Copy link to article


The quickstart guide to generative AI for business leaders

Written by the team at Datasparq.

Is generative AI a threat, or an opportunity? We sat down with John Wyllie, Managing Director at Datasparq, to get his take on what steps business leaders can take to make sure that generative AI has a positive impact on their business.

The new wave of generative AI systems such as ChatGPT, Bard and Midjourney are defining a new era in business—and they’re evolving at lightning speed. The impact is set to be enormous: research suggests that 40% of all working hours may be impacted by LLMs (large language models) like GPT-4—and that’s not counting the other types of generative AI. 

Human, or machine?

Until recently, content created by machines—or, AI—lacked a certain quality. Simply put, it lacked a human feel. It was easy to tell human-content from machine-content. But, in the latter half of last year, that changed. When the first version of ChatGPT launched on 30 November 2022, the quality of its outputs clearly struck a chord. It took just 2 months for ChatGPT to reach 100 million users—the fastest of any consumer internet app, ever. 

The newest hire at your favourite media company (image generated by Midjourney)

Perhaps unsurprisingly, this year has seen generative AI take centre stage. It’s broken out of the techsphere and piqued the interest of ordinary people around the world. (If your experience has been anything like mine, it’s been impossible to log on to LinkedIn without hearing from at least a handful of self-proclaimed “AI gurus”.) The sentiment is mixed. Some are excited about the incredible possibilities this new, groundbreaking technology offers humanity. Some are concerned their livelihoods are at risk. Others fear being left behind.

Whatever your disposition entering into this article, I’m sure we can agree on one thing. The potential impact of generative AI—whether you’re gleefully submitting to our new AI overlords or collecting canned-foods for your doomsday bunker (just kidding… kinda)—is enormous. Already—while, still in its early stages—its impact has been profound. I’m of the opinion that—like with all technology—generative AI will be a mirror of humanity. So, it’s on us to build, use and regulate it responsibly—and for the benefit of all. With that in mind, let’s dive into it.

What is generative AI—and how does it work?

Generative AI is a type of artificial intelligence that generates content such as text, images, videos, music or 3D models. Generative AI models tend to be trained on huge corpuses of data. These models identify patterns in the data they’re trained on, which they use to generate new content.

Generative AI has only a syntactical understanding of the data it's trained on. That means that it relies on statistical patterns—it doesn’t actually “know” that Beijing is the capital of China, for instance. (For further reading, try The Chinese Room Argument, first published by philosopher John Searle.) Because of this, on occasion, generative AI models produce outputs that seem plausible—but are actually nonsense. (For sensitive use cases, fact check. This is a good example of one of our core beliefs here at Datasparq—that AI will augment human intelligence, not replace it.)

Who are the big players in generative AI?

There are a number of big players in the space. Some of the most well-known companies include:

OpenAI: OpenAI is a research company that was founded by Peter Thiel, Elon Musk and Greg Brockman. It’s headed today by CEO, Sam Altman. It was OpenAI's ChatGPT that caught the attention of the media late in 2022. OpenAI’s GPT-4 language model is one of the most powerful generative AI models in the world.

Google: Google is another leading player in generative AI. Google has a number of different AI offerings, including Bard, BERT and more. In addition, Google provides many of the tools developers use to create and run AI systems.

Microsoft: Microsoft is another major player in the space. Arguably, it was adding AI to Bing that garnered the most media attention for their own AI offering. They partnered with OpenAI and the AI functionality in Bing used GPT-4. They offer a range of tools to create and run AI systems.

NVIDIA: On the hardware-side, NVIDIA recently became the first chipmaker to be valued at $1tn. Their hardware powers much of the generative AI ecosystem. (Jensen Huang, CEO of NVIDIA is a pretty cool guy, too.)

And let’s not forget about open source offerings: Additionally, there is an increasing number of open source LLMs, fine-tuned for specific use cases. Developer communities and smaller companies are playing a significant role in the generative AI space. Take a look at this list from Eugene Yan, Senior Applied Scientist at Amazon, on Github.

How can you use generative AI? (Real-world examples)

The number of use cases for generative AI grows every day. It isn’t surprising, given the sheer volume of AI startups entering the scene. 32% of the startups in Y-Combinator’s Winter ‘23 batch fell within the AI/ML category—a staggering figure. Today, I thought I’d share a few examples of how businesses are successfully using generative AI to cut operating costs or add net-new value.  

Marketing campaigns: The Coca-Cola Company recently launched a marketing campaign, Masterpiece, powered by AI. Heinz launched a rather smart ketchup ad campaign featuring images—and a narrative—created by generative AI.

“Just like humans, AI prefers Heinz” Heinz ketchup campaign (source)

Personalised experiences: Language learning app, Duolingo, used GPT-4 to enhance and personalise users’ educational experiences

Content creation: Generative AI can be used to create new content, such as articles, blog posts, and even books. Here’s an example of Sky News trialling ChatGPT for a writing gig.

Generating insights: Given that generative AI tools are able to analyse large amounts of data—spotting spikes, anomalies and trends—consumer banks are well-positioned to supply valuable personalised insights to their customers.

Coding: Tools like GitHub’s Copilot are being used to dramatically increase developer productivity. Here’s an example from Duolingo.

Product creation: Nike collaborated with designers using generative AI to create new shoe designs.

Customer service: Generative AI can be used to improve customer service by providing customers with friendly, conversational answers to queries, enabling customer service workers to be radically more efficient.

So, is generative AI a threat—or an opportunity?

This is what John Wyllie had to say…

“Despite the impressive, almost human-like, capabilities of LLMs, I believe they will be an opportunity for most businesses rather than a threat. Like all new technologies, it's not perfect—and its limitations will give businesses time to pivot and reposition themselves accordingly to play to their strengths.

But, it does mean that businesses should take a step back to evaluate the strength—and defensibility—of their proposition—what is it about their offering that their customers really value? Understanding these core aspects will help assess if LLMs are an opportunity or a threat. 

But, it does mean that businesses should take a step back to evaluate the strength—and defensibility—of their proposition—what is it about their offering that their customers really value? Understanding these core aspects will help assess if LLMs are an opportunity or a threat. 

One business we worked with provides consulting services and data from publicly available sources—it's plausible that this could be retrieved by a simple ChatGPT query in the near future—but there are a number of things this company's customers really value that ChatGPT will find it harder to do: the high-touch service of a human relationship is important—and the level of convenience that offers. As well as the level of provenance and trust in the information given as part of their service.

Content creation and the creative industry is where the biggest threats seem to be felt. On this I agree with Dave Karpf's observation that LLM technologies will find a niche for delivering 'satisficing' experiences.

Contrary to the excitement and buzz around LLMs, there are still some significant risks to consider before adopting them fully. Costs are one. The computational power required to train these models is significant. The creators and suppliers of these services will need to find a sustainable commercial model, and it's not clear where this will end up. So, the incremental cost of these LLMs must be outweighed by the incremental benefits (not to mention the environmental cost for the hardware to train/run these models). The second is regulation. Many politicians (as well as the entrepreneurs of these technologies) are keen for them to be regulated—so remaining flexible in how you deploy these until this is defined is crucial.”

How to make the most of generative AI—according to John Wyllie

  1. Stay informed—educate yourself on what they can and can't do to help you. Rather than relying on your LinkedIn feed for anecdotal evidence about what LLMs can do, source your own primary research by using them yourself to start evaluating end-user applications. 

    For example, try using Google's Bard rather than Google's search engine when doing some research on a topic. Does the convenience of summarised, synthesised prose outweigh the lack of provenance?

    When preparing for that next presentation, ask ChatGPT to suggest some section headings and key messages you might want to include. Was it easier to write the perfect prompt than to write the headings yourself?
  1. Experiment with the technologies—understand what it takes to build and support these solutions. Alongside your own personal end-user research, it's important to stay abreast of the development technologies that will help you and your businesses benefit from these. 

    Experiment as much as you can with the different technologies and techniques to build LLM powered solutions—and begin to understand the limitations and costs. Is it feasible (and cost effective) to build your own private LLM? Can your team support it if it goes wrong?
  1. Consider practical use-cases—how would you use 1,000 interns? This is often a question we ask our clients when helping to identify valuable applications of AI (especially those with an automation bent). It works well for LLM applications too. Where in your business could you benefit from 1,000 interns to do something better, quicker or cheaper? If 1,000 interns can help, there's a good chance AI can too. 

    Datasparq's Product Primer
    is a useful canvas we use to identify, qualify and validate these opportunities.
  1. Keep an eye on the regulation and costs—if one of your use-cases or experiments is particularly successful, you'll want to ensure it's sustainable. Be forewarned on any upcoming regulation that might limit you; and keep an eye on costs of the service providers you're using and engineer your solutions in a way that will continue to be affordable when the API fees settle.

  2. Make the most of the hype—get experimenting, create a proof of concept and demonstrate your innovation to your customers, staff and other stakeholders. Be part of the hype-cycle to benefit from the marketing exposure—but at the same time, be realistic. Don’t let the GenAI hype distract you from the more valuable and important (if more boring and prosaic) applications of data and AI in your business.

Undertaking a threats/opportunities assessment is the best way to understand the landscape in your industry—and we’re always happy to help. Our AI experts help businesses put the right strategies in place to be winners in the generative AI race. Whether you’re an enterprise, or a private equity fund looking for a generative AI DD, we can help. Get in touch with the subject “Generative AI assessment” if you’d like to find out more.

Don’t put it off—contact us now to get started

More insights

When you're ready, we're here

Contact us
Welcome to the club!
Oops! Something went wrong while submitting the form.