Conviction: When Belief Becomes a Substitute for Understanding

While watching some YouTube videos recently, a story about DeepSeek caught my attention. Not simply because of another technical breakthrough, although the engineering itself is impressive. Its DSpark work focuses on making AI inference dramatically more efficient by attacking the problem differently, using better drafting, verification, confidence management, and hardware awareness rather than simply throwing more computing power at it. The reported results are significant, but what interests me most is not the number. It is the mindset and behaviour behind them.

There is something worth studying in that. Constraint did not eliminate ambition; it sharpened the focus on the challenge. A real problem was identified, an idea emerged, and that idea was attacked, tested, refined, and pursued. The path forward was not, apparently, “spend whatever it takes until the problem surrenders.” The thinking itself had to get better. They had to innovate.

That is where conviction begins. Not with belief, enthusiasm, a mandate, or avarice, but with a worthy idea that has been beaten up, challenged by evidence, exposed to practical reality, and still earns the right to be pursued.

There is another kind of behaviour I have seen throughout my career, and it is almost the inverse.

Intent Without Vision Becomes Willful Aspiration.

“I want the outcome. I have already decided the means. Don’t confuse me with facts.”

I have seen variations of this pattern with the fax machine, email, the internet, enterprise software, cloud computing, digital transformation, and now artificial intelligence. All are worthwhile and valuable, but not a panacea. Technologies evolve and mature, but the pathology does not.

Someone sees the promise of a technology and decides that acquiring it will make things better, as though purchase itself were somehow equivalent to understanding what the technology needs to do, or what the organization needs to do with it. A budget is approved, a program begins, a deadline is declared, the software arrives, consultants arrive, presentations multiply, and activity becomes evidence of progress.

That is when expenditure is industrialized.

Organizations can become extraordinarily efficient at spending money around an idea they have never properly understood. They mistake acquisition for transformation, deployment for adoption, and automation for improvement.

I have seen this firsthand, in different forms, many times throughout my career. One example involved an industrial company whose leadership became convinced that the business needed an ERP system. That conclusion, in itself, was not unreasonable. The company was growing, its systems were long-standing and established, though fragmented and siloed, and there were legitimate operational problems to address. The problem was not the aspiration. The problem was everything that followed.

Several systems were evaluated, but the process focused primarily on selecting technology rather than deeply understanding the organization that would have to live with it. Implementation responsibility was assigned to someone with great enthusiasm, but without the experience needed to grasp the scale of organizational change being attempted.

Early in the implementation, the ERP consultants said something entirely reasonable:

“We need to understand how you work. What do you need this system to improve?”

My advice, as a technology service provider at the time, was simple. You have to be able to describe your processes and systems with the same familiarity, clarity, detail, and comfort with which you would describe how you tie your shoelaces. That sounds almost absurdly basic, but it is not. An ERP system is not a magic box into which a company pours its existing behaviour and receives transformation in return. The work requires an organization to understand and describe itself. Don’t get me started on PLM.

How does work actually flow? Where are decisions made? What happens when the normal process fails? Which exceptions are legitimate? Which workarounds exist because of genuine need, and which exist because nobody has challenged them in fifteen years, or ever, because doing so might invalidate someone else’s bright idea? Those questions are uncomfortable, and technology cannot make them disappear. In fact, if you are open to the process, technology will expose them.

If you do not inspect what you expect, you are doomed.

The organization also established an aggressive implementation timeline before fully understanding the work that transformation would require. The logic was familiar: set the deadline, drive toward it, and completion will follow. Completion, perhaps; success, not so much.

The implementation continued, but over time an enormous amount of effort was spent trying to make the software conform to the organization while the organization resisted changing itself. The underlying message was always some variation of the same thing: we have done it this way for years, we have never done it differently, and we are confident it will not change. They failed to recognize their dependency on deeply embedded institutional knowledge, and the risk they faced when that talent left or retired.

Meanwhile, the rest of the world continued to change around them, adopting new technology, reshaping how work was organized, and increasing delivery while becoming more efficient.

They wanted to bend the software instead of bending themselves. That distinction is the essence of transformation.

The system did not struggle because the company lacked intelligent people, because nobody worked hard, or because nobody cared. Many of the people involved cared enormously. It struggled because the organization wanted a transformed outcome while defending and clinging to the behaviours that made transformation necessary in the first place.

One experience from that period captured the larger problem in a single room.

A persistent production problem had been circulating through the organization. It appeared in one part of the business, but its roots crossed several others. Different groups were living with different parts of the pain, each seeing the problem from inside their own silo, but nobody had a complete view, and the organization had never challenged the problem as a whole.

After watching the issue go around in circles, I proposed something radical only because the organization had made it radical: put everyone who actually knew something about the problem in the same room at the same time.

Operations, engineering, production, design, administration, management, and several of the people who understood the work because they lived it every day. The point was not to hold another meeting. The point was to make the whole system visible to the people who actually knew the work.

The discussion began with someone close to the operational pain describing what was happening. Others added what they knew. One by one, different pieces of the problem appeared, yet nobody could quite see how they connected.

Then, in a moment of frustration, one of the senior leaders stood up and explained what was wrong.

Once the whole problem was visible, his experience connected the dots almost immediately.

Within minutes, a costly problem people had been circling without resolution suddenly made sense. People in the room were stunned, not because the answer was impossibly complex, but because the right knowledge had never been assembled in one place at one time. The buzz and energy were palpable.

Then came the comment that has stayed with me for decades. “Hosting the meeting cost too much money in labour.”

Think about that for a moment.

The organization had just demonstrated, in real time, the value of cross-functional understanding. A persistent problem that had survived inside fragmented perspectives became visible and solvable once the right people were brought together. Yet the cost that mattered most was the immediate cost of having those people in the room, not the long-term cost, time, and effort that understanding and resolving the problem could save.

The cost leadership could see was the cost of the meeting. The cost it could not see was the accumulated waste of never having the meeting.

That moment captured something I have seen repeatedly since. Organizations can become obsessed with visible cost while remaining astonishingly tolerant of invisible waste.

The spreadsheet shows the meeting. It does not show years of rework, failed collaboration, or siloed thinking. It shows a few hours of labour, but not the compounding cost of allowing the same misunderstanding to move from project to project and year to year.

That kind of thinking exhausts people who see the larger system, because they are constantly asked to justify the cost of addressing a problem while the cost of failing to see the whole problem remains comfortably unexamined.

Over time, enormous effort continued to be expended. The organization moved closer to a working system, while the outcome itself remained impossibly out of reach. That closeness only made the situation more painful. Because close can be deceptive.

You can see the other side, and the destination is right there. Perhaps, measured in a straight line, it does not even seem very far away. But close might as well be standing at the edge of the Grand Canyon expecting to cross it with a tricycle in a single leap.

You can improve the seat, lubricate the wheels, measure tire pressure, create a steering committee to improve pedalling cadence, hire consultants to benchmark tricycle performance, and spend millions building a dashboard that reports revolutions per minute. But unless you are addressing the systemic gap itself, an inch might as well be a mile, and you will still not be any closer to the other side of the canyon.

That is the cruelty of proximity without clarity.

I have since seen variations of the same pattern at far greater scale, across different organizations and technologies, with larger budgets, greater complexity, and some of the smartest specialists I have ever met. Yet the pathology was familiar.

Give us new, but do not make us use it differently. Transform us, but do not require us to change. Modernize the environment, but preserve the assumptions, behaviours, governance structures, and ways of working that created the need for modernization.

Even leaders genuinely committed to change can find themselves spending enormous amounts of energy just creating the political and organizational conditions necessary for meaningful work to begin. Technology may be available, money may be available, intelligent people may be everywhere, and yet the organization still searches for a happy path that delivers the outcome without requiring the difficult work of real change.

The organization sees the other side of the canyon, but keeps looking for a way across that avoids confronting the terrain.

This is why I find the DeepSeek story so interesting. It represents something fundamentally different.

They identified a problem worth solving while knowing they would be working under severe constraint. They had a seed of an idea. They examined it, challenged it, and allowed the work to become a process of innovation and discovery. Conviction grew because the idea continued to survive contact with reality.

That is different from deciding what you want and demanding that technology make it true.

One begins with inquiry and asks, “What is the actual problem, and what might solve it?” The other begins with an answer and says, “I have already chosen the solution. Now make the problem fit.” That distinction matters enormously as we enter the AI era.

In an earlier article, I wrote about the industrialization of abdicated intent. I said that AI can scale confusion and encourage organizations to abdicate thinking when they automate before doing the difficult human work of establishing clarity, purpose, judgment, and standards. AI does not invent the ambiguity. It exposes it, accelerates it, and multiplies the consequences.

I think there is another part to that argument. The danger is not only abdicated intent. It is belief masquerading as conviction.

Conviction is earned. Belief is easily declared.

A leader can believe that ERP will transform a business, that cloud will create agility, that a digital platform will create collaboration, or that AI will eliminate cost, remove labour, accelerate decisions, or solve a productivity problem.

Conviction demands more. It should require evidence, understanding, and the humility to discover that the original idea was wrong, incomplete, or pointed at the wrong problem. Those involved should be prepared to “inspect what they expect,” review the evidence honestly, and be willing to change direction when the evidence requires it.

That includes the intellectual honesty to walk away from a chosen vehicle when you discover a canyon, a mud puddle, or that there is simply no there, there.

Without discipline, intent becomes willful aspiration. And willful aspiration is often extraordinarily expensive.

The financial wasteland is easy to measure, at least in theory. Program costs, software licensing, consulting fees, rework, duplication, replacement systems, and labour consumed by years of effort can be counted. The emotional wasteland, though, is far worse.

Good people who care become exhausted. They learn that raising concerns is dangerous or pointless. Trust disappears, cynicism grows, and the people who can see the canyon eventually stop explaining why the tricycle will not cross it. I have seen that happen time and again. Some leave. Others stay and learn to pedal theatrically.

Then there is the wasteland of lost opportunity, and that may be the greatest cost of all. While an organization spends years forcing a poor idea to survive, better ideas never see the light of day. Attention is consumed, capital is committed, political identity becomes attached to the chosen direction, and admitting the mistake becomes more difficult with every passing year.

So the organization doubles down. More money, more technology, more effort, more pedalling, more will, all toward a goal it has never fully understood.

This is why I do not believe we should fear AI primarily because it is powerful. Of course, there are real and legitimate reasons to be cautious about AI, and some deserve serious attention, but one of the most immediate risks is much older than artificial intelligence.

It is us.

It is our tendency to confuse desire with vision, our willingness to substitute authority for understanding, our ability to industrialize expenditure around poorly examined ideas, and our habit of asking technology to save us from understanding the real problem, doing the necessary work, and changing ourselves in the process.

There is a profound difference between believing in the promise of technology and having the conviction to pursue a worthy idea through technology. One is consumption. The other is innovation.

DeepSeek offers a useful counterexample because its innovation suggests another path. Start with the problem, find the idea, beat it up, let others challenge it, follow the evidence, allow constraint to sharpen the thinking, and build conviction because the idea continues to earn it.

Then execute with everything you have. That is not technological faith. That is disciplined pursuit.

And before we spend countless millions more improving the tricycle, exhausting the people who can see the problem, and consuming the opportunity to do something better, perhaps we should ask whether we understand the canyon at all.