Working with AI in your content workflow? Here are 3 things that help

The best teams don’t bolt AI onto old systems, they start fresh

The best don’t ‘use’ AI to speed up work; they redesign from scratch and then use AI to accelerate. If you use AI as a lazy shortcut, you’ll get shortcut results. Use AI as a partner, and suddenly, your work proves you give a damn.

→ Craftsmanship and AI aren’t enemies. They’re better together. Mediocrity is your real enemy.

Let’s look at the current state of AI, then explore the 10 content stages and 2 other key lessons.

Break your work into smaller steps and decide where to use AI and where not to
You need a project manager. Working with AI adds steps to the process as you also increase scale and volume
Most of the work is in the AI configuration and keeping it up to date

Why it matters

If you want to get the most out of AI, use AI to transform your work, not to automate it. We all should know by now that the true power of AI lies in being your partner, not just a tool.

→ But most companies still try to bolt AI into old workflows. That costs you time, instead of saving it. And you’ll be missing out on a huge opportunity to improve the quality of your content.

Now is the time to move from experiments and pilots to rewriting how the content job gets done.

When done correctly, you should be able to create amazing content in under 4 to 6 hours. For my own content, I go from ideation to publication in 2 to 4 hours.

The real status of AI in 2026

The gap between AI hype and AI impact has never been bigger.

90% of companies claim they ‘use AI’, but 67% still stuck in pilot mode
64% say AI boosts innovation, but only 39% see real EBIT gains
The top 6% think differently. They don’t ‘add AI’, they rebuild workflows, roles, and processes

→ The winners don’t ‘use’ AI to speed up work; they redesign from scratch.

Credits: McKinsey 2025 AI report, via WizofAI.

Now we know this, let’s dive into the 3 lessons I learned while working with AI for 3 years.

Lesson 1. AI only works if your process is built for it

In general, there are 10 stages in the content process. For the sake of clarity, I’ll refrain from diving into the rabbit hole and all specifications. This is not a detailed how-to guide for producing content.

Rule of thumb: the more AI you use in the first stages, the more time you’ll spend in the later stages reviewing and correcting to get something more or less decent (read: still mediocre content) The more time you spend in the beginning, the more you can speed up in the end with AI, while you also use AI to improve the quality.

	What you do:	If you can use AI:
1.	Strategy and goal setting Start with purpose and clarity. Know what you're saying, why you're saying it, and who it's for. Without purpose and clarity, you're just typing with ambition and hoping the internet does you a favor.	You can use AI for this, but honestly, without you doing the thinking, you’ll get mediocre output.
2.	Research and topic development Do your homework. Read, listen, scroll for information, opinions, and different angles.	You can go full AI as a shortcut and prompt your way into it. I prefer to combine my own mind and craft with technology to add depth and meaning.
3.	Structure before writing: outline Even Ann Handley says it: structure first, draft second. Don’t freestyle; start with a clear outline.	Here it depends. If I (ghost)write for a client, I use AI for the outline to be more efficient with time. If I write for myself, I still do it by hand with pen and paper, because it slows down my thinking and improves the quality of the content. → For client work, get approval on the outline before moving to the next step. This is why you need a project manager, see lesson 2. → Also important, if you use AI, you must have configured it to think and act like you, see lesson 3.
4.	Writing the first ugly draft It’s not supposed to be good; it’s called the first ugly draft for a reason. Don’t wait for brilliance. Just get it down, and you’ll fix it later.	Here, again, it depends. For a client, I use AI and ask for approval again before starting the time-consuming next step. For myself, I trust good old Word, and I just start typing.
5.	Editing and refining with precision Kill your darlings and be the devil’s advocate if your content still aligns with the strategy and structure, steps 1 and 3. This is where you turn messy into meaningful. Cut the fluff, sharpen the message, punch up the flow. If it doesn’t serve the reader, kill it. With love. And now it’s also the moment to add your visuals and or infographics to see the full result.	This is where the craftsmanship shows, and the content hackers will reveal their true colors, getting stuck in mediocre, generic content. And this is also where I go fully AI. But only because I’ve configured it and created my bespoke proofreading GPTs; I use 3 to 4 different ones, each for very specific tasks.
6.	Review and approval Even the best content won’t survive 10+ rounds of feedback. And if your content needs 6 or more approvals, your marketing mandate might be broken.	No AI involved here. This is pure stakeholder management.
7.	Design and formatting, obsessively Don’t bury great writing in lazy design. Use formatting, spacing, and white space to help your message land, not hide. Do this and review your work exactly as your audience sees it. Always ‘in situ’, copy combined with visuals, most likely on a mobile. Never do this final review of your copy in a Word document. It's best to do this on ‘staging’ in your CMS.	Again, no AI. This is more craftsmanship and a profound understanding of communication, marketing, UX, design, and psychology on a small mobile screen. Note, you can use AI to fill your CMS, but you still need the manual work for reviewing, correcting, and editing. Therefore, I prefer to do it all by hand; it’s faster, and you can do part of the design and editing while filling the CMS.
8.	Final sign-off This is the last stop before going live. If you need approvals, get them now or forever hold your passive-aggressive voice.	Again, no AI involved here. This is more of pure stakeholder management. If you did a great job in step 6, this is easy and just to cover your ass for one more time.
9.	Publishing	If you did your job as you should in step 7, it’s just a push of the button.
10.	Adjust for other channels Content formats differ; adapt accordingly in copy and in design.	Yes. You can use AI for this. If your structure is solid, your final content is top-notch, and your AI is configured correctly, this is the easiest part.

Ethical and quality guidelines

An important note about ethics, brand, and data safety is that you have to create your guidelines.

The bias in AI is real; it still hallucinates; online manipulation is around the corner; and, not to mention, the unintended, unaware sharing of confidential information with ChatGPT.

Lesson 2. You need a project manager

This is one of the toughest lessons I’ve learned in 3 years. The client and I were both surprised by the importance of having a project manager in content production; we both didn’t expect it, and underestimated the impact of the scale and volume. Twice.

→ Using AI for content adds steps to the process, see above. And at the same time, you want to scale and turn up the volume of your content.

If you work with just a few content pieces, you’ll think you can do without. But as time passes and more content is produced, we’ve learned you can’t keep track anymore.

You really need to keep the overview to stay accountable and make progress. And that’s a project manager’s job.

Lesson 3. AI won’t help until you configure it

This is indispensable and goes much further than just adding your tone of voice to the settings of ChatGPT or any other tool. It’s about your values, vision, depth, rhythm, tone of voice, and lexicon.

But also, about adding your content structure so you always have the same setup, building your GPTs for proofreading, improving, and evaluating based on personas and audiences.

→ I spend about 2 weeks on AI configuration, testing thoroughly with the team, and asking for blunt feedback so we can optimize it together.

And once you’ve configured it, the real work starts.

Having your configuration in place means keeping it up to date. Not only because you and the team learn and grow together, but also because the technology matures. With every AI update, things change. And what works today can feel ‘off’ tomorrow.

In practice, this means you update and or refine your settings at least once a month. These can be minor tweaks, bigger refinements, or complete rebuilds

Summary

If you really want to get the most out of AI and prepare yourself for the future, you redesign how the work gets done.

The prerequisites, based on lessons learned:

Break the job down into smaller pieces. And check where you can use AI and where it’s better to keep the human craftsmanship in place.
Get a project manager on board. As you add steps to the process, you also scale and turn up the volume.
Obsess on your AI configuration and update that at least every month

Sounds like a lot of work and investments?

Yes, it is. But it pays off in the quality of your content and the happiness of your team.

Or you choose to use AI as a lazy shortcut, but don’t get mad if you get shortcut results.