Strategy #14: When AI accelerates decision-making, who still holds the responsibility?
In a previous article, we discussed how AI creates the most value when organizations have “discipline in decision-making.”
Because having AI doesn't automatically mean an organization will make better decisions.
AI may increasingly assist in writing, summarizing, translating, analyzing, and offering alternatives, but the crucial question remains: how do humans and organizations use this knowledge to make decisions?
But in this article, I'd like to invite you to take it to the next level.
If AI not only helps us work faster but starts to enter the "pre-decision process," who still owns the decision-making power?
This question isn't just arising in company boardrooms; it's happening in far higher-risk areas such as security, war, and the use of military force.
The case of the US-Iran war is a very interesting example.
According to publicly available information, AI is reportedly being used to support military operations in multiple dimensions, from gathering data from various sources and analyzing battlefield data to target identification and helping prioritize objectives. Systems like Palantir's Maven are described as command-and-control platforms that use AI to analyze battlefield data and assist in target identification. Reuters reports that this system was used in numerous operations related to the attacks against Iran.
But the key point is that we shouldn't jump to conclusions beyond the evidence that AI is making decisions instead of humans.
The U.S. Department of Defense policy continues to state that automatic and semi-automatic weapon systems must be designed to allow commanders and operators to exercise appropriate human judgment in the use of force.
Therefore, the interesting question may not just be, "Does AI make decisions instead of humans?"
But it is
As AI gathers data, analyzes options, prioritizes, and accelerates the decision-making process, do humans still have enough time, space, and real power to question the system's recommendations?
This is an issue that businesses should pay close attention to.
Because even though businesses aren't on the battlefield, they're facing similar problems, only the impact differs at different levels.
In the business world, AI may not target military objectives, but it could help prioritize customers, assess debtor risk, screen job applicants, analyze suppliers, propose investment plans, or advise on how to make decisions in specific situations.
Then a human approves it.
The question is, did the human actually “make the decision” or simply “approve” what the system made easy to approve?
This is a fine line that organizations in the AI era must be wary of.
Many organizations like to use the term "human-in-the-loop".
This means that there are still humans involved in the process; there are people reviewing and approving.
But in the age of AI, the term "human-in-the-loop" may not be enough.
Because the presence of humans in the process doesn't necessarily mean that their judgment is always meaningful.
If AI were to choose the data to see.
AI prioritizes that data.
AI is offering an alternative.
AI is a risk assessor.
AI makes some options seem more suitable than others.
Humans may still be living... It's a loop, that's true, but it doesn't mean humans are still under command.
What organizations need in this era is not just a human-in-the-loop system, but... Human-in-Command
Humans must not be merely those who click the approve button at the end of the process, but must be the owners of the questions, the owners of the decision criteria, the owners of the risks, and the owners of the responsibility.
This is what Decision Governance means.
Decision Governance is not about adding more approval steps.
But it's about designing clearly how key organizational decisions are made, who owns the data, what data is used, where AI will play a role, what humans will monitor, and when the results don't meet expectations, who needs to learn and adjust the system.
In the pre-AI era, organizations might design governance to control “decision-making power.”
But in the AI era, organizations must design governance to control "influence before decision-making" as well.
Because AI may not make decisions directly for us. but AI may determine what information we see and what information we don't see. Which option do you see first, which one comes later, and which one seems the most credible?
Strategically, this is very important.
Because whoever controls the framing of the choices generally has a significant influence on the outcome of the decision.
For example, if AI helps analyze sales figures and suggests cost reductions in certain areas or closing branches, management might be able to make decisions based on those suggestions more quickly.
However, if AI lacks data on long-term customer relationships, doesn't understand market sensitivities, or isn't aware that cost-cutting certain measures or branch closures could damage brand trust, it could be problematic. A decision that appears to be informationally correct may not be the best strategic decision.
Therefore, what needs to be checked is not just the AI's answer.
However, we need to examine the questions we ask the AI, the data the AI uses, the assumptions the AI makes, and the alternatives the AI doesn't suggest.
Organizations that utilize AI effectively must adhere to at least four principles.
- It must be clearly stated whether the AI is merely an analytical assistant or has the right to perform certain actions independently.
- It is necessary to define which decisions are low-impact and which are high-impact.
- It's important to designate who the accountable owner is for that decision. This is something we rarely do.
- A feedback loop is needed to monitor how AI-assisted decisions actually yield results. Most of the time, we let things run their course and then make adjustments on the fly.
In particular, decisions that affect people, customers, safety, investment, trust, or the reputation and brand of the organization should have clearer guidelines than general tasks.
What's worrying isn't... AI is too good.
But the problem is that organizations haven't yet designed responsibilities to keep pace with the capabilities of AI.
In the past, when people made wrong decisions, we could still ask: Who proposed it? Who approved it? Who is responsible? And who should learn from it?
However, in the AI era, if decisions are made based on data selected by the system, models analyzed by the system, recommendations proposed by the system, and approval made by humans under time constraints, the question of accountability becomes much more complex.
This is why organizations shouldn't wait for mistakes to happen before designing governance.
Because good governance must be designed before a crisis, not after a crisis.
The lessons learned from using AI on the battlefield are therefore not limited to military matters.
But it reflects a big question for every organization:
While AI makes decision-making much faster, humans still have a chance. Think, question, object, examine, and take responsibility. Is it meaningful?
In the business world, the answer to this question may determine the difference between organizations that use AI to enhance decision-making and those that only use AI to accelerate existing decisions.
Speed is important.
However, speed without governance can lead an organization astray faster than it can, like driving a car at high speed but the steering wheel and brakes aren't designed to handle the increased speed.
In the AI era, the advantage isn't always about who makes the fastest decision.
But it all comes down to who can make decisions faster while still maintaining accountability, transparency, and human judgment.
In a previous article, we said that the AI era doesn't need humans who think less, but rather humans who think more clearly.
This article will therefore continue with one more sentence.
The AI era doesn't just require organizations to approve decisions faster, but also organizations that clearly understand who owns the decision and who is responsible when that decision impacts the organization's future. Ultimately, AI may help us see the answers more quickly. But the responsibility for the answer always remains with humans.
References: Reuters, “Pentagon to adopt Palantir AI as core US military system, memo says”; US Department of Defense Directive 3000.09, “Autonomy in Weapon Systems”.
