The importance of context when using AI for content creation

AI tools can spit out content with a simple prompt and a click of a button, whereas humans have a lot more context when writing something. How do you bridge that gap for AI

Image generated by DALL-E

Isn’t it fascinating to see ChatGPT churn out content on any topic with a one-line prompt? Whether you want to write about travel, technology, science, or marketing, LLMs, including OpenAI’s GPT that powers ChatGPT, are trained on an astronomical amount of data that enables them to generate content on pretty much any topic.

When it comes to marketing though, context behind the content is everything. Content produced for marketing needs to speak to and resonate with the audience and hopefully drive a desired action or behavior. Moreover, it needs to embody the beliefs, value propositions, and differentiation of the brand or product that is being marketed.

The challenge with using AI to generate content

Even the simplest of products have more than one audience. Take an email marketing tool, for instance. The end user of the tool is an email marketing specialist/manager, the executive sponsor is likely the director or VP of marketing, and the budget approver is the CMO. The value proposition of the tool is different for each of these audiences - features and ease of use for the end user, team productivity and throughput for the executive sponsor, and ROI for the executive sponsor and CMO. Now imagine using ChatGPT to generate product one-pagers for each of these three audiences. How well do you think ChatGPT is going to be able to personalize the content for each? How much context would you need to provide every time you need to generate a piece of content?

Now add a layer of multiple products to create content for, and tones of voice to use for each audience or content type. The combinations of audience+product+tone+content quickly get out of hand. The agility that AI provides in creating content starts becoming less attractive with the requirement of providing and maintaining the depth of context needed to create the content.

The risk of a lack of context in AI-generated content

The biggest risk of providing partial or no context to generative AI tools is the creation of generic, undifferentiated content that fails to showcase a brand’s unique personality and value proposition that needs to resonate with the audience.

I ran an experiment with ChatGPT sometime ago where I asked it to write out 100-word introductions to Uber and Lyft, two very different, but well-known brands in the same space. The result is shown below.

Providing context to generative AI tools

One way to address this is to provide enough context to the tool before asking it to generate a piece of content. This involves creating a prompt that does the following,

1. Instructs the tool to behave as the specific talent you need to generate the content - a short-form copywriter, a blog writer, an email marketing specialist, etc.

2. Provide guidance on how to approach the content creation task - best practices, guard rails, do’s and don’ts, etc.

3. Explain what your brand and product offer the intended audience of the content, along with your differentiation and messaging hierarchy to ensure strong message pull-through in the content.

4. Describe your audience, their behavior, challenges, and their role in the buying process.

5. Define a tone of voice to use. The more granular and detailed you are in describing this voice, the better the output. For instance, simply asking the tool to use a serious or playful voice may be insufficient to create differentiated content. You may want to describe your brand’s personality and values in detail, along with how these come to life in the content to truly add uniqueness to it.

You would need to do this for each product, audience, content type, and tone of voice you plan on using.

How we do this inside TheWordsmith.ai

A few months ago, I embarked on a journey to create an AI copilot for marketers that truly understands a brand and product inside-out, and why and for whom it is creating content. Coming at it from a marketer’s perspective, I built it such that brand and context form the foundation of any content or response the AI generates.

The Knowledge Base

At the core of TheWordsmith is the knowledge base, which is a combination of,

1. The functional aspects of a brand/product - what it is, how it differentiates itself in the market, its value propositions, its ideal customer profile, etc.

2. The brand’s purpose, personality traits, and a sample piece of content that together determine a unique tone of voice for the content to be created

Additionally, we train the AI on best practices, approaches, and frameworks for creating different types of content.

Managing multiple knowledge bases

As articulated earlier, even the simplest of brands may have more than one product or solution, and speak to several audiences within the context of each. So we added the capability to create multiple knowledge bases with the ability to dynamically switch between them when creating content for a specific audience and product. This is by far the most unique aspect of TheWordsmith.

Ease of building knowledge bases

Marketers will have already done all the hard work in segmenting their audience, creating messaging hierarchies, product/landing pages, and more on their websites. We leverage the same and let users provide the URL of the relevant page for a particular knowledge base and make the AI extract and analyze all the relevant context needed and build out the knowledge base.

Once you’ve built the set of knowledge bases needed for your content needs, it is a matter of selecting the appropriate one from a drop-down list in order to switch contexts.

To sum it up

Whether you are using ChatGPT, or another tool designed for marketing content creation, it is critical to assess,

1. How strong are its capabilities to ingest and maintain the content you need before asking it to generate content.

2. How easily are you able to build and maintain this context on an ongoing basis

3. How efficiently can you maintain and switch context between the various brands, products, and audiences that you market to.