Agency Anecdotes: Likes and learnings about AI as a content writing assistant

A creative agency using AI? Blasphemy! 

Confession time. 

About a year ago, with the launch of our new brand, we started writing a new swath of content for our followers (like you) to enjoy. In the wake of the COVID pandemic, our previous content writing efforts had waned – and it was time to get back on the horse. 

But our team, the market, and the tools available to marketers had also changed over that time. Which meant that by starting this content journey anew, we had the opportunity to re-envision our methodology. 

And leveraging AI was an obvious place to start. 

Your gut reaction to this admission might be shock or outrage. We get it. 

After all, AI is being heralded by many in the industry as a death knell for creative jobs. Despite it being around for several years already (where does the time go?), there’s still a lot of trepidation about how to use AI to make people’s work better – not make it obsolete. 

So, when we started leveraging AI for our content, we were clear about our intentions. Its utility for us was as an assistant to our writers, not a replacement. 

And in that capacity, there’s a lot to like. 

How we’ve used AI to assist our content marketing writing 

First a quick overview. [shameless plug alert] 

At Show and Tell, we have many subject matter experts in a variety of disciplines: from demand generation, to digital marketing, to linear media, to advertising creative… We do it all, and all well.  

So, for each of our content articles, we start with the person with the in-house expertise. Not a machine. 

Step 1 – Get the fodder 

We ask the team member best suited to a particular topic to write down their thoughts, in their words. We often complement this with a recorded conversation (‘interview’ would be overstating this), so we capture their words, idioms, and personality.

Both the format in which we ask them to jot down their thoughts, and the conversation we record are structured to make sure we’re gathering the information in a way that it’s useful for you when we publish it.  

Step 2 – Feed the beast 

We provide our AI tool-of-choice (Claude and ChatGPT each have their advantages, in our experience) with the written content and the recording, and give it specific directions on what we want. Using our subscription to these tools, we’ve already set up a brand tone and voice that the AI will adhere to, so our prompts can be more specific to the article at hand.  

A couple points to make here. 

First, you’ll note the inputs to the AI, are all people-generated.  Our ideas, our point of view. We’re not combing the ‘net or injecting other source information to the AI to plagiarize, or claim as our own. That’s important (and integral). 

Second, the ‘pre-work’ – setting up the AI to generate content in your tone of voice, learning and iterating what makes a great (and not so great) prompt, etc. – is also human-powered. 

Step 3 – Edit, edit, edit 

When the AI generates an article based on our input(s) and direction(s), it once again goes through a human filter.  

Or rather, several: 

• The writer assigned reviews and edits to make sure it makes sense (and regenerates, with new prompts, if it doesn’t).  
• The subject matter expert reviews and edits to make sure their perspective is captured accurately.  
• Our content editor reviews it to make sure it has some personality to it – specifically, our personality. And that it fits the general voice and tone of the suite of content it’s published alongside.  
• A human proofreader reviews the final draft before publishing (even if the AI has done a proofreading pass), because quality matters. 

Some of these steps can, of course, be consolidated. But none of them are skipped. They all matter, especially with the changing undercurrent of how people are searching and finding content on the internet (ironically, also an AI-assisted environment). 

Step 4 – Iterate (and edit, edit, edit) 

Once our article is drafted to our (human) satisfaction, we use the final draft to iterate all the peripheral output we need to get it to you. This includes the metadata for SEO purposes, the email we send to newsletter subscribers, and suggestions for (platform- specific) social captions. 

For each of these outputs, we repeat step 3: multiple real people with subject matter expertise (in email marketing, social media, SEO writing) review the AI’s output. Nothing’s published without a smart person signing off. 

3 things we like about AI-assisted content 

So, after a year of repeating and honing these steps to publish a couple dozen articles, here are our takeaways.  

1. It speeds up the first draft 

The notion that there’s nothing as daunting as a blank page is true. The biggest strength of AI-assisted writing is it gets you to a first draft, fast. In our early days, those ‘first drafts’ weren’t very good. But as we refined our prompt writing, and as the AI allowed us to integrate brand voice ‘style guides’ that we could use consistently, the initial output from the AI got much closer to the final piece. 

This isn’t to suggest that writing, as a craft, isn’t valuable. But not all our subject matter experts are great at writing on their subject matter. And while having a gifted, nuanced writer assigned just to massage our experts’ excellent insights is valuable, even necessary… it’s kind of like using a bazooka as a flyswatter. 

Enter the flyswatter. Taking content from nothing to something is easily the most time-consuming part of the process. And the AI virtually eliminates it. 

2. It saves a very specific kind of time 

With the first draft being generated with the assistance of AI, that means our time and thoughts are more squarely focused on where we provide the most value: the input, and the edit. 

Or, if you prefer, what we’re showing you, and how we tell you about it. #BrandedContent 

Because the AI output is only as good as what you feed it, the majority of our time in this new process is in step 1 – having our experts articulate and document (in writing, or on the record) their ideas and analysis. That’s the stuff that keeps you coming back for more of our content, so it makes sense that it’s where most of the time should be spent. 

Our second-highest time investment is now in step 3 – the edit.  

Because we use the machine. But we don’t trust it. If we’re publishing something with our name, expertise, reputation and credibility attached, you better believe we’re going to fine-tooth-comb it. And on that note… 

3. It’s best when it’s people-powered 

One of the reasons our clients have tried and failed to implement AI content to their satisfaction (and have hired us to help) is they’ve fallen into the erroneous assumption that AI makes this all happen ‘instantly.’

But they ignore that AI needs input in order to do… anything.

And AI-generated content needs vetting to sound like it comes from a real person – let alone a specific person like your in-house experts. 

In other words: people still need to put in the time, the effort and especially the insight. AI assistance doesn’t remove the need for your experts’ crucial and valuable involvement. In fact, it depends on it. 

Triage time: Do you need an AI assistant, or an AI expert? 

If you’ve been curious about how AI can help your organization generate content more efficiently, hopefully you can draw from our experience.  

But now that you’re better informed, consider: Are you equipped to leverage the AI yourself, or are you better off working with a partner that’s already done the troubleshooting? 

Do you have the bandwidth to do content marketing properly?

Most organizations are generating thought leadership content off the side of someone’s desk. This is especially true of small-to-medium sized businesses, who don’t have the budget (or luxury) of having dedicated hires devoted to their content marketing. And the experts they depend on for the fodder for that content already have jobs – being experts – and lack the bandwidth to write and reflect about what they do, while also doing it. 

For those organizations, we can help.  

From setting up your content strategy, to programming style guide parameters, to managing the various steps we’ve outlined above – our team works with yours to generate content that’s powered by you, your people, your expertise, and your point of view.  

We do this as a standalone for some clients, and as part of an integrated marketing strategy for others. 

Need guidance to develop an AI content writing strategy?

For those organizations willing and able to take these learnings and apply them in-house, we can also help.  

Strategy and direction setting at the outset is one of the most overlooked steps in content marketing. And one of the most valuable.  

We can work with your team to help you establish a clear direction on who you’re targeting and what your content will focus on, how you’ll program and manage it, and what kinds of prompts will (and won’t) work to get better quality AI output, on your first try.

To learn more about AI-assisted content, and how you can put machine learning to work for you, reach out to speak to one of our in-house experts.