The Quiet Costs of AI That Are Already Affecting Its Net Value

There has been a lot of discussion what AI can do. It can surface information faster than any research team. It can run analyses in minutes that used to take days. It can distil a hundred-page report into a crisp summary.

All of this is true. And all of it is good.

But there is a quieter, more complex conversation that does not get enough airtime -  one about what happens in the space between what AI can do and what we, as individuals and organisations, are actually ready for. The benefits are real. So are the transitional risks that come with them.

These are not just organisational challenges. They are personal ones too: for each of us navigating a new way of working, learning, and relating to information and to each other.

1. More Information, But Not More Judgement

AI can pull research from vast sources and deliver it faster than any team. But a junior analyst who has learned to triage sources, spot a suspicious statistic, or sense when something does not quite add up - that skill was built through trial and error, and experience, not speed.

When individuals hand research over to AI, the judgement muscle can quietly atrophy. At the same time, we are operating in an era of significant misinformation, which makes quality control more important, not less. The risk is that AI produces volume, while our individual and collective capacity to interrogate that vast volume has not kept pace.

Investing in the human side of information governance is not optional, it is urgent.

2. Speed Is Not the Same as Understanding and Insight

AI can run analysis quickly. But analysis is not just an output; it is a process through which people build intuition, notice anomalies, and develop the felt sense of what seems off. The professional who has spent hours inside a dataset, who remembers the painful quarter when a particular metric behaved unexpectedly, brings something that AI currently cannot replicate.

Think about what we lose when AI does the analysis and we simply receive the conclusion. Individually, we miss the chance to build pattern recognition. As teams, we gradually lose the shared fluency that allows us to have meaningful conversations about trends and results. Speed at the front end can create fragility in our people and in our thought process at the back end.

3. Summaries Are Efficient. Learning Is Not.

AI can summarise a report impressively well. But when someone reads a summary rather than the source, they have had a fundamentally different, and shallower, encounter with the material.

Reading, absorbing, and wrestling with information is how people learn. It is how they develop perspective and form the kind of informed scepticism that makes them effective professionals. When AI takeovers this process, the individual's relationship with knowledge becomes thinner over time.

At first, this feels like an efficiency gain. Over time, it becomes a dependency. The shift happens gradually and quietly. Fluency and dependency can look very similar; the difference is whether the person using it could still interrogate, challenge, or own the output if the tool weren't there.

The knowledge gap between what the machine holds and what any of us actually understands will widen quietly, without obvious warning signs, until it matters enormously.

4. We Are Used to Trusting Computers. AI Is Different.

One of my clients said something during an AI training session that has stayed with me: "We are used to trusting computers. But with generative AI output, we need to be more critical and question what we get, more than we are used to."

This is a significant mindset shift. When a spreadsheet returns a number, it is the number. When an AI generates a response, it is a plausible-sounding version of an answer - one that can be confidently wrong, subtly incomplete, or shaped by the prompt in ways we did not intend. The interface looks trustworthy. The output may not be.

For individuals and teams, developing genuine alertness to AI output - a kind of critical reflex that does not come naturally when we are working fast, is one of the most important skills of this moment.

It is not about being sceptical of AI for its own sake. It is about staying in the driver's seat.

5. AI Does Not Really Push Back. And That Is a Problem.

Here is something we hardly acknowledge: people are becoming more fluent at prompting AI; and AI, of course, does not disagree unless we ask it to. It does not say no to what we ask it to do. It does not have bandwidth limits. It does not push back, challenge assumptions, or ask why - forcefully or passionately, like human beings sometime do.

Working with human colleagues is sometimes harder precisely because of this friction. A peer who questions our logic, a direct report who flags a concern, a supplier who says no this time; these are uncomfortable moments, but they are also how we refine ideas and plans, test assumptions, and build the collaborative muscles that teams depend on.

There is an uncomfortable parallel here with social media, where people have grown accustomed to expressing views directly like entitlement, without the moderation that comes from face-to-face accountability.

AI carries a similar risk: it becomes attractive to work out ideas, frustrations, or requests through a machine that never challenges us, rather than through the messier but more valuable process of working it out with real people.

If collaboration is not made intentional, if we do not deliberately preserve the habits of constructive friction, AI adoption risks quietly eroding the team dynamics that organisations need most.

6. Change Management Is the Hardest Part — and the Most Human

Different people will come to AI with very different levels of appetite, aptitude, and trust. Some will move quickly. Others will be more cautious, not because they lack capability, but because caution is often wisdom. Both responses are legitimate.

A few things are worth being explicit about:

• It is okay to try and fail. There is a real risk of wasting time before the efficiency gains materialise. That is not a sign that AI is not working; it is a sign that integration takes time, and people should not be penalised for honest attempts that do not land immediately.

• Experience needs to be shared, with the team and with the AI itself. The quality of AI output is directly shaped by the quality of guidance it receives. When individuals hoard their learnings or develop their own private ways of working, the organisation loses consistency and the AI loses coherence. Making knowledge-sharing a visible norm is part of the integration work.

• Be realistic about the time allowed for transition. Apply an inclusive lens when deciding how long to give people to adopt AI into their workflow. There might be more than one way to get to Rome.

Do this simple exercise: ask your team to describe how they see AI: as a person. An assistant? A researcher? A super calculator? A junior analyst? The answers will vary, and that variation is the point. It surfaces assumptions, tells us how different people understand AI's role, and opens a conversation that is much more useful than any policy document. These mental models will shape how people use AI in ways that are hard to observe otherwise, so surfacing them early is a practical tool, not a soft one.

What This Asks of Each of Us

Integrating AI well is not just a technology or process challenge. It is a human one - for organisations and for individuals.

It requires alignment on purpose, honest assessment of where skills and habits genuinely stand, and the patience to acclimatise rather than simply implement. It asks us to stay alert, stay curious, and stay genuinely engaged with the people we work with - as new tools make it far too easy to sidestep that engagement.

The organisations and individuals who get the most from AI will not necessarily be the fastest adopters. They will be the ones who remain clear-eyed about what they are aiming for and gaining, what they must not lose, and what it takes to bring people through this transition with both capability and confidence intact.

If any of this resonates with what you're experiencing in your organisation, I'd love to hear about it. Whether you're at the start of your AI journey or navigating the messy middle, the conversations worth having are the human ones.