AI’nt That Special
Crafting Customer Value with Artificial Intelligence
AI — The Great Disruptor
Generative AI is changing everything. In just a few short years, it’s completely transformed how we think about creativity, work, and innovation. From art to business, to how we solve complex problems — AI is opening doors we never knew existed.
Let me paint a picture: imagine tools that can write stories, create images, compose music, even generate video — all at the click of a button. These aren’t just abstract ideas; they’re happening right now. For today’s talk, I’m focusing on large language models — like ChatGPT — the technology behind much of this revolution in text generation.
But here’s the thing: with great power comes great complexity. On one hand, generative AI is sparking incredible innovation. It’s making industries more efficient and offering possibilities we never thought possible. On the other hand, it brings its own set of challenges — ethical concerns, social implications, and economic disruptions.
Take education, for example. AI can create personalized learning experiences, generate custom content, and even provide instant tutoring. It’s a educators dream! But, at the same time, it raises tough questions: What happens when students use AI to plagiarize? What if we become too reliant on AI for knowledge? And what about those who don’t have access to these tools — are we widening the digital divide?
Generative AI is powerful. But, like any tool, how we use it — and who benefits from it — depends on us.
The Race to Find Value — Enter the Age of the Chatbot
While we wrestle with the challenges of generative AI, the technology itself is evolving at a mind-blowing speed. Think about this: the models behind ChatGPT, Llama, and Gemini have exploded from 3.7 billion tokens to over 1.5 trillion! That’s like going from a bicycle to a rocket ship in just a few years.
And as these models grow, so do their capabilities — faster than we can sometimes keep up. It’s both exciting and daunting. How do we navigate this rapid advancement responsibly? How do we make sure we’re using this incredible power wisely, instead of letting it run away with us?
Here’s what we know for sure: generative AI isn’t going anywhere. Its impact will only get bigger.
But the real question isn’t whether AI will disrupt our industries — that’s inevitable. The real question is how we choose to harness it. Will we use it to create positive, meaningful change? Or will we let it run wild, creating more problems than it solves?
Let’s look at how some of the biggest companies in the world are applying generative AI with impressive results.
Take Duolingo, for example. They launched Duolingo Max, an AI-powered chatbot that gives learners personalized feedback and even roleplays real-life conversations to help them practice. The impact? A 54% increase in paid subscribers in just the first quarter of 2024. And 20% of their existing subscribers upgraded to this feature.
Klarna, the fintech giant, also jumped on the AI bandwagon, launching a chatbot to handle customer support. By automating repetitive tasks and analyzing customer interactions, they essentially created the work equivalent of 700 support agents. The result? A $40 million boost in profits in 2024.
Then there’s Alibaba, the Chinese e-commerce giant. They introduced a generative AI chatbot to help scale their customer support, offering personalized nudges and reminders that influenced shopper behavior. The outcome? A 25% increase in customer satisfaction and a 20% reduction in training time for their support teams.
These companies are showing us that AI isn’t just about cool tech — it’s about driving real business results.
So, with all this buzz around generative AI, it seemed obvious, right? If we wanted to keep up with this rapid trend, we needed to jump in and build our own AI-powered chatbot. After all, everyone else was doing it, so why not us?
At Go1, we set out to explore the wonders of generative AI, and what better place to start than with an AI chat experience to help our learners find relevant courses? We worked hard, excited about the potential impact this shiny new tool could have for our customers.
We launched the experiment, eagerly awaiting the results. 60,000 learning sessions later, here’s what we saw: learners who engaged with the chat bot had fewer enrollments.
We didn’t understand. McKinsey had told us that “Organisations seeing the largest returns from generative AI are more likely than others to follow a range of best practices”⁵ — and yet here we were with a slick new chatbot, clear the trending best practice — and none of these graphs showing an aggressive trend up and to the right.
The lesson? Just because AI is the latest and greatest tool doesn’t mean it’s always the right one. Plenty of companies have found success with chatbots, but they didn’t jump on the AI bandwagon without asking the most important question: ‘Is this the right solution for our customers?’
It’s like having the world’s most advanced hammer and suddenly thinking every problem looks like a nail.
The opportunity of AI — strengths and weaknesses of generative AI
Before we jumped back into generative AI, we had to pause and ask ourselves: What’s the actual problem we’re trying to solve for our customers? And where does generative AI shine, and where does it fall short?
Let’s start with where AI excels.
One of its biggest strengths is content generation. Whether it’s automating product descriptions, creating images, filling in metadata, or drafting marketing copy, AI can scale content creation at a level humans simply can’t match. It’s a game changer, especially in industries where personalization is key.
Generative AI also excels at knowledge synthesis and summarisation. They can analyse vast amounts of information, extract key insights, and distill complex concepts into concise and accessible formats. This capability streamlines the creation of learning content by automatically generating summaries, translation, and taxonomies.
Generative AI is also effective at establishing context. It can very easily identify there is a significant difference between a ‘birthday card’ and a ‘red card’. And this contextual awareness leads to more nuanced and accurate ways to power search experiences through the latest wave of semantic search technologies.
But here’s the flip side: AI isn’t perfect.
Generative AI can struggle with accuracy — having a tendency to generate information that is not grounded in reality or supported by their training data. This can manifest itself as fabricated facts or statistics, or invented references or citations — or producing educated guesses.
Despite their impressive outputs, LLMs fundamentally lack true understanding or reasoning capabilities. They will often produce educated guesses, and show they are incapable of genuine comprehension when pushed — although they’re highly effective at mimicking understanding.
And let’s not forget: AI can be slow and expensive to run, especially at scale. That means longer wait times for users, leading to frustration, and potentially creating more friction than value.
So yes, generative AI is powerful — but it’s not magic. It has strengths, and it has real limitations. The key is knowing when to use it, and when to step back and ask: Is this really solving the right problem?
There’s a lot of ways to skin a cat (but should you)
The most successful applications of AI aren’t the ones that mimic the heard. They’re the ones that align with real-world needs.
I’ve seen firsthand both the promise and challenges of leveraging AI to create value for our users. From backfilling metadata and enhancing search with semantic technologies to creating personalized learning experiences, the key takeaway is clear: success in AI lies not in jumping on trends, but in carefully aligning solutions with the actual needs of customers.
So, as you go forward in your own journeys with AI, ask yourself — what problem are you trying to solve? Once you know that, the right solution will present itself
Remember, AI isn’t just a hammer looking for a nail. It’s a tool. And it’s up to us to decide how we wield it.
References
- Jaimon. A (2023), Revolutionizing Search with AI: Diving Deep into Semantic Search, QED42
- Pinjari. R (2023), Unlocking Possibilities: Real Vector Databases Demystified!, Medium.com
- Mitchell. T (2024), What is vector search, couchbase.com
- Unknown (2024), Klarna AI assistant handles two-thirds of customer service chats in its first month, Klarna.com
- Singla. A, Sukharevsk. A (2024), The State of AI, McKinsey & Company
- PYMNTS (2024), Duolingo Credits GenAI for 54% Leap in Paid Subscribers, pymnts.com
- Abdulkadi. A (2024), Earnings call: Duolingo outlines growth strategy and Max rollout, investing.com
- https://aibusiness.com/ml/the-alibaba-challenge-how-to-effectively-engage-with-a-billion-customers-#close-modal
