This article originally appeared on LinkedIn.

We blinked and all of a sudden the effect of (Chat)GPT and related technologies was everywhere, immediately. Why has this technology become so popular overnight? Well for one, it does an excellent job generating content — and quickly. It can also generate answers to complex questions, cliff note a book, tell you top grossing movies, and explain key terminology. What seems to impress everyone the most is the output it can generate with the simplest prompts.

That’s also the problem. Being able to generate content with simple prompts can lead to a greater quantity of content — but not necessarily the appropriate quality of content for your desired audience and market. When you have a ton of services tapping into common AI with little or no context on your audience, all you will end up with is a broadly available, unemotional, computer-generative voice — not the voice your brand or audience is necessarily looking for. So how can you overcome this problem and make the best of the technology? Below are three key areas to focus on that’ll help you drive greater impact and attract the right audience with your AI-generated content.

All examples provided below include generated responses from GPT.

1) Know your audience

Rule number one of communication is ’know your audience.’ While we’re currently enamored with the evolution of ChatGPT and generative AI’s output, we still need to solve the problem of context and audience to apply this technology in broader applications.

Here’s an example to highlight the importance of context and audience when using GPT:

Can you explain cryptocurrencies?

Cryptocurrencies are digital or virtual currencies that use cryptography to secure and verify transactions and control the creation of additional units... They are powered by a distributed, peer-to-peer network of computers and use a consensus algorithm to confirm and record transactions... The most popular cryptocurrency is Bitcoin, which was created in 2009. Other popular cryptocurrencies include Ethereum, Litecoin, and Ripple.

Can you explain cryptocurrencies such that a non-technical person can understand?

Cryptocurrencies are digital forms of money that are created, used, and stored online. They are not tied to any government or bank, and instead use decentralized networks of computers to manage and secure the transactions. Transactions with cryptocurrencies are fast and secure, and can be used for a variety of purchases, from buying goods at stores to investing in stocks and other financial assets.

Can you explain cryptocurrencies such that a child from France can understand?

Cryptocurrencies sont des monnaies numériques qui peuvent être utilisées pour acheter et vendre des produits et des services. Elles sont différentes des monnaies traditionnelles parce qu'elles sont virtuellement stockées et transférées entre les personnes sans passer par une banque. Les cryptomonnaies sont sécurisées grâce à des technologies très avancées.

While the first, basic prompt provided an impressively accurate and comprehensive result; was it appropriate for the other audiences I intended to communicate with? Absolutely not.

I reached out to Microsoft’s own Pablo Castro (CVP, Distinguished Engineer), who is leading much of the thought leadership in their platform's application of AI:

Appropriate context and grounding data are going to be key to helping users determine the trustworthiness of responses generated by ChatGPT-like experiences in applications...

It’s an excellent point that using the right context and voice with your audience will build trust in your adoption of this technology with your customers.

2) Use audience context to improve AI output

The future of generative AI technology, such as ChatGPT, lies with customer data platforms and similar technologies. Many businesses have a singular audience and an on-staff copywriter or marketer who can easily generate original content for them. But let's scale this to a broader market, to a diverse population of customers across many products and services. Another example, what if I ask GPT to sell the same pen to an artist versus teachers? What are the similar versus dissimilar talking points generated?

Sell a professional artist this pen

This pen is perfect for a professional artist. It features a fine-tipped, waterproof, and fade-resistant ink that will not smudge, making it ideal for precision drawing. The ergonomic rubber grip makes it comfortable to use for extended periods of time, and the durable aluminum body ensures it will last for years to come. It also comes with replaceable ink cartridges, so you never have to worry about running out of ink.

Sell an elementary school teacher this pen

This pen is perfect for elementary school teachers! It's made with high-quality materials, so it's sure to last a long time. It has a comfortable grip that makes writing easier and more enjoyable. Plus, it comes in a variety of vibrant colors, so it's fun and stylish. The ink is smudge-proof and quick-drying, so your teacher won't have to worry about messes. With this pen, they can easily write out lesson plans, take notes, and grade papers with ease.

Even the simplest audience context can greatly improve the quality of generative AI for personalized content creation with your customers.

Open, customer data platforms (CDP), such as Sitecore CDP, specialize in aggregating information about your customers and their omnichannel interactions as first-party data — meaning, it’s your data just for your customers you do business with. By combining this first-party data about your visitor, with GPT capabilities, you can achieve one-to-one, personalized content in real-time.

Combining CDP and ChatGPT (Retrieval-Augmented Generation, or 'RAG's)

  1. Take input from your visitor, including their known visitor ID (from a cookie or a secure login) to submit to your website or app
  2. Combine the visitor input with context fetched from the CDP based on historical visitor information (geo, segment, preferences, language) to create an improved prompt for GPT
  3. Return a personalized response from GPT, which makes your visitor feel like they are a known individual versus a frustrated, generic bystander

3) Consider generative biases

Even without providing context or audiences to your generative AI prompts, businesses must be aware of implied biases that can come out of this technology. As an example, Americans will usually draw an affinity to a particular news source for information — but even that news source may publish material that you – the audience – do not always agree with. The same applies to generative AI.

Secondly, as you incorporate information about your audience into generative AI, there are some aspects of your audience you may want to leave out. For example, incorporating household income into your audience profiles could result in the information getting added directly to generated content as, “As a high-income resident of Kansas, …”

My guidance on how to handle this is two-fold:

  • Be careful about what information you provide to generative AI about your customer(s) without proper boundaries.
  • Establish processes with knowledgeable people to review auto-generated content from AI to approve and continually tune your processes.

In my experience, including audience context is necessary, but do not forget to instruct it specifically on how to apply or exclude the audience detail in generating content for you.

Key takeaways

While ChatGPT and other generative AI technologies are a major leap forward, the industry runs the risk of re-normalizing to the same voice, same output—same news outlet.

When you’re looking at applications of generative AI technology to elevate your brand and speak to your various customers, the difference will be in your ability to maintain knowledge of your individual customers to improve personalization.

Tell the AI who you want to talk to by investing in technology that can build that personalized audience context for them – and for you.

Using generative AI technology can also lead to unintentional biases in your content. That risk can be mitigated by giving the AI as knowledgeable of a request as possible with thorough instruction and having someone review the content generated for errors – but I encourage you to remain diligent about how you protect your brand image.

I am looking forward to seeing how Microsoft and OpenAI continue to innovate this notion of audience and context as a first-class component of GPT interactions – to build trust in the technology and broader business applications.

Jake Hookom is Vice President, Commerce and Platform at Sitecore. Follow him on LinkedIn.