A Marketer's Guide to AI-enabled Content Workflows

The article is based on my talk at the Revenue Marketing Summit, Seattle .

AI has become a mandate at most organizations, and marketing is arguably the most ripe function for AI-driven change. Today's Large Language Models (LLMs) bring marketers closer to the holy grail of marketing - send the right message, at the right time, to the right audience. 

Marketers are dabbling in generative AI with mixed results

Let's start with a snapshot of where we are today. Salesforce recently surveyed over 1,000 marketers to understand their usage of generative AI. They found that over half the marketers are already experimenting with generative AI, and another 22% plan to join the fray soon.

The potential benefits are substantial - savings of up to 5 hours per week, increased productivity, faster content creation, and super personalized communications. 

Yet, AI generated content suffers from multiple challenges. LLMs hallucinate and produce wrong information. The writing can be basic and lack an angle or personality. You often run into limits on how much content you can paste into ChatGPT or other LLMs at one time. And sometimes, the outputs of AI models can be harmful or biased. 

Crowded tooling space makes choices difficult

And then there’s another challenge. Marketers realize that ChatGPT is great but they want more robust capabilities for their real workflows. They start looking for other generative AI tools. And that’s when they run into an avalanche of tools. 

We’re in the midst of an AI gold rush. The popular generative AI directory, futurepedia listed ~1000 generative AI tools in December 2022. In June 2023, that number crossed 350,000!

We need to pick the right use cases to narrow down the tool selection. This can be done with the help of a framework devised by Barak Turovsky. He evaluates use cases along two axes to determine a fit for generative AI.

• Fluency: How natural sounding the output is.
• Accuracy: How correct the output is.

His recommendation is as follows -

1. High fluency, low accuracy: Great fit for generative AI. Examples include social posts, basic blogs, children's books, and poems. These forms of content need greater fluency but accuracy isn’t mandatory.
2. Low fluency, high accuracy: Not a great fit for generative AI. Examples include search queries with clear answers (e.g., “When was Barack Obama born?”). Generative AI models have dated information and need to be supplemented with data sources to create value.
3. High fluency, high accuracy: Seems like a good fit for generative AI, but cannot be trusted blindly. Examples include travel recommendations and business emails. You need a person to check the answer.

How Generative AI helps in the content writing process

I've tested dozens of tools and read extensively about how marketers are using AI for content writing in the enterprise. Based on my understanding, AI is highly suitable for three areas of content creation:

First, idea generation and research. Services like Bing AI and Perplexity AI are great for finding ideas and information. ChatGPT and Claude (by Anthropic) can analyze documents and create summaries so you can extract snippets and talking points for your piece. Always double check the facts though!

Next, AI tools like MidJourney and Adobe Firefly are awesome at turning natural language into creative images. I've used Lexica.Art to search hundreds of AI-made images. But beware - AI art is controversial since it can infringe on copyrights. Adobe avoids this by using their own image data. For safe use, stick to non-commercial applications.

Finally, copywriting and editing requires the most human oversight. Once you have an outline, give AI tools like ChatGPT, Claude, Poe (by Quora), Jasper and Writer the inputs, and watch them create drafts with the tone and voice you like. You would need to fact check and revise the final copy, and polish it for semantics. These tools can also personalize content for different audiences or create more assets from your original.

How to combine AI and marketers into impactful workflows

AI is getting pretty good at the groundwork for content creation. But it still can't match human judgment, creativity and strategic thinking. The key is using AI to simplify and scale your efforts, not replace them.

AI is great at scaling repetitive tasks or analyzing huge amounts of data. And marketers provide the creativity, contextual thinking and complex decision making that the business really needs.

The key is a symbiotic relationship where AI and marketers play to their strengths. Each one alone would be outperformed by a teamwork approach.

Three workflows for collaboration

Based on the opportunities and risks discussed so far, let’s talk through three workflows for content creation. These workflows allow AI and human collaboration to yield the most impact in your content process.

Workflow 1 - Use AI for research or first drafts, marketers for editing and approval

As a content creator, AI can be a huge help when you're stuck in writer's block. It's great at suggesting thought starters, or at summarizing research into short snippets you can add to your writing research.

AI can also suggest blog titles, draft your first email copy, come up with social media posts or even ad copy.  It handles the tedious work of generating initial ideas so you can focus on the creative stuff.

But here's the thing - you still need humans to make that content shine. AI may generate ideas, but it can't match your creative instinct or truly understand your brand's voice. It's up to you and your team to refine AI's drafts, double check the facts, and ensure the end result sounds uniquely like you.

Now, there are some ways to help AI match your brand voice more closely. Try giving AI multiple examples of your best content as prompts. Services like ChatGPT and Writer.ai are great at this - with more examples, they get better at enforcing your brand tone and standards.

The more prompts you provide, the better AI gets at sounding like your unique brand. Check out resources like OpenAI's prompt engineering best practices for tips on how to write powerful and effective prompts.

Workflow 2 - Marketers build core messaging while AI personalizes at scale

Your marketing team creates your core messaging and content. They shape the strategies and stories at the heart of your brand.

These content assets, like PDFs, landing pages and more, are uploaded to an AI system. The AI uses them as a base to create short-form versions for different channels, localize it for other languages, and even generate persona-specific variations with the right tone and voice for each audience.

The AI can customize and scale your content, but your marketing team still oversees everything the AI produces.

Workflow 3 - Apply AI to data analysis, marketers to strategic decisions

AI is great at finding patterns and connections in huge amounts of data. It can take semi-structured data and organize it into tables so you can analyze it better.

For example, you could give ChatGPT or Claude a chunk of text and ask them to create a table with the details you need. ChatGPT is shockingly good at this! The tables let you do meaningful analysis on data that was hard to understand before.

AI models like ChatGPT and Claude are also awesome at processing massive amounts of unstructured text. This makes it way easier to spot important patterns and trends you might have missed. You can give them transcripts, articles or reports and ask for summaries. ChatGPT can remember up to 8,000 words of context and Claude up to 75,000 words - enough for hours of audio or pages of text.

Another way AI helps with data analysis is by generating code. You can ask ChatGPT or Claude to write Excel or Google Sheets formulas in plain English. For example, "Write an Excel formula to find all the Johns in this table and show their average age." ChatGPT will give you a perfect formula. This is so useful when analyzing campaign results or comparing strategies!

AI simplifies finding meaning in huge volumes of data, while humans provide the intuition and critical thinking to turn insights into impact. Marketers determine the questions and make decisions around how insights influence strategy.

So how are you thinking about generative AI

In conclusion, generative AI will massively disrupt the content landscape. Experts call it the commodification of content, where content becomes a product like any other. I’d encourage you to find your answers, and share your experiences with generative AI. 

So, I leave you with some final questions to think about:

1. Do you have a vision for how generative AI will elevate your content marketing?  
2. What is your plan for upskilling to become an AI-enabled marketing function? 
3. Do you have identified use cases or workflows to start piloting generative AI on a small scale?