When I used AI as ‘just a writing tool’, I didn’t give in to the temptation and I stubbornly stuck to starting every piece old school: with pen and paper. Handwriting the title in my own words. Scribbling down the reader’s problem, my solution, and why the reader should care. Once that was set, I opened AI and found the rest while drafting. By this rather old-fashioned start, I safeguarded the essence and my point of view. 

→ Now that my workflow is built around AI, all of that thinking happens before AI writes the first word. My work moves from writing to thinking.

For the ‘9 AI dilemmas’ blog series, with AI at the core of the process, I spent about 4 hours on strategy and setting up, 3.5 of them sparring with AI to sharpen my thinking. 

With that in place, the production of each piece in this series takes 3 to 5 hours from briefing to publishing, and my bar is much higher than when I wrote alone. 

For a client team, the same setup cut a blog from 12 to 16 hours to 4 to 6.

How did I achieve this? I moved the sharpening that used to happen while writing to the front, on purpose.

How you use AI decides quality and time savings 

The visual below shows the same piece of work in different stages: before AI, the expectations, AI bolted onto the old workflow, and AI at the core. There are 10 blocks of work as the starting point. All is illustrative, based on experiences with corporate clients and for my newsroom. 

1. Without AI, the workload sits in creating. 

2. Working with AI: the expectation is a big time saving. In this picture, 40% or more.

3. If you bolt AI onto your current workflow, the time AI saves on writing evaporates in reviewing: round after round of quality checks to stop it sounding generic. 

In this picture it saves the least time, 10%, and it risks serious quality issues: shiny on the outside, empty on the inside.

4. With AI at the core, the workload moves upfront: to configuring and thinking. That’s the homework. If you want to keep the quality high, you invest time before you save it. The first rounds can cost more than they save, especially the first time you build the setup.

The more you use it and give feedback, the better it gets, and the faster you get to work that’s high quality and on brand. At scale, the investment pays back.

4 hours of setup, once, for the whole series

For this ‘9 AI dilemmas’ series, I started by building a project in Claude. In Copilot I’ve built an agent for a specific campaign, which we share with the team. The idea is the same: one place where everything lives and links, a space to come back to, like a SharePoint folder for your files. 

Then I sharpen the topic with AI: the strategy, the message, and the thinking underneath it. From the objective, the reader’s problems and how I solve them, the number of pieces. And the structure per piece to maintain consistency. I wrote down the detailed tone of voice, and how AI should work with me, pushback included. 

I also created the visual brief and stress-tested it. I built all the checks, for another round of pushback, sanity check, and fit for my reader. And I set up the tracker to keep the overview and status, connected to the tools; Notion and Canva for me, SharePoint and Word for client work. 

Time spent: 4 hours in total, once, for the whole series. 3.5 of them were thinking.

Four hours is fast. The first time, expect a few days.

Before, all of this lived in my head and took shape while working. Now it has to be written down, in words AI understands, before any of the production work starts. 

There’s a catch. My best ideas often turn up on a run. I used to find the right words while I wrote. Now I find them in the reasoning with AI. Skip that step, and the idea stays in my head while the draft drifts.

Every piece starts with a debate

Before I start with a piece, I ask AI to suggest the exact topic, based on the strategic plan and what we’ve already created, but I decide.

Then the reasoning starts: 1 to 2 hours of arguments between me and AI is quite normal. Because I asked for pushback in the setup, I get it and AI doesn’t spare me. The title and bottom line of the piece always come first. If those are sharp, the rest flows. Otherwise, the article drifts.

Something I didn’t expect: my bar went up. I’m stricter on message and topic than when I wrote alone. I can’t fully explain why. Maybe because I have a knowledgeable sparring partner I can be brutally honest with, and trust. That only works because of the setup.

AI and I write together, and I still make the visuals by hand

AI follows the process I set at the start: based on the title and bottom line, we write the outline first. When that is set, we write the draft. We rewrite and sharpen it together. In the last step, we run the checks. 

→ Outline, draft and rewrite takes 1 to 2 hours.

For the visuals, AI helps me to write the brief. I built in a feedback loop: AI checks the visuals of earlier pieces for consistency, because things change while working.

I design everything by hand, in Canva. I like this manual work, and AI-made visuals always lack a heartbeat. 

To be honest, I’m also not that good at instructing AI on visual design, let alone getting it to adjust what it made. So I love working hybrid. Besides, Canva makes it so quick to resize for my website, LinkedIn, Medium and email.

The secret: AI runs the checks before anyone reviews

My drafts reach stakeholders close to approval. This is because of the checks that I built into the process. 

• One looks for flaws in the reasoning
• One checks the fit with my audience
• One reasons from my stakeholders’ perspective 

The biggest difference came from that stakeholder check. Before anything goes for approval, the agent answers one question: what feedback would the stakeholder give? The review happens before the review.

Last step: publication

The piece then lands in Notion automatically, through a process I set up at the start. I check it one more time by hand. 

Then the publishing happens in 30 minutes, manually by me, in the CMS, optimising the copy for mobile and desktop. Around 20 mins go to design: visual anchor points that make the copy easy to skim and skip.

After every piece, the setup gets better

After publishing, I go back to the setup. I ask AI to check the latest status, and to update its memory and settings where needed. 

→ The configuration is an ongoing process.

The same setup works for a team

What I do for my newsroom by myself, also works in an office and team setting. For one of my clients, I built a campaign agent with the same setup, and shared it with the whole team. 

• Before, a blog or client case took around 12 to 16 hours from brief to approval, spread over 2 weeks to 6 months.
• Now with AI, it takes 4 to 6 hours, done in 2 days. Configuring the AI and process took about 4 hours: 2 upfront, the rest in small updates as we learned.

My advice

The sharpening that used to happen while writing now has to happen upfront, on purpose. Skip it, and nothing catches it. AI turns a thin brief into a fluent draft every time, and you end up with something anyone, or any AI, could have written. That’s not a time saving. That’s a brand problem.

There’s a quieter cost too. What you don’t make yourself doesn’t stick. You know the work less well, so you defend it less well: in the meeting, with the client, in front of the board. That’s why I still spend hours reasoning before a word is written. The thinking is where the piece becomes mine.

This also takes skills. You have to learn how to talk to machines and how to brief properly. My setup is quick because I’ve failed and learned. Working with AI isn’t buying a tool. It’s changing how you work.

If you want this setup for your team, I’ll build it with you, and teach you how to do it yourself and how to manage it. 
 

9 AI dilemmas

This piece shows the work behind the 9 AI dilemmas sfor the C-suite. Each dilemma names the decision, what it costs to leave it open, and what I’d do.

The one you recognise immediately might be the one you’ve left open too long.

1. Faster with AI means more errors
2. Secure AI means weaker AI
3. Your board wants AI to make marketing cheaper. Your brand can't afford it

Join 2,750+ professionals

AI changes how the work gets done. I test it in my own work and write about what actually works.

Free. Unsubscribe anytime.

→ Subscribe

You can do both, by design: save with AI and stay distinct.

1. Obsess over configuring AI on your point of view, for everything you make. Then decide at management level how much human work each piece gets, so the team's workload doesn't decide for you.
2. Where little is at stake, let AI produce it fast. Where your name, reputation and conversion are on the line, more human time goes in, and that's where the savings go.

This flips at both ends:

• When customers buy on price and availability, AI can produce most of it fast.
• When your thinking is what you sell, nearly every piece needs that human time.

Right now, the savings target decides

Marketing budgets aren’t growing. In Gartner’s 2026 CMO Spend Survey, marketing spend went from 7.7% of company revenue to 7.8%, while CMOs were still expected to grow the business, work more efficiently and make AI work. 56% say they lack the budget to deliver their 2026 strategy.

So the savings target is already in the budget, while the workload and deadlines stay the same. Every piece gets the fast AI treatment, including the ones that carry your reputation. Nobody chose which pieces get more human attention.

Hit the savings target with a basic AI setup, and you sound like everyone else. Hold AI back, and you lose the budget fight.

Taking the savings is a reasonable call

The rest of the board expects marketing to save budget by working with AI. The return on a distinct brand is hard to prove. And competitors are cutting too, so why pay more to stand out?

Most content is skimmed, not read. Customers rarely notice the difference between good and very good. Let alone your brand voice.

Fair, for content nobody chooses you for. But where people decide whether to trust you or buy from you, good enough is how you disappear.

AI doesn’t have to sound generic

Many leaders think it does. So they see two options: cheap and generic, or human and expensive.

Few know there’s a third: 

→ AI configured on a clear point of view. It takes time and skills most teams don’t have, and a point of view written down in plain language. That last part is usually missing.

AI can copy your tone, not your point of view

Many teams give AI their brand book and a few adjectives, like bold, human and authentic. But every brand uses those words. So what AI writes sounds generic.

What you need is to feed AI with your point of view.

A point of view is what your company believes about its field and its customers' problems, especially where it disagrees with how things are usually done. It's a stance on something contested: what you think, and what you'd argue for in public. If nobody could disagree with it, it isn't one.

Your point of view also works as a filter, twice.

• Before you produce anything, it tells you what to say yes or no to.
• Afterwards, it's your quality check: does this fit what we stand for?

Without a point of view, every decision gets made from scratch, by the loudest voice in the room or by the deadline.

When I configure AI for a client, it takes me up to two weeks. More than 10 documents, written so AI agents can reuse them, and a lot of testing before I sign off. First for the brand, then for the people who speak for it. Very little of it is a 1:1 copy-paste from the brand book.

When I write for a CEO on a strong configuration, 95% is approved with a few small tweaks. Without it, it's review round after review round.

Not every piece deserves the same hours

Some content decides whether people see you as the expert, trust you or buy from you: the homepage, the campaign, the pitch, what your CEO publishes.

Other content just has to be correct and on brand: product specs, service answers, frequently asked questions. AI works on both. The difference is how much human time each one gets.

In many teams, that difference doesn't exist. The same people give the CEO's opinion piece the same time as a product description. Or both come out of AI with the same quick edit.

→ Where to invest more human time is a management call, made once. Not a deadline call, made by the team every day.

Spend the AI time savings, don't bank them

The time AI saves on volume content pays for more human time on the content that matters most. Spend it there. Not on cutting the team, and not on publishing more.

AI saves around 40% of your time at first. Done right, about 10% ends up as savings. The rest goes back into configuration and the thinking that makes you distinct. Your board still gets savings, just smaller than the spreadsheet promised.

Gartner's survey points the same way. Even though AI is assumed to cut people costs, labour's share of the marketing budget rose from 21.9% in 2025 to 24.5% in 2026, a sign that AI's value depends on people, skills and execution. And what holds CMOs back most is a lack of AI expertise and talent in-house, cited by 38%.

→ The ROI isn't in the hours saved. It's in what you do with them.

Start here

Write down your company's point of view on what matters most to your customers. Five sentences, in plain language, without the brand book.

Then show it to a colleague on the board. If they'd write something different, or it's new to them, you've found what your AI is missing.

Cheaper where you can, distinct where it matters

Right now, the savings target makes that call. It should be yours.

If your board expects savings and your brand can't afford to sound generic, I'll help you write the point of view and set up AI to work from it. Email me directly.

9 AI dilemmas

This is the third of 9 AI dilemmas for the C-suite. I'm publishing them over the coming weeks. Each piece names the decision, what it costs to leave it open, and what I'd do.

The one you recognise immediately might be the one you've left open too long.

1. Faster with AI means more errors
2. Secure AI means weaker AI
3. Your board wants AI to make marketing cheaper. Your brand can't afford it (this page)

Sources

• https://www.gartner.com/en/newsroom/press-releases/2026-05-11-gartner-2026-cmo-spend-survey-finds-cmos-allocate-15-point-3-percent-of-marketing-budgets-to-ai-but-only-30-percent-are-ready-to-scale-ai-capabilities
• https://www.gartner.com/en/newsroom/press-releases/2026-06-08-gartner-marketing-survey-finds-awareness-and-conversion-account-for-62-6-of-total-media-spend
• https://www.thinklikeapublisher.com/use-ai-to-create-better-content-not-just-more/

Join 2,750+ professionals

AI changes how the work gets done. I test it in my own work and write about what actually works.

Free. Unsubscribe anytime.

→ Subscribe

Stop pretending Copilot is the best AI. Sell it on what it is: the same AI as ChatGPT and Claude, with a Microsoft layer around it. That layer makes it feel slower, but it's where the value sits: your email, files and meetings, inside Word, Excel and PowerPoint, safely. 

The one-hour training everyone got never showed that. Learned properly, Copilot comes close to working with ChatGPT and Claude. Not equal, but close, and it keeps your data in.

• 85% of employees can’t apply AI training to their actual jobs.
• Nearly 60% say the programmes aren't designed with people like them in mind. (source: Docebo)

When a team needs what Copilot can't do, buy them the business version of an AI tool that can, with the same data protection.

The legal consequences

Right now, every employee decides where your data goes, one prompt at a time.

And the legal part lands on you. Since February 2025, the EU AI Act requires you to support your people's AI literacy and keep a record of what you did. When personal data ends up in a public AI tool, the company answers for it under GDPR, not the employee.

Buying the tool your employees like sounds fair

Some leaders skip the Copilot licence and the expensive training, and buy the business version of ChatGPT or Claude. People already like it and know how to work with it. The contract protects the data. And no AI tool fits every job anyway, so why force one?

Fair. But in the companies I work with, the approved tool is almost always Copilot or Gemini. The platform you already run on made that choice for you, and undoing it is expensive: a second AI tool means a second contract, a second data agreement and a second training.

→ So what matters is whether your people learn what the one you've got can do. And it works differently from ChatGPT or Claude.

Out of the box, Copilot loses

Most people met AI at home first. In PagerDuty's 2026 survey, 89% of office professionals who use AI for work first came across the tool in their personal lives. So they learned ChatGPT or Claude: open a chat, ask, paste.

Then they try the same thing in Copilot. It feels slower. The answers feel thinner. The language feels generic and corporate. Back to the familiar tool. The same survey found that two-thirds have used AI tools at work they believed weren't allowed.

And let's not sugarcoat it: ChatGPT and Claude feel ahead. People aren't comparing the AI. They're comparing Microsoft's packaging, your company's settings and how well they learned to work with it.

→ Judged as a consumer AI tool, Copilot loses. That's the wrong test.

Its power sits in your own systems

Copilot's strength isn't the model. It's what it can see: your email, your Teams meetings, your files in SharePoint. It works inside Teams and Outlook, and moves between your apps without you copying anything across.

Ask it to prepare a meeting from the email thread, last week's notes and the deck in SharePoint. Ask it to find a file, an email or a vague conversation from months ago. A personal ChatGPT account doesn't, because it sees none of it.

You think your people are trained, but they never saw what Copilot can do

That power stays hidden until you learn the deeper layer and change how you work: setting it up for your work, digging into the corners of SharePoint and your inbox, using it as a sparring partner on confidential work. The more you work with it, the better you get, and the more of your work it has to draw on.

Standard training doesn’t get there. Usually it’s one session for everyone: the novice, the sceptic and the power user, finance and marketing. That’s how you end up with the 85%.

Bespoke training costs money. And you can’t see the value you’re not getting, so it rarely makes the budget.

→ A training for everyone fits nobody.

You never have to think twice before you paste

When I work in Copilot, I don't think about what I paste. When I use ChatGPT or Claude, I stop every time and check what's in it. For confidential business work, I don't use them at all. I don't want to be the one who causes a leak by accident.

Everyone knows they should pause. Nobody plans to skip it. Then, on a busy day, they paste the one document they shouldn't have.

Safety that's always on is the part that's undersold. It works before anyone's trained.

Start here

Ask three people at different levels one question: what do you use Copilot for, and what do you still do in ChatGPT or Claude?

If the first answer is meeting summaries and the second is everything else, you know two things. Where your data goes. And what your training missed.

For most of your people, close enough and safe is the better trade. Until they learn it, the tab they keep open decides where your data goes.

If your people still go around Copilot, I'll help you build training that fits their work, and be honest about what Copilot can't do. Email me at fleurwillemijnvanbeinum@gmail.com.

9 AI dilemmas

This is the second of 9 AI dilemmas for the C-suite. I'm publishing them over the coming weeks. Each piece names the decision, what it costs to leave it open, and what I'd do.

The one you recognise immediately might be the one you've left open too long.

1. Faster with AI means more errors

2. Secure AI means weaker AI (this page)

3. Your board wants AI to make marketing cheaper. Your brand can’t afford it

Sources

• https://www.pagerduty.com/newsroom/shadow-ai-workplace-survey-2026/
• https://www.docebo.com/company/newsroom/docebo-releases-the-ai-readiness-gap-report
• https://www.whitecase.com/insight-alert/eu-ai-omnibus-enters-force-amending-ai-act
• https://learn.microsoft.com/en-us/microsoft-365/copilot/microsoft-365-copilot-application-card
• https://www.thinklikeapublisher.com/faster-with-ai-means-more-errors/

Join 2,750+ professionals

AI changes how the work gets done. I test it in my own work and write about what actually works.

Free. Unsubscribe anytime.

→ Subscribe

36.5% of marketers say incorrect AI content has already gone public. And 83.5% are expected to produce more because of AI.

My advice: build it into the process

Build quality into how your team works with AI. Put serious time into the setup before AI starts writing, so it works from your facts, your sources and your voice. Then add multiple checks along the way, not only at the end.

This slows you down at first. After that, you're faster and safer with every piece. 

→ Quality control shouldn't be an editing step. It should be part of the AI workflow itself.

Yes, the setup slows you down, for a few weeks

You bought AI to go faster, and now someone tells you to slow down first. That's a hard sell when your AI investment has to pay back this year.

But the pace changes once your setup is in place. You build the setup once and use it for every piece after. The more you publish, the less manual work each piece costs. 

If you only produce a handful of pieces, it flips. Then the setup costs more than it saves, and one very careful check at the end is enough.

More than 70% of marketers already spend one to five hours a week fact-checking AI output. A team that checks everything by hand at the end pays the full price every time.

→ The setup is a cost once. Skipping it is a risk every time.

AI errors are predictable

AI doesn't fail at random. The same kinds of problems show up again and again. AI tends to fail in five predictable ways:

• makes things up
• leaves things out
• uses outdated information
• answers a different question than the one you asked
• uses language your brand would never use

That's good news. What you can predict, you can check for. That means replacing a vague "have a look before it goes out" with a specific check for each failure mode.

→ Predictable errors are preventable errors.

Checking only at the end is too late

Many teams check AI content at the end, if they check it at all. By then, an error is baked into a finished piece. Catch it, and you have rework nobody planned for. Miss it, and it goes out under your name.

Put the time in earlier. Configure AI before it writes and build checks into the process, then check again just before publication.

Skip those checks and you get cases like DPD, whose chatbot swore at a customer, or lawyers filing case law that AI made up. In both cases, AI-generated output reached the outside world when it shouldn't have.

With my clients, we haven't had an AI error make it through to publication. Not because AI doesn't make mistakes. Because we double down on configuration and checking throughout the process.

Start here

Ask your team to walk you through one piece made with AI: from the setup, to the brief, to the checks, to publication. 

Then ask one question: does anyone ask AI to check and challenge what it wrote, or do we only check the links?

If it's only the links, quality isn't built in. It's hoped for. 

Fast and safe is a decision

Speed and safety can go together in a process that’s built for both. Set it up once, use it often, and check your setup regularly, especially when the AI or model changes.

Without that process, deadlines and workload decide how much checking gets done.

If you want to move fast without compromising on quality, I can help you build the process. Email me directly.

9 AI dilemmas

This is one of 9 AI dilemmas for the C-suite. I'm publishing them over the coming weeks. Each piece names the decision, what it costs to leave it open, and what I'd do.

The one you recognise immediately might be the one you've left open too long.

1. Faster with AI means more errors (this page)
2. Secure AI means weaker AI
3. Your board wants AI to make marketing cheaper. Your brand can’t afford it

Sources

• NP Digital, AI Hallucinations and Accuracy Report (February 2026)
• HubSpot, State of Marketing 2026
• DPD chatbot incident (January 2024)
• AI hallucinations in court filings

Join 2,750+ professionals

AI changes how the work gets done. I test it in my own work and write about what actually works.

Free. Unsubscribe anytime.

→ Subscribe