The allure of editing is the opportunity it provides to be at the center of the relationship between the writer and the audience.

Editing teaches me to understand the writer’s mind — how they think, how they write, what they know, and how much of that translates to their computer screen. Having written and edited for the SaaS tech industry in the past, I can say that, to some extent, I understand writing; to some extent, I am also the audience.

The heart of an editor’s role is striking a balance between the author’s thoughts and what the target audience wants to read while ensuring writing quality. It is also crucial to contribute to team goals and meet our service level agreement (SLA). 

To guarantee quality and meet deadlines, it’s essential to set up efficient processes. Ask any editor, and they’ll likely say an article is never truly “finished” — it’s the deadline that decides its fate.

When AI became all the hype, with the promise of efficiency, I couldn’t stop thinking about its potential for editors. The nature of a new technology is scary at first, but that’s what I thought about Facebook (umm…Meta?) so many years back, and look where that is now. So, I decided to give AI-assisted editing a go.

I could say the rest is history, but then I’d be lying to you. But it does have a lot of promise, and I don’t say that on a whim but after rounds of experimentation trying to incorporate AI into editing processes at G2. 

And to answer the biggest question ever – will AI steal my job? Let’s find out. 

The promise of AI to transform content editing workflows

I manage the editing team at G2. On a normal day, my team of global editors works on and publishes a variety of articles written by in-house writers, freelancers, and external thought leaders. Efficiency-wise, the odds were never in our favor, but we’re a strong, talented team, so we persevered.

When ChatGPT became everyone’s new favorite content creator, G2 launched a program to do “ten percent more” with AI — an initiative to use AI to improve daily productivity. This motivated me to challenge the team and explore whether we could use AI to resolve our underlying problems of meeting our efficiency goals.

Setting up our test run

My expectation at that point was to employ AI to give some time back to my team so they could start thinking about the writer’s voice and audience needs and not just run through copy edits and customary line edits to fit our brand voice.

To this end, we started by setting up a test run.

Experiment round 1

No agenda-driven experiment can run well without a methodology in place. With the above efficiency challenges in mind, we set up the framework for our first round of experiments.

Every editor on the team was assigned a particular AI tool and a few articles from the editing pipeline. The ask was for them to edit these articles with the AI tool and document their learnings. 

The process was subjective. It was up to the editors how they wanted to use the AI tool and for what use cases.

Their only objective was to answer a questionnaire and support their answers with screenshots of the tool, as well as their perspective or understanding of how it went. 

Here’s what the questionnaire looked like:

So, what did we learn in the first round?

The good:

• AI helped improve the structural flow and logical progression of content, effectively contextualizing scattered ideas.
• It proved adept at filling content gaps and expanding on existing content, identifying possibly missed ideas that could be incorporated into the article, and serving as a good source of inspiration for framing sentences or when you are stuck creatively.
• AI was helpful in improving grammar and spelling accuracy.
• AI has the potential to save time during the content editing process, but it depends on the type and complexity of the content piece.
• AI-generated compelling, alternative title options and refined headlines.
• AI effectively edited content to match specific brand guidelines or editorial standards.

The bad:

• AI sometimes misunderstands the context, leading to incorrect suggestions.
• There were instances where the AI-generated content was robotic or unnatural.
• Integrating AI into existing workflows may be challenging. 
• Reviewing and verifying changes made by the AI tool is time-consuming.
• AI is not well-suited for contextual editing or long-form articles.
• AI seems better suited for writers than editors.
• At times, AI made unnecessary changes to the context or damaged the originality of the writing.
• Even after inputting a specific prompt, AI may rewrite the article instead of adjusting grammar or sentence structure.
• Certain editing tasks seem beyond the current capabilities of AI.

Yes, these are mixed results! 

While one editor said the AI tool helped build context gaps, another team member said it was unsuited for contextual editing.

But all in all, the results were definitely better than we expected.

Experiment round 2

For our second round, we moved from a subjective to a more objective process. 

Editors were instructed to use AI in every step of the editing process — from the first round of style edits to developmental editing to final proofreading. And while they did so, they were required to note the time it took them to complete the editing round. 

Here’s an example of what it looked like:

We followed the same technique for analyzing the AI tool’s work. This included prompts related to improving article introductions, suggesting changes to sentence structure, and proofreading.

What did we learn in the second round?

• AI did not help reduce the time taken to edit but helped bring down effort related to ideation.
• AI can help with audience persona alignment and generate an ideal customer profile (ICP).
• AI takes the cake with proofreading.
• AI helps identify the use of unnecessary metaphors, misplaced jokes, awkwardness, and redundancy in the article and helps change the language to fit the desired brand voice.
• AI can scan for sarcasm and help make the text more constructive.
• AI can identify absolutes in text, such as “never,” “always,” etc., and suggest changes to the sentence.

Compared to round 1, this round gave us more promising results, leading us to conclude that AI could help us in more ways than one.

But there was still a long way to go.

Establishing use cases

After two rounds of experimentation with two different methodologies, we now had a pretty solid idea of what AI could and couldn’t do. Despite limitations, it was becoming increasingly clear to me that AI could be the messiah of our efficiency challenges.

This meant that it was time to delve deeper and conduct at least another round of experimentation. But this time, we needed a more acute structure to our methodology in order to figure out AI’s specific use cases in editing.

Experiment round 3

The third round consisted of creating a template with a prompt for each use case, which can be used in every sub-step of the editing process. Each prompt we used in this stage was tried and tested during the first two rounds of experimentation to provide the desired results.

The template was also accompanied by a set of guidelines for the editors, which were informed by our learnings. The guidelines included which AI tool to use for certain use cases. For example, while Google’s Gemini could be used to find contextual gaps, ChatGPT would be helpful to ideate better article introductions. This documentation would serve as a singular set of guidelines (or source of truth) for the team.

Once the template was ready, the editors reviewed and ran the prompts with the suggested AI tool. The final step was to share their thoughts on the above as a team and resolve any roadblocks.

Here are some examples of what it looked like:

The editors could now just copy-paste the standard set of prompts into the AI tool and measure efficiency gains or losses.

Moment of truth: experiment results

Now it’s time for the big reveal.

However, it is important to acknowledge that an efficiency gain to this extent is not possible after one or two iterations. It may also not be true for all cases. We’ve had several instances where an editor spent more time with AI than they would with manual editing.

But for now, let’s celebrate this win.

Key takeaways

AI does a great job of looking for gaps in writing. But it may, every once in a while, identify a gap that isn’t there. More often than not, AI will change the voice of the writer, making a human’s voice sound like robots from Skynet days. AI will almost definitely start your articles with “In the ever-evolving business landscape” or worse, turn phrases like “employee happiness” into “laborer’s prosperity.”

The trick is to find what works for you. We found our best prompts after working through several bad ones. The best use case of AI is what you can make.

My final verdict on using AI in editing is that it does resolve our efficiency challenges. But it may not always give us the best path forward. Editors need to use their best judgment at all times. 

It does, however, become the tool that we didn’t know we needed. 

With AI, we can now confidently meet a strict deadline and adhere to our SLA. AI also helps us get “unstuck” sooner. While editing, we’re often stuck thinking if there are gaps in the article we didn’t find or if there are better ways to structure a sentence.

AI helps speed up ideation and, hence, improves efficiency.

Based on our learnings, the following AI content guide highlights the only use cases where we found AI helpful with editing. But bear in mind that it depends on the type and complexity of the article.

Download your AI content editing guide!

So, will AI steal my job?

Short answer: not yet. AI doesn’t have a personality yet.

While AI may excel in grammar checks and filling structural gaps, it doesn’t have original thoughts, and that’s the edge, the best, and the biggest thing we, as humans, bring to the table.

AI cannot replace the human element of being able to identify context, nuance, sarcasm, voice, and intent.

The good, the bad, and the ugly of AI-assisted editing

When I first asked my team about working on an AI experiment, I started wondering if I’d make AI mandatory or optional to use if the experiment was successful. Now that I’m here and we’re successful, I’ve chosen to keep it optional. In creative processes, freedom is essential. 

While we discovered the value AI brings, integrating it into existing workflows was not without challenges.

There is a huge learning curve with any new technology. Every editor has their own unique style of editing. Incorporating a tool into that workflow means disrupting their creativity. So it takes some time to get used to this change. But once your experiments bear fruit, it is difficult to go back to entirely traditional editing processes. 

More than that, what is needed is perseverance.

It is not an easy job to keep going at it even after rounds of experiments fail. Using AI as a tool to enhance and not replace their existing skills and the desire to do ten percent more, editing teams could definitely find a win with AI at the end of it all.

Looking for more info about how AI is changing the world of content? Learn more about the future of AI in writing.