The HIPOPOTESIS AI Manifesto
We are neither hyped nor afraid.
Artificial intelligence is one of the loudest forces in our industry right now. We give it room, but we also keep human judgement in front of it, on every project without exception.
AI is a tool in our hands.
Never a replacement for the hands.
The position
The conversation has collapsed into two camps shouting past each other. We refuse both, and we have a third answer that takes a little more work and a lot more honesty.
The optimist
Automates everything, ships it unchecked, and calls the speed a feature. Confuses output with insight, and confidence with truth.
The pessimist
Rejects it on principle, romanticises the old way, and falls behind on methods that genuinely make the work better.
The realist | Us
Studies what AI does well, names exactly what it does badly, and puts in writing where it belongs.
We chose the third one a long time ago. On every project, in plain sight.
> generate: HIPOPOTESIS_ai_principles
We are techno-realists, neither optimists nor pessimists. We weigh what AI does well against what it does badly, and we decide on purpose.
AI is scaffolding around our work, never the work itself.
A person runs every interview, and a human expert checks every insight against the evidence before it reaches you.
We use AI to think faster, to challenge our own thinking and to check our work for contradictions, never to think for us.
We don’t trust AI. It’s fluent, fast and confident, and wrong often enough that we verify everything it touches.
A senior human reads, validates and signs off on every output. The machine never has the last word.
We declare where AI was used on every project, in plain language and in writing.
We use AI task by task, only where it helps the study. Most of the work stays manual, and strategic vision belongs to human experts.
We manage data within rigorous boundaries. Good input needs a person to curate and filter it.
Full transparency is non-negotiable, and our clients can audit our methods.
We protect intellectual property and never run your proprietary data through unsecured public systems.
Human judgement is the final seal. Technology can draft a path, but experts decide what is true.
Where the line is drawn
The judgement you’re paying for stays with people.
Task by task, only where it helps the study, under full senior supervision.
Read that left column again. A big part of our industry is comfortable with that. We aren’t.
The uncomfortable truth
AI is fluent, fast and confident. It’s also a machine built to produce plausible text, not true text. Those are not the same thing, and the gap between them is where research can go to die. It hallucinates, invents quotes or fabricates statistics that look feasible. It cites studies that don’t exist. And it does all this in the same calm, authoritative voice it uses when it happens to be right. This is what makes it dangerous in research: the errors don’t look like errors. That’s why at HIPOPOTESIS every output is reviewed, line by line, by a senior expert. The machine never has the last word. Humans do.
Cautionary tales
These aren’t internet rumours. They are documented public incidents from the last few years. We leave out the names on purpose; the point is the pattern, not the shaming.
Case 01
A global consulting firm delivered a six-figure report to a government department. It cited academic papers that didn’t exist, footnotes that led nowhere, and a quotation invented and placed in the mouth of a real judge. The client only learned AI had been involved after an outside academic spotted the fabrications.
Outcome: a partial refund.
Case 02
A respected publication ran dozens of AI-written articles under a staff byline. An internal audit later found that more than half needed corrections, some of them substantial, on facts as basic as how compound interest works. Readers were never clearly told a machine had written them.
Outcome: corrections on 53% of the articles.
Case 03
Experienced professionals filed formal documents built on six prior cases that the AI had simply made up, complete with invented quotes and citations to rulings that never happened. They were sanctioned, and it has happened again and again around the world since.
Outcome: a $5,000 fine and a finding of bad faith.
Case 04
A major international airline deployed a customer service chatbot that invented a discount policy for a grieving passenger. At a civil tribunal, the airline argued it couldn’t be held responsible for what its chatbot said, in effect treating it as “a separate legal entity” responsible for its own actions.
Outcome: ordered to pay, in a ruling still cited today.
Case 05
A finance employee at a multinational joined a video call with their CFO and several colleagues to authorise an urgent transfer. Everyone on the call except the employee was an AI deepfake, and nobody noticed until the money was gone.
Outcome: $25,000,000 stolen.
Behind these cases are respected firms and seasoned experts who trusted the output and skipped the review.
We will never put your name on work like that. That’s a promise about how we work, whatever the technology can do.
Our position
A young technology is being overused to pass off cost-cutting as “statistical scale”. We don’t build high-stakes strategies on automated echoes.
Reported n = 40Real n = 12
Many firms now interview a handful of real people and fill the gaps with AI-invented profiles to simulate weight. We refuse to fake volume.
Critical business decisions need real cultural nuance, friction and human paradox, and synthetic simulations lose all three.
Simulations have a place in isolated micro-testing. For core strategic intelligence, our rule doesn’t bend: real humans first, always.
Our standing rule
We don’t hide it or pretend it does the hard part. On every project we declare, in plain language and in writing, where AI helped and where it didn’t. Open each assisted step to see exactly which task a tool touched, and which one stayed human. Most firms charge you for senior human hours while they hand the work to a machine, undeclared, at the same price. We think that’s dishonest. You’re paying for expert people, real time and real accountability, and when AI saves us an hour, we spend it checking the work again.
AI disclosure | Sample
Methodology in action
We go beyond generic consumer panels. Our team handpicks industry leaders, specialist operators and senior stakeholders who know the real, unvarnished friction in the market.
Every qualitative signal is cross-checked against quantitative data and real behaviour, so no finding rests on an assumption or an automated shortcut.
We translate complex human paradoxes into boardroom-ready intelligence: the clear, defensible direction a decision worth millions needs, which a descriptive report rarely gives you.
Opinions are cheap, so we put ours on the record, in public, in front of tens of thousands of professionals.Our LinkedIn Learning courses
Straight answers
Something else on your mind? Ask us directly. You’ll get an answer from the people who would run the work.
Yes, and we’re open about it. We choose it task by task, only where it helps the study: to transcribe recordings that a person then checks, to sketch a first map of a large body of material, to read our analysis again looking for contradictions, and to flag issues in our own writing. Most of the work stays manual. We never use it to talk to participants, to invent data or to decide what a finding means, and every use is declared in your project.
Only as an assistant, and never on its own. On a large study it can propose a first view, such as recurring themes or the kinship ties across a set of interviews, and it can check our analysis for contradictions and weak claims. A senior researcher checks every proposal against the transcript or the data, changes what’s wrong and drops what doesn’t hold. The coding frame, the interpretation and the conclusions are ours, and so is the responsibility.
We see synthetic users as an illusion of robustness. In the research industry, this young technology is often used to pass off cost-cutting as statistical scale: a firm interviews a handful of real people and pads the sample with AI-invented profiles to simulate weight. We refuse to fake volume. High-stakes strategic decisions need real cultural nuance, behavioural friction and human paradox, and automated echoes lose all of them. Simulations have a limited place in isolated micro-tests, but they never enter our core research. Every interview we run is moderated by a person, and at HIPOPOTESIS every high-value project starts, runs and finishes with real human experts and real participants.
Always. Every project includes a plain-language AI disclosure that states exactly where a tool helped and where it did not, down to the individual task. You’ll never have to wonder whether you paid for human expertise or for an unsupervised machine.
Because both are true, and holding both is the whole point. AI is a useful assistant and an unreliable authority. It produces plausible text, not necessarily true text, and it can hallucinate facts, quotes and sources with total confidence. So we use it for support and verify everything with senior human review. Trust sits with the people, not the tool.
It makes some steps faster, and we put that time back into the work: another reading of the material, another check of the numbers. Used as a second reviewer, it also catches contradictions and weak claims a person can miss on the fifth read, so fewer errors reach you. What you pay for is senior people who answer for the result, and that hasn’t changed. Cheap automated research feels like a saving until it sends a launch or a market entry in the wrong direction.
Why blue to green
From ignorance to wisdom
01 | The sky
Hypotheses, hunches and assumptions that nobody has tested yet.
What we do here
“Our customers would never pay for that.”
02 | The descent
Data you already own: scattered, unread, or contradicting itself.
What we do here
“We have three studies on this, and they disagree.”
03 | The ground
Evidence you can act on and defend in front of the board.
What we do here
“We know why it happens, and we can prove it.”
Our research is the gravity that pulls an idea out of the sky and plants it in the ground.
Walk the Research Cycle