputs. In many cases, the balance between creativity and accuracy achieved by these models can be highly valuable, especially where perfect precision is not critical.

When "Good Enough" is Good Enough

Throughout my career, I've seen numerous scenarios where perfect accuracy isn't necessary but optional. For example, in brainstorming sessions or preliminary data analysis, getting close enough to the correct answer quickly is often more valuable than spending time on perfect accuracy.

LLM-based coding assistants are another great example. These tools propose solutions and ideas that may not always be perfectly accurate but can inspire new approaches and save significant time in the development process. For instance, a coding assistant might suggest a snippet of code to automate a task. While the proposed code may need adjustments, it can provide a valuable starting point and spark ideas the user might have yet to consider.

The Microsoft Work Trend Index [2] highlights that 70% of early Copilot users reported increased productivity, and 68% saw improved work quality, illustrating that "good enough" solutions can drive significant value. According to the report, users found that Copilot helped them get to a good first draft faster, saved time on mundane tasks, and improved overall productivity and creativity. This data underscores the practical benefits of accepting "good enough" solutions in the workplace.

Such tools, infused by GenAI, can help users by providing initial drafts that can be refined and polished with human oversight. This approach not only accelerates workflows but also fosters innovation by allowing humans to focus on more critical and creative aspects of their work. By leveraging AI to handle routine tasks and provide a starting point for more complex problems, businesses can free up people to focus on strategic and creative endeavors.

This concept extends beyond coding to various business applications. For instance, in marketing, an AI-generated draft for a campaign can provide a framework that marketers can build upon, saving time and allowing for creative refinements. In customer service, AI can generate summaries and initial responses to queries, which human agents can then customize to ensure accuracy and personalization.

The notion of "good enough" isn't about settling for mediocrity; it's about understanding that in many business contexts, approximate solutions can pave the way for faster, more innovative outcomes. Accepting this can lead to substantial productivity gains and open new avenues for creativity and problem-solving.

Randomness as a Strategic and Creative Force

Randomness can be a powerful driver of innovation and strategic value. Just as the CVM algorithm uses randomness to solve counting problems efficiently, LLMs leverage it to generate diverse and creative outputs. This randomness not only facilitates problem-solving in unexpected ways but also opens new avenues for strategic thinking and business planning.

However, the real magic happens when AI's creative randomness is harnessed by human intelligence. Beyond creative contexts, GenAI can simulate different scenarios and potential outcomes, helping businesses prepare for various possibilities and make more informed decisions. For example, AI might generate several market entry strategies, each considering different economic conditions or competitor actions, prompting business leaders to explore options they might not have otherwise considered. AI-generated content or strategic scenarios serve as starting points, offering fresh perspectives and innovative ideas.

This concept is not just about generating random outputs but about using these outputs as valuable inputs in a human-driven decision-making process. It's about recognizing that AI, with its inherent randomness, can get us close enough to an answer we can refine and improve upon. It's not about replacing human judgment but augmenting it with AI's ability to explore possibilities quickly and efficiently. It's about making the most of what AI has to offer (for now) and using it to our advantage.

Embracing Hallucinations for Business Value

Generative AI often gets a bad reputation because of hallucinations, but there's a silver lining if we look closer. Many times, "good enough" solutions can be incredibly valuable. AI might not always hit the bullseye, but it can get us close enough for humans to take it the rest of the way, saving time and sparking creativity in the process.

Think about it: AI can handle the grunt work, letting us focus on the finer details. This partnership between AI's broad-stroke approach and our precision can drive innovation and efficiency in ways we hadn't imagined. By turning AI's quirks into opportunities, we can discover hidden gems that propel our businesses forward.

Randomness and hallucinations in AI aren't just quirks—they're powerful tools for innovation. Whether it's solving problems efficiently or generating creative and strategic outputs, the value lies in using these AI-generated ideas as springboards. It's not always about perfect accuracy; sometimes, it's about seeing the potential in "good enough" solutions and letting AI augment our creativity and decision-making.

What if AI throws out a wild idea that seems off at first, but upon closer inspection, it sparks a new direction you hadn't considered? This blend of AI's randomness and human ingenuity can lead to breakthroughs. It's about leveraging AI to get us close enough, refining those outputs, and turning what might seem like a flaw into a powerful advantage.

Take the Hamlet example from earlier: Knowing it has precisely 3,967 unique words might be interesting, but knowing the number is nearly 4,000 is often good enough. It's the bigger picture that counts, and AI can help paint it quickly and effectively.

So, what business applications can benefit from "good enough" accuracy? Reflect on your own experiences. Sometimes, a bit of randomness and imperfection can be the catalyst for innovation and efficiency in your work. Personally, I find more and more use cases where this turns out to be true - confirmation that embracing a bit of quirkiness can lead to unexpected and valuable outcomes.

References

[1] Computer Scientists Invent an Efficient New Way to Count

[2] Microsoft 2023 Work Trend Index Special Report

"I don't know what still newer marvels will make writing twice as easy in the next 30 years. But I do know they won't make writing twice as good. That will still require plain old hard thinking [...]." - On Writing Well by William Zinsser

Natural language is an omnipresent interface. We use it in our lives every single day, and we've yearned to use it for technology in the same way. Large Language Models are the bridging stone getting us closer to that wish. Yet, many of us, including myself, are not proficient in using this interface particularly well. Luckily, society has had many years to reflect on this topic.

While traveling to Japan, I started reading William Zinsser's On Writing Well. The introduction made me pause. In it, he describes how the arrival of word processors triggered two opposing developments: Good writers got better and bad writers got worse. It was not because words changed. Instead, by lowering the barrier to writing, the new technology introduced a non-obvious catch. In his own words:

"Nobody told all the new computer writers that the essence of writing is rewriting. Just because they're writing fluently doesn't mean they're writing well."

Writing is not getting clearer by adding an increasing amount of words to it. Good writers know they need to experiment, revise and reshape their creation until every element serves a purpose. Computer screens avoided the need for retyping, making them faster and better.

Something similar can be observed with the emergence of low-code and no-code tools. Creating applications or any automated business logic became easier than ever to transform ideas into working solutions. The number of deployed automation increased as the barrier of entry drastically decreased. It is no longer required to study different types, methods, and algorithms, as well as the strengths and weaknesses of the language in which they've been embedded. You can simply start building. But what is true for writing is also true for software development: Just because they're developing fluently doesn't mean they're developing well.

Well-crafted software is not just a combination of individual features stitched together. It is an enabler for scalable value creation. Therefore, with low-code and no-code platforms rising, the same thing happened again: Good developers got better and bad developers got worse. Low-code does not mean either the problem or the solution is low-complexity. Often, these platforms lower required implementation efforts, allowing good developers to focus more on experimentation, refinement, and improvement - until every element serves a purpose.

In contrast, other developers stop working on a functionality once they see the expected results. Why bother with it any longer if we can add more logic to it at any given time? The user experience, however, will differ with both approaches, similar to how some texts or books are more enjoyable than others.

Low-code does not mean either the problem or the solution is low-complexity.

Considering both examples, there might be a pattern that has less to do with the advancement of technology but rather how we humans react to them. Generative AI will again be one of these advancements that will reduce the barriers to creating: Be it content, applications, or, more broadly, experiences. As history often rhymes, I'm willing to bet that two things will happen again: Good creators will get better and bad creators will get worse.

Creating any type of experience will be simpler than ever before. Chain any number of words together, and you will be given output in your desired medium. However, while some people will use this to rapidly experiment, iterate and revisit how they create novel experiences, others will chase the latest blueprint or mega-prompt instead. They will apply them to recreate different variations of the same, over and over again.

What do I expect to be the differentiation factors between those two groups: The ability to form and formulate clear thoughts. The combination of strong thinking and good communication will remain the craft to hone, with Generative AI being the lever to provide unprecedented levels of impact.

The remaining question is: Do you want to treat it as a craft to be mastered or a tool to be used only when needed?

AI-powered email analysis with ChatGPT revolutionizes customer service by effortlessly assessing language, sentiment, and urgency, leading to faster responses and happier customers. However, be mindful of potential security risks and data protection concerns when using third-party APIs.

Unleashing ChatGPT to improve Customer Support

Are you curious about how Large Language Models (LLMs) like ChatGPT can revolutionize customer service and support? Allow me to share a scenario with you that demonstrates the potential!

Imagine this - you're a support agent, and you're bombarded with a ton of emails every day. Some of them are complaints, some are queries, and some are just random chit-chat. How can you prioritize the most urgent issues or efficiently direct them to a colleague who speaks the same language? Better yet, how can you obtain this information effortlessly without any manual intervention? This is where AI come into play.

Incorporating the power of LLMs into the process of handling customer complaints can significantly streamline the management of incoming emails. By utilizing AI's ability to effortlessly analyze language, sentiment, and urgency in real-world complaint emails, we can prioritize cases and responses more effectively. This enables quicker response times to high-priority issues, which in turn leads to improvements in overall customer satisfaction.

I experimented with ChatGPT by providing it with a real-world complaint email for analysis. The results I obtained from a single prompt were fascinating. ChatGPT successfully detected the email's language, identified the sentiment and urgency of the content, and provided a concise one-sentence summary. Allow me to demonstrate how you can achieve this too!

How to Craft a GPT Prompt for Analyzing Inbound Emails

To enable quick experimentation, I prefer using a Jupyter Notebook to make requests to the OpenAI API. First, import the openai package, provide your API key, and define a simple function to call the API:

import openai
openai.api_key  = 'sk-Your-API-Key'

def get_simple_completion(prompt, model="gpt-3.5-turbo", temperature=0):
    messages = [{"role": "user", "content": prompt}]
    response = openai.ChatCompletion.create(
        model = model,
        messages = messages,
        temperature = temperature, # degree of randomness
    )
    return response.choices[0].message["content"]

Next, we need to obtain the contents of an email. To simplify this example, let's assume we have successfully extracted the subject and body of an email for further processing. To make things more interesting, I have chosen an example from the Consumer Complaint Database published by the US Consumer Financial Protection Bureau and translated it into German. Our extracted email reads as follows:

email_subject = "Problem mit einem Kauf auf Ihrer Abrechnung"
email_body = "Ich habe eine Belastung auf meiner Kreditkarte für eine Kreditkartenkontonummer, die mit XXXX endet, gemeldet. Ich hatte Sie angerufen und Sie haben mir versichert, dass Sie die Belastung entfernen werden. Allerdings haben Sie mein Konto geschlossen und die Belastung nicht entfernt, obwohl Sie gesagt haben, dass sie es tun würden. Mein Kredit-Score wird jeden Tag schlechter, weil die Belastung immer noch da ist. Ich habe sie mehrmals kontaktiert und sie haben mein Problem immer noch nicht gelöst."

Now for the most interesting part: the actual prompt. We will follow a simple prompt structure by first defining the task at hand, which is to identify the language, sentiment, and urgency of the email. Additionally, we will request a one-sentence summary in English, which will be useful for further processing tasks. From here, we simply add the email's subject and body as input and ask for the output as a JSON. All that remains is to pass this prompt into our previously defined function and display the response.

prompt = """
Task:
Analyze the email subject and body provided below for language, sentiment, and urgency. Also, provide the 1 sentence summary in English. Limit your response by only outputting information in the following structure as JSON:
Email Language: <Detected Language>
Email Sentiment: <Detected Sentiment>
Email Urgency: <Detected Urgency>
Email Summary: <1 Sentence Summary in English>

Input:
Email Subject: '{email_subject}'
Email Body: '{email_body}'

JSON Output:
""".format(email_subject=email_subject, email_body=email_body)

response = get_simple_completion(prompt)
print(response)

With this simple script, we have now accomplished multiple tasks at once: We transformed unstructured information from an email into structured information to assist with the initial assessment of an incoming case. We detected the language, allowing for automated routing. We identified the sentiment and urgency, helping us to better prioritize the task. We received a summary in English, independent of the original language used, providing us with additional opportunities to better analyze, track, and compare this request to others. All of this, with just one call, giving us the following result:

{
  "Email Language": "German",
  "Email Sentiment": "Negative",
  "Email Urgency": "High",
  "Email Summary": "Customer reports a problem with a purchase on their billing statement, stating that the charge was not removed as promised and their credit score is being negatively impacted."
}

Closing thoughts

This simple demonstration illustrates how incorporating large language models like ChatGPT into customer service processes can significantly enhance the management of inbound emails. By harnessing AI's capabilities to analyze language, sentiment, and urgency, support teams can prioritize cases more effectively, resulting in quicker response times and heightened customer satisfaction. With just a single API call, multiple tasks can be executed, converting unstructured information into actionable insights and streamlining the entire support process. Is there a catch, though?

Utilizing a third-party API and providing them with potentially sensitive data can be a cause for concern. By integrating an external API into the support process, businesses may inadvertently expose their customers' private information to potential security risks. This could include personal details, financial information, or other confidential data that should be handled with the utmost care. Therefore, it is essential for companies to thoroughly evaluate the security measures and data protection policies of any third-party API provider before integrating their services. In doing so, they can ensure that customer data remains secure while still benefiting from the enhanced efficiency and effectiveness that API-driven support processes can offer.

ChatGPT and other Large Language Models can help us generate ideas for transforming our business. However, it is sometimes worthwhile to explore alternative options to achieve similar results. Nonetheless, it's fun to push the boundaries in a responsible and controlled manner!