AI Is Changing How We Decide. We Still Don’t Know If It Is Making Us Better at It.

I recently ran three separate VoxQI analyses around ChatGPT, Claude, and Grok conversations and sentiment.

Across the searches, VoxQI examined more than 10,000 observations (and growing) drawn from Reddit, TikTok, X, YouTube, and Google advertising data. 

I conducted the research to understand how people are discussing these tools in actual use: what they trust them to do, where they remain skeptical, and how AI is entering everyday workflows and decisions.

The three platforms occupy different positions in the public conversation.

ChatGPT has become the default reference point for artificial intelligence. People use it as a first-pass researcher, writer, advisor, and problem-solving tool. Even conversations about competing systems frequently return to ChatGPT as the benchmark.

Claude appears more often in discussions about serious production workflows: software development, automation, research, structured writing and systems that require sustained context. 

The scrutiny is correspondingly practical. Users are not merely asking whether Claude is intelligent. They are asking whether it follows rules, corrects faulty premises and produces reliable work over repeated use.

Grok is discussed as both a tool and a cultural object. Its visibility, personality, connection to X and association with platform politics shape perceptions alongside its technical performance. It attracts experimentation, comparison and attention, but its credibility is also more entangled with questions of governance, moderation and brand trust.

Those differences matter. But the stronger finding was what the conversations shared.

People are already consulting AI before making decisions about:

The model does not always make the final decision. It often influences how the person frames the problem, which alternatives they consider, and how confident they feel about a possible course of action.

That distinction is important.

AI may improve the substance of a decision. It may also make an existing preference sound more coherent. Those are not the same outcome.

A fluent answer can surface missing considerations, organize evidence, and challenge assumptions. It can also create a convincing explanation around a weak premise.

The decision-maker may leave better informed.

The decision-maker may simply leave more certain.

At the moment, we do not have a sufficiently clear understanding of when each is happening.

Across the three analyses, users repeatedly described checking AI outputs through other sources, other models or other people.

They were looking for:

The behavior is neither blind adoption nor wholesale rejection.

It is closer to:

This may be one of the most important changes in how AI is entering professional and consumer life.

The early conversation focused heavily on what AI could generate. The current conversation is increasingly concerned with what can be trusted, repeated and acted upon.

The market is moving from generation toward qualification.

The relevant question is no longer only, “Can AI produce an answer?”

It is, “What would make this answer sufficient for a consequential decision?”

The research also suggests that people are becoming less loyal to a single model.

They move between ChatGPT, Claude, Grok, Gemini, DeepSeek, Perplexity and specialist systems depending on the task. 

Some use one model to produce an answer and another to criticize it. Others add search engines, social platforms, professional communities and verification layers.

In practice, people are assembling decision stacks:

Experience + AI + external evidence + peer judgment + current market signals

The composition changes according to the decision.

A low-risk task may require only a fast AI response.

A higher-stakes decision may trigger a much more complex process: several models, outside sources, colleagues, public discourse, and specialist evidence.

This leads to a question I believe deserves much more serious investigation:

The answer is unlikely to be the same across every person, model or decision.

The three models do not occupy the same place in the public conversation. Each is being evaluated through a different mix of utility, trust, culture and workflow fit. But together they show how fragmented AI-assisted decision-making has already become.

The differences are meaningful.ChatGPT is often where the process begins. Claude appears more frequently where users are building or operating systems. Grok is more entangled with culture, personality and platform identity.But none of the three is being treated consistently as a final authority. 

Across the VoxQI analysis, users moved between models, external sources, peer communities and verification tools. The emerging behavior is not loyalty to one system. It is the construction of a decision stack:

That shared pattern matters more than the competition between individual platforms. It suggests that AI is not simply replacing one source of advice. It is being inserted into a larger process for deciding when the available information feels sufficient to act.

Confidence has practical value.

Leaders cannot operate indefinitely in a state of uncertainty. Decisions must eventually be made with incomplete information. Experience, judgment and intuition have always helped close the distance between what is known and what must be done.

AI now enters that distance.

It can help decision-makers identify possibilities, structure trade-offs and retrieve relevant context more quickly. It may allow a founder, executive or operator to examine a question from several angles before committing resources.

But increased confidence can also become dangerous when it is produced by coherence rather than evidence.

A polished recommendation may conceal:

The core risk is not simply that AI can be wrong.

Humans have always been wrong.

The more interesting risk is that AI can make an incomplete interpretation feel finished.

This research is becoming the starting point for a new Digitalist Hub project.

I will be speaking with founders, executives, investors, agency leaders and operators about how AI is entering consequential business decisions.

The project will examine:

The objective is not to stage a simplistic competition between people and machines.

Human judgment and AI do different things.

The more useful question is how they interact—and whether the combination leads to better decisions.

The first phase will consist of structured conversations with experienced decision-makers.

Participants will not be asked to reveal confidential plans, client identities, private financial information or anything restricted by an NDA. They will be able to participate by name, by professional role or anonymously.

Where appropriate, named contributors will receive visibility in the resulting report, including an approved biography, quotation and professional or company link.

The interviews will help identify recurring decisions where evidence remains incomplete and the consequences of error are material.

A later phase may test one anonymized business decision through three stages:

The point will not be to declare a universal winner.

It will be to observe what changes:

That is where the project becomes especially relevant to VoxQI.

Interviews can reveal what leaders believe is unresolved. VoxQI can examine how the same issue is appearing across public discourse, consumer conversations, cultural signals and live market behavior.

The value is not in collecting more opinions.

It is in finding something consequential hiding inside them.

AI is already becoming part of how decisions are researched, debated and justified.

That process is moving faster than our understanding of its consequences.

Organizations are introducing AI into strategy, operations, marketing, product development, investment and customer intelligence. Yet the important question is not simply whether the tools are being adopted.

It is whether they are improving the quality of the decisions made with them.

The public conversation around ChatGPT, Claude and Grok suggests that users themselves are still working this out. They appreciate the speed and utility. They compare models, build workflows and consult AI in increasingly practical situations. At the same time, they verify, question, challenge and look for external confirmation.

That tension is not a weakness in the adoption story.

It is the story.

We are entering a period in which advice is abundant, reasoning is inexpensive and confident recommendations can be generated instantly.

Evidence, judgment and accountability remain scarce.

The next competitive advantage may not belong to the person or organization with the most AI.

It may belong to the one that knows what to trust, what to question and what still needs to be understood before acting.

AI is already influencing consequential business decisions. The question is whether it is improving judgment—or simply increasing confidence.

I am speaking with founders, executives, investors, and operators about where AI is changing how they evaluate, prepare for, and make important decisions.

No confidential information is required. Participants may be named, attributed by role, or remain anonymous.

The central question is simple:

Participants may be identified by name, by role, or remain anonymous. No confidential company or client information is required.

Have a decision process worth examining? Request a research conversation.

Contact Through LinkedIn