Forge & Spark AI for content marketing spilled coffee

AI Wins and Flops—Lessons from an Agency’s Journey Using AI for Content Marketing


Table of Contents

Forge & Spark Media is a content agency dedicated to serving purpose-driven brands including non-profits, social ventures, and fellow B Corps. While that description might not scream “AI early adopters,” our team is committed to delivering genuine value to our clients and audience—and as an agency owner, I’m genuinely interested in exploring the best ways possible to do that as efficiently and effectively as possible. So when viable AI tools for content marketing started coming online back in 2022, we got curious.

An Agency’s Journey Using AI Tools for Content Marketing

Forge & Spark Black Cat Looking

We set out to explore and test—keeping firmly in mind our commitment to always deliver effective and premium-quality content that feels authentic to an audience. So we began by applying AI tools to a range of content marketing activities from content ideation to SEO, research, content development, content editing and optimization, image searches, and visual development.

Along the way my team has seen our share of hits and misses, wins, and utter flops, helping us to sort out for ourselves how best to use it—and when absolutely not to.

Forge & Spark Do not enter trail sign

In this article, my team shares some of those wins and flops with you, offering their insights on how you, as a marketing or business leader, might use AI to meet your marketing goals and create content you can feel proud of.

How content marketers are using AI in 2024

Agency and marketing leaders have varying opinions on the best ways to use AI in marketing, as demonstrated in this 2024 AdWeek article. A clear theme that’s emerging, however, is that AI can be a boon to efficiency—and should have a place in the day-to-day operations of businesses (including ad shops, and creative content agencies like ours).

The counterpoint—which shouldn’t be ignored, especially for purpose-led brands looking to communicate brand values and purpose with integrity—is that we must also use caution, and be aware of our responsibilities.

In my own discussions with agency owners and content marketers, I see AI being used to streamline workflows and enhance efficiency in several ways:

  • Content ideation and research: AI is an excellent companion for content marketers when it comes to analyzing data and market trends, identifying the most engaging topics in a given industry or field and tailor content to a specific target audience.
  • Producing content at scale: With AI doing some of the heavy lifting, content creators are able to spend more time ideating, editing, and curating—and less writing from scratch.
  • Editing and improving content flow: Tools like Grammarly and Hemingway have been around for years helping to streamline editing, and now there’s a whole new raft of tools to help anyone create polished content that minimizes (with some fact-checking!) your technical and grammatical errors.
  • Image searching and visual development: While AI is still a bit shaky on the generative visual front, delivering some downright odd imagery (some fun examples below), it’s helping designers (and non-designers) to optimize visual content; for instance, by refining keyword searching and introducing a huge range of new tools for editing and improving images and video.

But with anything, some attempts turn out better than others. What’s working and what’s not? We asked our team about their recent wins and flops.

The Wins: What AI is really good at when it comes to content marketing


I was trained as a writer and journalist, so when I gave myself the task of checking out AI tools, I admit to feeling highly skeptical about AI’s wordsmithing capabilities. I remain cautious, but am now optimistic about—and delightfully surprised by—AI’s capabilities. As a writer, I know how challenging it is to start writing anything from the blank page. So I’ve begun using tools like Narrato and Jasper to help me come up with content and blog ideas, as well as to help me ‘tune’ my crazier original concepts into more search-optimized content that people will likely find and engage with in their quest for information online.

I’ve been using Narrato in particular to help me come up with winning SEO meta descriptions and titles; in fact, we used it for this article.

I’ve also been using Perplexity as a research tool…  and am finding it often excels where Google falls down—providing immediate and clear answers, backed up by helpful links and sources, so that I can go directly to the research to verify where the copy has come from. This itself is a breakthrough for an AI tool, and so incredibly valuable for anyone concerned with actually knowing where the AI research and writing is coming from.

It’s not always stellar writing, I have to say—it builds, after all, from all the marketing content out there, which isn’t always praiseworthy—but I am nonetheless grateful that it gives me a place to start from.


I think there’s so much cool stuff going on in the world of AI, and it’s just going to keep getting better and better.

One of the big AI wins currently for visual storytelling are all the new AI-powered tools that help designers to refine and elevate photos, graphics and video. I’ve dabbled successfully with Canva’s Magic Design, for example, which creates templates based on text prompts and uploaded images. Figma’s AI-powered colour palette generator is great for foundational work and branding. And Adobe Photoshop’s Generative Fill, Smart Healing and Removal tools can shortcut image-editing processes that used to take hours—down to minutes or seconds.

I also see a lot of potential for using different AI platforms in combination. For instance, I’ve been using Chat GPT to help me build prompts for Midjourney, with some really interesting and fun results. I also buy prompts all the time from PromptBase to see what other people are using. Refining the outputs still takes time and effort, but Midjourney can provide ideas and starting points that you can bring into premium tools like Adobe to work with further.

I’ve learned that my art school training is very handy when it comes to figuring out keywords for prompts. Good thing I paid attention! Here’s a recent example of Midjourney art I created using the following prompt:

A minimal and surreal painting with a vintage color scheme of sepia, faded teal, and dusty rose; using acrylics and textured brushwork for a vintage and surreal atmosphere; inspired by vintage art and surrealism; mood is nostalgic and dreamlike.

Forge & Spark AI content marketing minimal and surreal images collage Forge & Spark AI content marketing minimal and surreal images

WIN: HEADLINE HELP—Lisa Manfield, Editorial Director

It’s true when they say that writing long is often easier when writing short, and I find that especially true with headlines. Headlines are the most important component on a page when it comes to grabbing a reader’s attention—and optimizing for SEO. But getting them just right is tough—and time consuming. I’ve used ChatGPT and Google Gemini for inspiration with crafting headlines—or at least getting a start on them. I admit, I have yet to see a headline that is masterful enough to use (AI seems to love clichés), but they’re often good starting points, and give me a foundation to work from to improve the headline and come up with something that is accurate and engaging.

WIN: RESEARCH AND BRAINSTORMING—Kirsten Rodenhizer, Senior Editor

I’ve found AI handy as an idea-bouncing buddy—using it almost as a virtual content assistant that I assign sub-tasks to support me. Chat GPT, Narrato and recently Google’s Gemini have proven solid for creating story outlines, and quickly getting a sense of the points typically covered in a certain subject area (which also tells you what you need to get beyond to distinguish yourself in the sea of AI-generated content out there right now).

I also love to have AI tools riding shotgun when I sit down to write boat-loads of social captions for our client projects; inspiring new ways to tackle well-worn subject areas. Much like working with a real assistant, however, you wind up spending some extra time carving out pieces to delegate, and then thoroughly checking the work, rewriting into your own voice, expanding, and elevating. It’s a trade-off.

WIN: CREATING BLOG BRIEFS—Carli Van Stolk, Content & SEO Strategist

Preparing briefs for writers can be time-consuming. But AI large language models (LLMs) like Gemini can seriously speed up the process. I love using Gemini for preliminary research and ideation, all while double-checking its output for accuracy. Next, I go into search engines and conduct manual research myself to see what’s in SERPs. With that information and Gemini’s initial insights, I can then create a search-optimized blog brief with topics, subheadings, and client-specific insights to cover the bases we want.


As a content strategist, I love using AI like it’s a brainstorming buddy who helps my ideas take off. I tend to brain dump all my thoughts into ChatGPT and prompt it to weave my ideas into a coherent story. This method is super-handy for presentations and slide decks when I have information (usually too much!) that needs to be distilled down into what’s most important for our clients or an audience. AI helps to build a narrative by organizing insights and recommendations into neat sections with headings and subheadings. It’s a tool to get me started and enhance my creative process—it’s like having the perfect wingwoman for our shared vision.

The Flops: What AI is really bad at when it comes to content marketing


As a marketer I know all too well that the copy I create (as a human) is not always Booker Prize quality. We’re often under pressure to write that newsletter intro or teaser, or to finish that blog post. We write copy to prompt a particular belief or action, rather than to convey a complex thought with nuance. It’s important to note that THIS COPY—flawed, rushed, intent on manipulating emotion and action—is the content that many AI tools have been trained on. And this is why you’ll likely find, as I have, that the writing produced by AI tools is often promotional, stilted, cliched, sales-y and—yeah, wrong a good chunk of the time. It’s going to get better and better, but for now, be wary of trusting any AI copy without a thoroughly human review.

FLOP: LONGER-FORM WRITING—Lisa Manfield, Editorial Director

Whether I’m writing blog posts, long-form content or social captions, it’s nice to have a first draft to start from. And I’ve found ChatGPT can easily deliver something to get the creative juices flowing. But I wouldn’t ever consider copying and pasting anything more than a few words here and there. Frankly, I find its copy to be generic, flat and generally boring to read. There is no personality to it, no human interest and often no limit to the clichéd phrases and formats it uses. Perhaps in time that will change, but I feel that, while it can give a leg up on the writing process, in no way can it currently replace the masterful crafting of a human writer.


AI can be a useful tool in its lane, but a journalist it is not. Out of curiosity, I’ve tried introducing prompts like “adopt a storytelling tone” or “write in the style of a magazine feature.” The results are generally pretty silly and saccharine. Good luck adding real human voices, too—it has coughed up fairly comical results when asked to read interview transcripts and incorporate quotes. I’m sure that will improve, but it’s not there yet.

Factual accuracy is also a huge area of concern for me with AI tools, in my work as an editor and in my own writing. They seem to Hoover up bits and pieces from all kinds of sources, reliable and not. Lately I’ve seen drafts from writers who have pulled in AI-generated content, not realizing that a certain detail was totally invented—by an eager-to-please bot scripting for effect, not accuracy. Yikes!

FLOP: SOCIAL CAPTIONS WITH CHARACTER—Carli Van Stolk, Content & SEO Strategist

I find ChatGPT really tough for creating social captions that aren’t generic. Despite using detailed prompts—including explaining who you are as a user, the purpose of the post, background information, and the audience you’re writing to—the output for social captions follows the same formula. And always includes way too many emojis.

Personally, I find it most effective to ask ChatGPT a question about the topic you’re writing a caption about, and cherry-picking relevant information from there. For example, say you work for an agency, and one of your clients is a fine dining restaurant that’s now serving duck confit. Instead of using a prompt, I find it most effective to continually ask ChatGPT questions, and then pull relevant information from its answers into an engaging social post.

FLOP: REALISTIC GENERATED IMAGES—Lara Kroeker, Creative Director

People were so excited about generative AI platforms like Midjourney, DALL-E2 and Stable Diffusion XL—which all use prompts to create photorealistic images and other kinds of artwork, like illustrations—when they first launched. But I’ve found generative pictures (especially the photorealistic variety) to be a flop in general. I’ve tried and tried, but never get anything close to usable. On first glance, the output usually looks good, but it’s always obvious, when you look a bit closer, that it was done with AI.

AI is particularly bad at rendering hands. Here is a funny collection of mishaps:

Forge & Spark AI content marketing AI handshakes

The lesson? Real, authentic photos and images always win—and AI is really a tool to enhance or clean up what you already have.

Our conclusion about using AI for content marketing: It’s a great tool, but it can’t do the job for you … yet

If there’s one theme coming out of our experiments with using AI for content marketing, it’s that AI is a great helper with a lot of potential—but it’s not a magic bullet. As a writer, designer, or marketer, you still have a responsibility to focus your efforts through the lens of strategic goals, audience, and brand voice when using AI to create branded content. And that simply isn’t something AI can do on its own.

Instead, think about all of these new AI options and platforms like an expanded tool-chest; full of handy helpers that can speed up your work, get you unstuck, expand, refine, and build on what you already have, and know deeply, as an expert in your field.

Our advice to other marketers and purpose-focused business leaders looking to navigate the AI revolution is to learn by doing: open ChatGPT, Jasper, or Midjourney (or the tool of your choice) and start doing some experiments of your own. No doubt this is just the beginning, and things are going to start evolving, quickly!

For our part, using AI tools in our content creation processes has ultimately brought us closer to achieving our purpose … of providing better content for our clients. And it helps to know that each one of these steps—win or flop—is progress.

Can we help YOU take the next step in your content-creation journey, AI-based or otherwise? Get in touch and let’s chat.

Articles You Might Like