Getting an incredible advantage in the age of AI
The widespread availability of AI algorithms has made it possible for businesses to incorporate AI into their operations and products. However, as Caspar Yuill, senior strategist at Affinity writes, with such AI ubiquity, what will be the competitive advantage?
AI algorithms are well and truly here. Now that anyone can get a hold of ChatGPT, it seems like everyone is trying to work out how to take advantage of it. But when everyone has access, how will one business get an edge on any other.
If you’ve ever seen Pixar’s ‘The Incredibles’, you’d know that the supervillain ‘Syndrome’ ends his evil-plot monologue with the line ‘And when everyone’s super, no-one will be’. Who knew a Pixar flick would give us such an apt summary of the state of AI in companies today?
Because today, AI is about to deliver Syndrome’s super vision to corporations worldwide. Only this time we have Mark Zuckerberg and Alphabet CEO Sundar Pichai to thank. Sadly though, the real world lacks a Mr. Incredible who can hurl them into outer space.
So why these tech giants? Well, frameworks like Facebook’s PyTorch and Alphabet’s Tensorflow have been open-source for years, and recently, generative models like GPT and DALL-E have become readily available from Microsoft Azure. Everyone has started using them. Over the next couple of years, companies will start to integrate them within their products and operations, and we’ll quickly reach a stage where everyone has AI models in production. But if everyone has these models, will anyone have an advantage?
In short, yes. But it depends on the quality and amount of data that you can get to train your AI with. Is it any wonder now why Google and Facebook are so keen on developing AI tech?
Facebook and Google’s vast data sets, and any other significant data sets out there, will be like a new kind of oil. And just like oil, they’ll be refined to become a source of competitive advantage. LinkedIn and Facebook realised this years ago and started limiting access to their datasets. In Facebook’s case, they cut off access to their open API, and LinkedIn closed house with its anti-scraping policy. Both of these actions were to retain as much private data as possible.
That data is valuable because it’s a way of training AI. Think of it this way: Connecting your business to AI is like hiring a thousand interns (Credit to Benedict Evans for this analogy). Asking your army of interns to immediately start writing content for your websites or social channels, or to man the customer service lines without any training isn’t going to go very well. They’re not going to know how your company offers refunds, its tone of voice or how to deal with a disgruntled caller. Now, if you take those same interns and train them with a dataset on your company brand or customer service principles, they suddenly become much more competent.
Take this idea one step further and imagine you set your AI interns loose on your products, instead of just your business functions. In fact, agricultural company John Deere is doing just that. They’ve been investing in automation tech since the 90s and are becoming one of the world’s most important AI companies today. Lately, they’ve been pointing out what a weed looks like to their AI using the video footage and sensor data they’ve gathered from their tractors for years. Now they have a custom model that can tell their tractors what weeds to pick built right into the product. Further iterations of models can account for new weeds, creating a system that is more valuable the more data John Deere feeds it. All of a sudden, a John Deere tractor can help a farmer maintain his crops in a way no other tractor company can. It’s an example of how a dataset gives them a superior product and a competitive advantage.
Think this is far-fetched? Try it yourself.
An AI project shouldn’t need massive investment to get started. Internationally-renowned startup investor Ash Fontana coined the phrase ‘Lean AI’ in his book ‘The AI-First Company’, and it’s probably the way to go to minimise wasted time and maximise your return. If you’re wondering where to start, here are some suggestions:
- Fine tune a GPT model on your company’s policy and process documentation. It’ll become a virtual office manager who can answer questions on your company’s procedures.
- If you’re a content-heavy business, fine-tune GPT with content you’ve already produced. The AI will learn your brand’s tone of voice, then draft new content for you. You’ll still need to edit and refine this content, but it’ll get you 60–80% of the way there.
- On the predictive side of things, use the data you already have to add more value to users. For example, automotive brands know how often customers with particular car models come in for a service. They could use this data to predict when people need to service their cars, and provide insight on the most common issues. Over time, they could use the feedback data to refine the accuracy of predictions.
Businesses that have been diligently squirrelling their data away have a huge opportunity with generative and predictive AI models. By taking existing models and training them on private data, companies can create applications for AI that offers them a defendable competitive advantage. And even if companies haven’t, creative uses of existing data like sales or customer service data can be used to start experimenting with these new models to do some incredible things.
Caspar Yuill is a senior strategist at Affinity.
Insightful and charming. Great article
User ID not verified.