Business Intent

An AI system that challenges business assumptions by investigating reality, finding contradictions, and discovering what companies may be getting wrong.

Comment

Inspiration

AI can now generate business plans, summarize markets, analyze documents, and answer strategic questions in seconds.

But there is a deeper problem:

What if the question itself is based on a false assumption?

Businesses constantly make decisions based on beliefs about customers, competitors, markets, opportunities, and risks.

For example:

"Enterprise customers are our best growth opportunity."

A conventional AI assistant might accept that premise and help create an enterprise strategy.

We wanted to build an AI system that does something fundamentally different:

What if AI tried to prove that the assumption is wrong?

That became Business Intent.

Business Intent is an AI-powered reality-testing system designed to investigate business assumptions, gather evidence, identify contradictions, generate alternative explanations, and help users update their understanding before committing to a decision.

What it does

A user provides Business Intent with a strategic question, hypothesis, or assumption.

For example:

"Should our company expand into Europe?"

Instead of immediately generating a recommendation, the system first investigates what would need to be true for that decision to make sense.

It can break the problem into multiple research dimensions, investigate relevant information, and bring the resulting evidence into a structured reasoning process.

The system looks for:

  • Evidence supporting the original assumption
  • Evidence contradicting the assumption
  • Competitor movements
  • Market signals
  • Customer and industry signals
  • Emerging opportunities
  • Risks
  • Alternative explanations

The result is not simply:

"Here is an answer."

It is:

Assumption → Evidence → Contradiction → Alternative hypothesis → Recommendation

The central design principle is that contradictory evidence is valuable.

Instead of filtering conflicting information away, Business Intent uses contradictions as signals that the model of the situation may be incomplete or wrong.

The Core Idea

Business Intent is built around a simple loop:

Observe → Model → Assume → Attack → Discover → Test → Update → Act

The system starts with an understanding of the user's question.

It then attempts to challenge that understanding by searching for evidence that could invalidate the initial hypothesis.

If contradictory evidence appears, the system investigates the contradiction instead of simply ignoring it.

The goal is to continuously improve the model of the situation.

This makes Business Intent less like a traditional chatbot and more like an AI investigation system.

How we built it

We designed Business Intent as an agentic intelligence pipeline rather than a single LLM prompt.

The high-level architecture is:

Strategic Question ↓ Intent / Hypothesis Decomposition ↓ Research Tasks ↓ Evidence Collection ↓ Evidence Analysis ↓ Contradiction Detection ↓ Synthesis ↓ Strategic Recommendation

The system separates information gathering from reasoning so that the model can work with newly collected evidence rather than relying exclusively on pretrained knowledge.

Each investigation begins with a specific question and produces evidence that can be evaluated in relation to the original hypothesis.

The final synthesis connects those independent signals into a structured strategic conclusion.

A Concrete Example

Consider the assumption:

"Europe is our highest-leverage expansion opportunity."

Business Intent does not simply search for information about expanding into Europe.

It asks:

What evidence would support this belief?

and:

What evidence would prove it wrong?

It may discover that European demand is growing, which supports the hypothesis.

But it could simultaneously discover that:

  • customer demand is stronger in another market,
  • competitors are becoming more concentrated in Europe,
  • acquisition costs are significantly higher,
  • regulatory conditions create additional friction,
  • or another segment has a substantially better opportunity.

The system then surfaces the conflict.

Instead of hiding the contradiction, it makes it the center of the investigation.

The output becomes a reality check rather than a confirmation of the user's original belief.

Why this is technically interesting

The difficult part of this problem is not generating text.

It is creating a system that can reason about beliefs, evidence, uncertainty, and contradictions.

Business Intent therefore treats the strategic question as something that can be investigated rather than something that should automatically be accepted.

This creates several important distinctions:

Question ≠ Assumption

Assumption ≠ Evidence

Evidence ≠ Truth

Contradiction ≠ Noise

These distinctions form the foundation of the system.

What we learned

Building Business Intent taught us that an intelligent AI application is not necessarily defined by the model underneath it.

The surrounding system matters just as much.

We had to think about:

  1. How to decompose ambiguous questions into useful investigations.
  2. How to determine which evidence is relevant.
  3. How to distinguish evidence from assumptions.
  4. How to search for counter-evidence rather than confirmation.
  5. How to reason about conflicting signals.
  6. How to turn uncertainty into an actionable next step.

The biggest conceptual shift was realizing that an AI system can be designed not only to answer questions, but to challenge the assumptions that produced them.

Challenges

Real-world intelligence is messy.

Information is fragmented, incomplete, constantly changing, and sometimes contradictory.

A system that simply searches for information can easily fall into confirmation bias: it finds evidence supporting the user's original belief and produces a confident answer.

We therefore designed Business Intent around the opposite behavior.

The system actively searches for information that could weaken or invalidate the original hypothesis.

This makes contradictions useful rather than undesirable.

Another challenge was maintaining the original strategic objective while allowing the investigation to explore multiple independent directions.

What we built

Business Intent is a working prototype of an AI system for strategic reality testing.

The current system demonstrates an end-to-end flow from:

Strategic Question → Investigation → Evidence → Contradiction → Intelligence → Action

Rather than building another AI assistant that simply makes existing workflows faster, we wanted to explore a more fundamental question:

Can AI help us discover when our understanding of reality is wrong?

Business Intent is our prototype toward that idea.

Future Vision

The long-term vision is to turn Business Intent into a continuously updating business reality model.

Instead of investigating a question only when a user asks, the system could continuously monitor changes across markets, competitors, customers, economics, operations, and strategy.

When the world's behavior begins to contradict the company's assumptions, the system could surface the change automatically.

The ultimate goal is not to build an AI that makes decisions for humans.

It is to build an AI that makes the human decision-maker's model of reality harder to fool.

Don't just ask AI what to do. Ask AI what you might be getting wrong.

Built With

Share this project:

Updates

Submission history