TL;DR
- AI should improve the quality of work, not just the speed at which it's produced.
- Generating content isn't the same as applying expertise.
- Poor AI use often transfers work from the author to everyone who has to review, interpret or correct it.
- The best results come from combining AI capabilities with human judgement.
- If you can't explain or defend the output, the work isn't finished.
- Ownership remains human, even when AI contributes to the work.
There is a strange thing happening in workplaces right now.
We have access to tools that can help us research faster, explore ideas, analyse information, challenge our thinking and turn rough concepts into something useful in minutes.
And sometimes we’re using all of that capability to produce… more words.
More documents. More slides. More emails. More content. All perfectly formatted, professionally worded and somehow saying remarkably little.
Most of us can recognise it by now. It looks professional and sounds convincing, but it’s generic, repetitive and full of words that don’t really add anything. You get to the end and realise you’re not actually sure what it was trying to tell you.
AI slop.
The problem isn’t that AI was used to create it.
The problem is that somewhere along the way, the human stopped doing the work. Or worse – the human stopped doing the thinking.
What we're seeing in hundreds of organisations
Over the past few years, I've had hundreds of conversations with organisations exploring AI, Microsoft Copilot and modern workplace technologies.
The businesses seeing the greatest value from AI are rarely the ones trying to automate everything.
Instead, they're using AI as a thinking partner.
They're using it to explore options, identify risks earlier, challenge assumptions and accelerate understanding.
The organisations struggling with AI adoption often have the opposite approach. They focus on generating more output without improving the quality of the thinking behind it.
That distinction matters more than most people realise.
AI doesn’t remove the need for expertise
Sophie Alpert recently wrote an excellent piece called There are no lossless transformations of natural-language text. One of the ideas that particularly resonated with us was that writing isn’t simply the process of transferring words onto a page.
Writing is part of thinking.
In professional work, that matters.
When a client asks us for a strategy, recommendation, technical design, roadmap or proposal, they’re not really paying for the document. They’re paying for the thinking represented by it.
And the same principle applies well beyond consulting.
A product roadmap represents choices and trade-offs. Technical documentation should reflect an understanding of how something actually works. A marketing campaign should contain an insight about its audience. A sales proposal should explain why this solution makes sense for this customer.
AI can help with every one of those things.
But AI is remarkably good at producing the shape of professional work.
That’s not the same as doing the professional thinking behind it.
The Transferred Work Principle:
AI hasn't eliminated effort. It has simply moved that effort somewhere else.
When AI generates content that is generic, unclear or poorly validated, the burden doesn't disappear.
Instead, it gets transferred to:
- Colleagues trying to interpret it
- Managers reviewing it
- Project teams correcting it
- Customers challenging it
- Stakeholders making decisions based on it
Many organisations measure AI productivity at the point content is created.
The better measure is the total effort required after that content is shared.
If five people spend thirty minutes interpreting a document that took ten minutes to generate, the organisation hasn't gained efficiency.
It has simply moved the work.
This is one of the most important principles organisations should consider when assessing the success of AI adoption.
The sweet spot is human + AI
We use AI extensively at WebVine, and we want our people to use it.
The interesting question for us isn’t whether someone used AI.
It’s how they used it.
Some of the most valuable conversations we have with AI don’t result in us using a single sentence from the response.
We might start with a half-formed idea and ask AI to help pull it apart.
What are we missing?
What assumptions are we making?
Give me three alternative ways of looking at this.
Find evidence that supports this argument. Now find evidence that contradicts it.
What would a CIO object to here?
What questions haven’t we answered?
Compare these options.
Challenge my recommendation.
That’s a very different relationship with AI from giving it a task, waiting for the output and moving on to something else.
For us, good AI-assisted work looks more like a conversation than a transaction: frame, explore, develop, interrogate, validate, own.
The WebVine AI Ownership Framework
At WebVine, we've found that effective AI-assisted work follows a repeatable process.
- Frame the problem and provide the context.
- Explore: Use AI to explore possibilities, research and challenge assumptions.
- Develop: Let it help develop and structure the work.
- Interrogate what comes back. Look for weaknesses, generic thinking, contradictions and missing evidence.
- Validate anything that matters.
- Own: Then edit it. Apply your professional judgement. Make the decisions. Own the result.
There’s a simple test we can apply before sharing AI-assisted work:
If somebody challenged me on this in a meeting, could I explain why it says what it says?
If the answer is no, the work isn’t finished.
This is really about ownership
At WebVine, one of our values is Ownership. The more we use AI in our own work, the more relevant that value becomes.
Ownership doesn’t mean doing everything yourself. Using the tools available to work smarter is part of doing good work.
Ownership means being accountable for the outcome.
If AI helped create the document, recommendation, code, analysis or campaign, you still own its accuracy, reasoning and quality.
If someone asks, “Why are we recommending this?”, the answer can’t be “because Copilot suggested it.”
AI can contribute to the work.
It can’t take accountability for it.
That’s still ours to own.
This is also consistent with Microsoft’s guidance on Copilot: AI-generated output needs to be reviewed and validated before it is relied upon or shared. The person using the technology remains responsible for how that output is applied.
This aligns closely with broader industry research.
Microsoft's Work Trend Index consistently highlights the growing role AI is playing in knowledge work and decision-making. However, Microsoft also emphasises that AI works best when combined with human judgement, domain expertise and organisational context.
Research from Stanford's Institute for Human-Centred Artificial Intelligence (HAI) similarly reinforces the importance of human oversight in AI-assisted work, particularly where accuracy, trust and decision quality matter.
This reflects what many organisations are now discovering. AI can accelerate work significantly, but it cannot replace accountability, professional expertise or critical thinking.
Bad AI use isn’t efficient
Accuracy is the obvious risk.
Generative AI can be confidently wrong. In technology consulting, where products, licensing, roadmaps and supported approaches change constantly, plausible-sounding advice can have real consequences.
But there’s another risk: trust.
Our clients use AI too.
They know what generic AI content (aka AI slop) looks like.
If they’re paying for professional expertise and receive something they could have generated themselves in five minutes, that reflects on both the person who produced it and the organisation whose name is on the work.
The same problem exists internally.
AI slop creates a hidden productivity problem.
Imagine someone saves an hour by generating a long document and giving it a cursory review.
Five colleagues then spend half an hour trying to interpret it, questioning things that don’t make sense, correcting assumptions and working out which parts actually represent the author’s position.
AI hasn’t saved an hour.
It has transferred the work to the reader.
Alpert makes this point particularly well: authoring effort happens once, while comprehension effort is multiplied across every person who has to consume the work.
Efficiency isn’t measured at the point where the document is generated. It needs to be measured across the work that follows.
What does good AI adoption look like?
Good AI adoption improves the quality of work, not just the speed of content creation.
Organisations gain the greatest value from AI when they use it to:
- Improve decisions
- Challenge assumptions
- Reduce repetitive tasks
- Surface risks earlier
- Accelerate learning
- Enhance professional expertise
Poor AI adoption focuses on output volume.
Successful AI adoption focuses on better outcomes.
Better work is the measure
Ultimately, the measure of good AI adoption isn’t how many prompts we send or how much content we generate.
It’s whether the work got better.
Did we consider something we otherwise wouldn’t have?
Did we find useful evidence faster?
Did we test an assumption?
Did we improve a recommendation?
Did we identify a risk earlier?
Did we remove repetitive effort while retaining professional judgement?
Did we create more time for the work that actually requires expertise?
That’s the opportunity we’re excited about.
Not replacing professional thinking.
Giving professionals better tools with which to think.
Use AI boldly. Delegate thoughtfully. Validate properly.
And never outsource ownership.
About the Author – Rachel Harnott
Rachel Harnott is the COO at WebVine. She works with organisations across Australia to improve the way they collaborate, manage information and get value from Microsoft 365, SharePoint and AI technologies such as Microsoft Copilot. Rachel is particularly interested in the practical realities of AI adoption, helping organisations use emerging technologies in ways that improve outcomes, support better decision-making and maintain accountability.
Frequently Asked Questions
Does AI replace professional expertise?
No. AI can assist with research, drafting, analysis and idea generation, but professional judgement, accountability and decision-making remain human responsibilities.
Who is responsible for AI-generated work?
The person using the AI remains responsible for reviewing, validating and approving any output before it is shared or acted upon.
What is AI slop?
AI slop refers to content that appears professional but lacks original thinking, meaningful insight, supporting evidence or clear ownership.
Can AI improve productivity without reducing quality?
Yes. The best outcomes occur when AI removes repetitive effort while allowing professionals to focus on expertise, judgement and decision-making.
What is the biggest mistake organisations make with AI?
Many organisations focus on producing more content rather than improving the quality of thinking behind the work. This often results in more output but less value.