The Human First Standard™ is a framework for keeping human judgment intact as AI scales inside organizations — what is drifting, how to measure it, and what holds the line.
Book the free consultBook the free consult Stay smarter than the machines you build.Every leader I talk to who is integrating AI quickly can feel that something is being lost. In their own thinking. In their team's judgment. In the quality of what ships. Almost none of them can name it, and none of them can measure it.
I call it Cognitive Drift: the slow, unnoticed degradation of human thinking, judgment, and decision-making when an organization integrates AI without defined human checkpoints.
Drift is invisible from the outside. The decisions still get made. The reports still get written. What goes missing is the friction — the argument, the wait, that doesn't feel right — that used to make the work correct.
This is why it compounds. There is no incident, no outage, no quarter where the number breaks. There is only a solopreneur, a team, or a company that within a month decides quickly, ships fast, but somehow cannot remember the last time the first answer got challenged. By then the instinct to argue is usually gone, and so too are the people who had it. That is Cognitive Drift.
Drift does not announce itself. It shows up as small changes in how work feels, and every one of them looks like efficiency first.
At a desk of one it looks different, and it is worse, because there is nobody else to notice. You stop having the first idea. The model’s draft becomes your draft. You stop being able to tell which sentences are yours — and the output goes up, which is why nobody stops.
Book the free consultBook the free consult If any of that sounded familiar, I want to hear about it.Every organization installing AI is redesigning how its people think. Almost none of them are doing it on purpose.
Cognitive systems design is my discipline: keeping human judgment intact as AI scales inside organizations. It extends cognitive systems engineering — Hollnagel and Woods, 1983 — out of the control room and into the company. They studied how people and machines share decisions in cockpits and power plants. I study what happens when the machine is a language model and the control room is an organization.
ML researchers design the machine. AI ethicists comment on the machine. The cognitive systems designer designs the human layer inside it — the decisions, the thinking, the ethics, and the talent the system depends on and routinely ignores.
Human First is where it operates — my body of work within the discipline, and the Standard is how it gets measured. It is also what happens when critical wellness theory meets AI: the body, the nervous system, attention, and load brought to a field that otherwise treats people as units of cognition.
Cognitive systems designer. Critical wellness theorist. Curriculum architect.
Founder of three institutions. Author of the Human First Standard™.
You need someone who has watched judgment break under real pressure — and built, in those same rooms, institutions that held.
Over two decades building learning systems. In production, not in pilot. AI inside them since 2020 — Claude, GPT and Gemini, in domains where being wrong costs someone their mental health, their physical health, or their livelihood.
I am Christina Stoltz — a cognitive systems designer and critical wellness theorist, and the author of the Human First Standard. I work at the intersection of mind, body, and machine, treating the people behind AI as embodied systems under load, which is what they actually are.
For more than two decades I have worked in rooms most people never enter, in peacetime and during civil uprising — women’s prisons, crisis centers, anti-trafficking organizations, sex worker coalitions, domestic violence shelters, and bride-kidnapping recovery agencies, across the United States, Russia, Tajikistan, Kyrgyzstan, and Uzbekistan. I did not observe those rooms. I built working things inside them. None of it was built to get us through the day. It was built to move the mission forward, and to keep working after I left the room.
I am fluent in both high theory and high action, and this work lives in the bridge between them. Rehabilitative prison education. A United Nations research consultancy protecting women entrepreneurs trading on the black market. A Fulbright fellowship advancing gender and development frameworks in deeply conservative Islamic societies. An American University of Central Asia professorship building the institution’s first gender and development research methodology and corresponding digital research archive — writing the curriculum because there was not one to inherit.
My territory is the human operating system behind machines — the condition, capacity, and judgment of the people building them, running them, and living with what they produce.
I do not sit between performance and ethics. I collapse the distinction. A neglected human system produces drifted decisions and drifted systems, which is why caring for the people behind the machines is not soft. It is the performance argument.
Writing a benchmark and making it hold is what I have done more than once. I have designed, operated, and scaled global infrastructures for over two decades — three times over. REQ.1, a nervous system literacy foundation — a 501(c)(3) I founded in Philadelphia in 2010, teaching regulation as a professional skill to people who work under chronic load. PLOOME, a pelvic education institute with a globally accredited curriculum, proprietary equipment, legal entities in three countries, and practitioners across more than thirty. And the International Wellness Credentialing and Accreditation Board, founded because a field with no floor keeps finding new ways to be worse.
I find the problem before it has a name, and then I build the thing that solves it. Pelvic education in 2010 — a non-clinical teaching pathway for the most dismissed layer of women’s medicine, at a point when the subject lived inside physiotherapy and almost nowhere else. Proprietary rehabilitation-adjacent, travel-ready, sustainably designed equipment in 2018 — a lane nobody was thinking about when I first brought it to market. AI-empowered curriculum production in 2020, while the tooling was still unnamed. Every time the same sequence: see it, build it, ship it, and let the market arrive.
I never waited for permission, and I never waited for a market to sanction an idea. I build the object, write the system, and put it into the world.
As Fulbright U.S. Research Scholar in the Kyrgyz Republic (2008–2009), Christina implemented research-grounded educational initiatives within university and NGO environments — connecting academic offerings to applied community outcomes.
Fulbright — “Non-Profit Social Impact Strategist”
As a sexual violence intervention advocate, university educator, and global activist, Christina Stoltz re-wrote the book on sexuality studies in Central Asia. Rarely do we meet a Fulbright Researcher with a stronghold in both academics and activism, but Christina’s grassroots empowerment approach to sexual education extends from the classroom to the crisis center. She is a woman dedicated to making a difference in the world and helping heal the people in it.
As National Consultant and Country Lead for the UN Safe Cities Initiative in Tajikistan (2009–2010) — research protecting women entrepreneurs trading on black market routes — Christina led country-level program implementation under UN governance standards, coordinating government, NGO, and institutional stakeholders against performance-based metrics — translating evidence and field signals into program decisions on the ground.
UNIFEM — “Non-Profit Development Consultant”
Our Central Asian Offices were buzzing about a Russian-speaking American woman who was taking the trauma recovery and non-profit sectors by storm. We hired Christina immediately as our Safe Cities Global Initiative Representative for the country of Tajikistan and she didn’t disappoint. Her extensive fieldwork, compelling research, and strong community connections brought new life and meaning to our worldwide wellness initiative.
Professor & Curriculum Designer at AUCA (2008–2010). Authored and led multidisciplinary curriculum across three departments (anthropology, sociology, journalism) for 150+ adult learners from 15+ countries. Built the university’s first digital research archive — a reusable resource adopted across cohorts and departments.
Adjunct Professor of Sociology at Southeast Correctional Facility (2006–2007). Designed and delivered credit-bearing curriculum for incarcerated adult learners — predominantly first-time college students from low-income backgrounds — driving measurable GED and college-credit attainment in a high-barrier learning environment.
International Policy & Program Specialist at Angel Anti-Trafficking Coalition (Russia & Central Asia, 2006–2008). Coordinated grant-funded program implementation and evaluation with regional stakeholders, transitioning from remote support to on-site work in Moscow.
Trauma-Informed Care Program Manager at Sezim Crisis Center in Kyrgyzstan (2008–2010). Built structured training and service-delivery models for staff working with adult learners and clients in high-stress, high-barrier conditions.
Trauma Recovery Specialist at the WISE Crisis Center and Shelter serving the New Hampshire and Vermont Upper Valley region. Facilitated trauma-informed crisis intervention for women needing advocates in social services, legal resources, and emergency housing — including enrolling women into social service programs, attending restraining order appointments and court cases as an advocate, and running education programming in the shelter.
Co-Keynote Speaker at Tuck School of Business at Dartmouth — Greener Ventures speaker series.
Tuck — “Greener Ventures: Keynote Speaker”
Christina is an innovative entrepreneur whose unique for-profit/non-profit hybrid enterprise utilizes physical fitness to promote social justice and destigmatize mental health taboos. She is working to change the world for the better and using movement as a metaphor for that change. By helping people shift the way they think about their own health and wellness, Christina is educating others on the inextricable link between personal growth, community development, and helping others in need.
Invited Speaker at the Wharton School of the University of Pennsylvania — featured by the Wharton Small Business Development Center for her hybrid education enterprise model.
Wharton Small Business Development Center — “The Small Business That Could”
Christina Stoltz is a solopreneur with serious guts. She trusted her instincts and honored her vision to create something truly unique in health and wellness. She heard many words of caution from seasoned business leaders who believed her vision was too ambitious and her hybrid model wouldn’t work. She went for it anyway. One year after her studio’s grand opening, she was featured in Forbes for her revolutionary alterna-biz approach to integrative wellness.
Christina holds an A.M. in Comparative Literature — Gender & Development (2007) and a B.A. in Russian & Comparative Literature, Cum Laude with High Honors (2006). Her original research was situated across Dartmouth’s Departments of Russian, Geography, and Comparative Literature, and the Dickey Center for International Understanding.
Dartmouth Alumni Magazine — “Recovering Balance Now”
Christina draws on her own experiences to find support and healing for trauma survivors struggling with PTSD, anxiety, and depression through her sister wellness enterprises. As Stoltz explains, ‘Fitness is a powerful metaphor for change. When your body feels capable and your mind is at ease, you remember that you have control over all other aspects of your life, as well.’
The patterns of how human judgment holds or breaks are the same whether the pressure is a human trafficking case, a founder’s fiscal crisis, or an AI integration.
This is the whole reason the Human First Standard™
exists.
More than two decades of watching systems fail under load is the qualification.
Cognitive Drift is the same failure in a new room. It is already inside companies that cannot name it, cost it, or measure it. So I built the instrument that measures it.
The Human First Standard™ measures an organization across four axes. Each tracks a specific dimension of drift and produces its own score; together they produce the Human First Index™.
The four pillars do not change with headcount. The unit does. At four hundred people, Talent measures the bench. At a desk of one, it measures you — whether your own judgment is appreciating or quietly depreciating while the output still looks fine.
Where AI makes or shapes decisions without a defined human checkpoint. Whether there are clear boundaries between what a human must decide and what a model is allowed to.
Catches: the leader who signs decisions they did not make. Concurrence mistaken for leadership.
Scored on how much of the decision surface has explicit human ownership, how many named checkpoints exist in the workflow, and whether escalation criteria are documented.
How much original synthesis — writing, reasoning, problem-framing — is done by people rather than through a model. Whether anyone can still reason from first principles before reaching for one.
Catches: output without input. Reaction to a synthetic first answer, mistaken for thought.
Scored on capacity for original synthesis against AI dependency — and whether people can articulate their reasoning before reaching for a model.
Whether a framework exists for should this be automated, not only can it be. Who reviews model output. Who owns the mistakes when they ship.
Catches: velocity without review. Capability without conscience. Errors without an owner.
Scored on whether governance is structural or post-hoc, and whether there is a named owner when an AI-shaped decision goes wrong.
Whether people are being replaced by AI or elevated with it. Whether institutional knowledge is being captured, or walking out of the building with the people who hold it.
Catches: firing the people AI should be making unstoppable. Judgment bleeding out unnoticed.
Scored on the replacement-versus-elevation ratio, knowledge retention, and senior bench stability.
A standard has to hold against whatever a company has actually deployed, not against whatever is fashionable this quarter. This is the surface area a standard has to cover, and where the human layer fails in each.
I do not build these systems. I assess the layer around them — who decides, who reviews, and who owns what comes out. Drift does not live in the model. It lives in the surfaces the model is wired into.
Claude · GPT · Gemini · Grok · Llama · Mistral · Command · DeepSeek · Qwen — hosted, and open-weight deployments running inside a company’s own perimeter.
Where it breaksA company standardizes on one model’s register and stops hearing it. Fluency gets mistaken for correctness, and nobody can say which parts of a document a person actually decided.
Microsoft 365 Copilot · GitHub Copilot · Gemini in Workspace · Claude in the browser and the spreadsheet · Salesforce Agentforce · Glean · Notion AI · Slack AI · Cursor · meeting assistants across Zoom, Teams, and Meet.
Where it breaksThe wrestle moves upstream into the prompt. Documents get shorter, feedback gets thinner, and the argument that used to happen in the draft never happens anywhere.
RAG over SharePoint, Confluence, Notion, Drive, and the ticket queue · vector stores (Pinecone, Weaviate, pgvector) · enterprise search layers · Model Context Protocol connectors into internal systems.
Where it breaksThe index becomes the institutional memory. What it cannot surface stops existing, and nobody chose what got left out.
LangChain and LangGraph · LlamaIndex · CrewAI · AutoGen · Bedrock Agents · Vertex Agent Builder · n8n and Zapier automations that quietly became production.
Where it breaksAn action is taken with nothing between the decision and the consequence. When it goes wrong there is no stage that claims it, because no stage was ever assigned.
LangSmith · Braintrust · Arize · Weights & Biases · red-team and jailbreak testing · model cards · NIST AI RMF · ISO/IEC 42001 · the EU AI Act obligations landing on anyone operating in Europe.
Where it breaksThe dashboard is green and the traces are clean, and nobody has asked whether the people reading the output can still tell when it is wrong.
Every tool on this page measures whether the model is right.None of them measures whether the organization still is.
There is no shortage of writing about AI and the human mind. Most of it is correct and none of it is actionable, because a warning has no threshold. You cannot tell from an essay whether your work is drifting — and you certainly cannot tell whether last quarter's intervention worked.
A standard is different. It has axes, a score, bands, and a re-measurement. It can be failed, and it can be held.
One number,
four sub-scores, and
a threshold you can be
held to.
Each pillar scores 0–100, weighted equally at 25%. The composite is the Index — and where it lands tells you what kind of problem you have.
The number is the headline. The readout — what the score means in your context, and which two or three moves come first — is the document that actually gets used.
Anyone can publish an index. The hard part is what sits underneath it — the scoring method, the review layer, and the credential that lets someone other than its author run it and arrive at the same answer.
That is not new work for me. Three institutions came before The Human First Standard™. Every one of them has survived being handed to someone else, and thrived on the structures I built.
I go where the institution does not exist, build it, and measure whether it worked. The judgment stays, and you keep the instrument that scores it.
No transformation program, no culture initiative, no eighteen-month roadmap. A conversation, a diagnosis, and an installation.
A score taken once is a photograph.A score taken on a cadence is a discipline.
Book the free consultBook the free consult No deck, no pitch. I want to understand how you’re running AI, and where I can help.Drift and discipline produce the same weekly output. They produce very different organizations.
You do not lose the ability to work. You lose the wrestle with the idea, and with yourself. That friction is the work, and nothing has ever changed without it.
You bought AI to make the thinking better.
Without a checkpoint, AI makes thinking optional.
One. The wrestle — the friction — goes first. It reads as speed, and for two quarters it is.
Two. The people who wrestled with you go next. Their judgment was never written down, because nobody writes down the obvious.
Three. Whoever comes next never watches anyone decide. Documentation does not transfer judgment. Watching someone refuse the first answer does. When no one refuses, there is nothing to watch.
Four. Everything ships, and everything looks like a slightly watered-down version of a once-good idea.
Every step of this reads as a productivity gain and a better P&L.
That is why nobody stops.
You can buy the tools back. You cannot buy the wrestle back. Bring in someone sharp and put them somewhere with no place to argue — inside a quarter they stop arguing too.
And this is not only what happens to a company. It is what happens to you. Your work does not get worse. You just stop being the reason it was good.
Recovery costs close to what the whole thing is worth.
Move one is a question.
Move four is the company.
I measure move one.
The objections I hear most from senior leaders, answered directly.
No. I am not an AI ethics board. Ethics is one of the Four Pillars, and I score it structurally: whether you have a framework for should this be automated, not just can it be, and whether anyone owns the mistakes. The other three pillars measure Decision, Thinking, and Talent. The ethical failure and the performance failure are the same failure. I score it that way.
A policy states intent. The Human First Index scores behavior. Under the Ethics pillar I measure whether your governance is structural or post-hoc, and whether AI errors have a named owner. Then I score three pillars your policy does not touch: where decisions still have human checkpoints, whether original synthesis is happening in the building, and whether senior judgment is being elevated or quietly replaced.
Not AI-first, not anti-AI. The risk is not AI. The risk is unstructured adoption that erodes human thinking, decision-making, authorship, and responsibility. Nothing in the Standard asks you to integrate less or integrate slower. It asks you to keep defined human checkpoints where the consequences land. Moving fast and deciding well should not be a choice. The Standard is how you get both.
Consultancies sell efficiency frameworks, productivity stacks, governance checklists, change-management playbooks. None of them have a measurable Standard, and none were built by someone who had built accreditation infrastructure before. I built three. You leave with a score, a readout, and a prioritized roadmap. I deliver every engagement personally. You are not handed off.
I score you through documented review, interviews, and operational artifacts, and you keep four things: the locked HFI baseline plus four pillar sub-scores; an eight-to-fifteen-page readout written for your context; a prioritized roadmap tagged by pillar; and a sixty-to-ninety-minute live walk-through. The Audit names what is drifting. The Sprint installs the fix.
Individuals looking for personal productivity tips, and anyone who wants theory rather than applied practice — or the AI enablement vocabulary your team has already sat through three times this quarter. It is also not for anyone who needs the score to come back clean. The Audit is only worth running if you are prepared to act on what it finds.
The slow, unnoticed degradation of human thinking, judgment, and decision-making when an organization integrates AI without defined human checkpoints. Teams that used to argue stop arguing; original synthesis is replaced by reaction to a synthetic first answer; the work of thinking quietly gets outsourced. Invisible from the outside — the outputs still appear. What goes missing is the friction that used to make the work correct.
A measurable benchmark assessing how well an organization preserves human judgment as it integrates AI. It scores a company across the Four Pillars and produces the Human First Index. Where others write essays, the Standard defines something a company can actually be measured against.
What the Standard measures. Decision — where AI shapes decisions without a human checkpoint. Thinking — the capacity for original synthesis still done by people. Ethics — whether there is a framework for should this be automated, not only can it be. Talent — the replacement-versus-elevation ratio, and whether institutional knowledge is being retained.
The number a company receives. Each pillar scores 0–100, weighted equally at 25%, and the composite is the Index, reported with all four sub-scores and a band. Produced by the Audit as a locked baseline, and re-scored after a Sprint to confirm the gaps actually closed.
The discipline this work operates in: keeping human judgment intact as AI scales inside organizations. It extends cognitive systems engineering — Hollnagel and Woods, 1983 — out of the control room and into the company.
Treating the people behind AI — knowledge workers, leaders — as embodied systems under load, which is what they actually are. The load language is rooted in stress physiology and in sports and military science. What is mine is its application to cognitive work.
The Audit is a two-week diagnosis: the organization is scored against the Four Pillars and you keep the readout. The Sprint is a four-week installation of the human-first system — the checkpoints, the review layer, the rules that hold under pressure — followed by a re-score.
Through the Decision pillar. I score how much of the decision surface still has explicit human ownership, how many named checkpoints exist in the workflow, and whether escalation criteria are written down anywhere. The failure it catches is concurrence mistaken for leadership — a leader signing decisions a model actually made.
A readiness audit asks whether your infrastructure can support AI. The Audit asks the opposite question: now that AI is running, what has it done to the judgment of the people around it. Readiness is a before question. Cognitive Drift is an after problem, and almost nobody is measuring it.
Two weeks, scoped to your integration footprint. It starts with the complimentary consult, and I recommend the Audit only if that conversation says you need it. Most of those two weeks is review and interviews; the readout and the live walk-through land at the end of them. A larger organization takes more calendar, not more scrutiny.
Whoever owns the consequences when an AI-shaped decision goes wrong. On a team that is usually a COO, a Chief of Staff, a Head of Engineering, or the CEO directly. If you work alone, it is you. If nobody can answer that question, that is itself the finding — and it is the most common one.
No, and the reason is where the checkpoints go. They sit at the decision points where a wrong call has consequences, which is a small fraction of the surface AI touches in a working week. Everything else runs at the speed it already ran. Drift and discipline cost the same to operate. They produce very different companies.
The filter is not headcount. It is whether AI is already making or shaping decisions inside your work and nobody has defined where the human checkpoints are. That is true of a fifteen-person company moving fast and of a four-hundred-person company mid-rollout.
Yes — that is the point of a standard rather than an opinion. The same four pillars are scored the same way every time, so the number means the same thing in your work as in anyone else’s. It is also comparable against your own earlier score, which is what makes the re-measurement after a Sprint meaningful.
A score taken once is a photograph. A score taken on a cadence is a discipline. The organization is re-scored on an agreed rhythm to confirm the checkpoints are still there and still being used.
You don’t lose the team to the
model.
You don’t lose the judgment to the synthesis.
You don’t lose the company to the optimization.
You install a Standard,
and the Standard holds.
The Human First Standard did not come from nowhere. I trained in critical theory, and that is where the reading starts — not with AI commentary. I use these writers as instruments: I extend the argument where it is useful, and I do not reinvent what they already named.
Also load-bearing, in passing: Adorno and Horkheimer on the culture industry, Derrida on the language we inherit without choosing it, Laura Mulvey on the gaze automating, Kahneman where a room needs the shorthand, Mark Fisher and Berardi on why the present is built to make the past unreadable.
This is not a reading list assembled to look serious.
I learned to read the world at Dartmouth in the early 2000s, deep in the Baker-Berry stacks: as a research fellow under the late Lev Losev of the Russian department, chronicling the economy of gender in post-Soviet Russia; as a graduate student in Comparative Literature working out of the Geography and Women’s Studies departments, mining the geopolitics of nonprofit discourse for my master’s; as a teaching and research assistant to the feminist political geographer Jennifer Fluri, on the discursive construction of self and nation; and as an adjunct professor, delivering the credit-led sociology curriculum I wrote for the women’s prison in Vermont. I didn’t know it then. I was being groomed for the work I do now.
My professors were adamant: if you don’t know critical theory, you will be a tourist in your own time. Everyone else told me I was wasting an education on dead theory. Then 2008 arrived, I was awarded a Fulbright, and most of my class was unemployed.
I have been making this argument ever since, and I believe the canon matters more now than at any point in my lifetime. None of this is derived from a model. It is derived from more than twenty years of reading the writers the world had already buried — and then building institutions with what they said.
So here is the challenge. Benjamin wrote about what happens to a work when it can be reproduced without limit — what leaves it, and what we agree to stop noticing once it is gone. That was 1936. I have yet to meet anyone building or selling AI who is not walking straight into that essay, and almost none of them have read it.
Read my paper on Benjamin.
Then tell me Cognitive Drift is a marketing term.