# How to answer ethical concerns about AI

> Source: <https://www.theaithinker.com/p/how-to-answer-ethical-concerns-about>
> Published: 2026-07-26 23:00:46+00:00

# How to answer ethical concerns about AI

### Listen to the strongest version of the concern. Sort it into one of four kinds. Concede what’s true, fix what you control, correct what’s mistaken, respect what’s a value.

Every time I talk about AI with someone, the conversation ends up in the same place. Not on models. Not on prompts. **On ethics.** Someone says the data centers are draining water tables. Someone says the training data was scraped from people who never agreed. Someone says they became a designer to design, not to review a machine’s output. **The objections that stick aren’t about capability anymore, they’re about conscience.**

In early 2025, I watched a French designers’ collective called Designers Éthiques lay out every one of these objections in [one 40-minute talk](https://peertube.designersethiques.org/w/jhHv5GLE5iV5ccYgs8rWe9). It’s stayed with me since. Not because it was contrarian, but because it wasn’t. These were rigorous professionals making the strongest version of the case: an eco-design specialist, a design researcher, an ergonomist. No trolling, no panic. **It’s the sharpest compact map I’ve found of what your own team members are thinking and mostly not saying in the meeting.**

If you’re leading a team through an AI transition, you’ve met these objections too. Maybe in a retro, maybe in a 1:1, maybe as a silence that never turns into usage. And you’ve probably been handed exactly one playbook for them: overcome the resistance. I think that playbook is wrong. **Your skeptics are mostly raising real problems, and the fastest way to lose them is to debate.**

The better move: **sort each concern into one of four kinds.** Some are true, and you change what you adopt. Some are risks you control, and you change how you adopt. A few are misconceptions, and you correct them with evidence instead of marketing. One or two are values, and you respect them. **The payoff isn’t a converted team. It’s AI use your team can defend out loud.**

The playbook, at a glance:

**The trap.** I’ll show you why winning the argument against a skeptic loses the team.**The sort.** Here’s the four-kind triage that replaces the debate: true, fixable, misread, or a value.**The map.** You get the seven concerns you’ll hear most, each with an honest verdict, the move that follows, and the one resource worth forwarding.**The practice.** What a defensible team AI posture looks like once the sorting is done.

By the end, you’ll have a verdict and a concrete next move for the seven objections coming your way this quarter, plus the one sentence that keeps a skeptic on your team. **You stop dreading the ethics conversation and start using it to make your team’s AI practice sharper.** No slides required, no philosophy degree either.

Let’s sort this out.

## Why winning the argument loses the team

[The talk](https://peertube.designersethiques.org/w/jhHv5GLE5iV5ccYgs8rWe9) opens with a moment I can’t stop thinking about. At a green-IT conference workshop, in front of the most skeptical crowd available, the speakers ran a session asking “AI: in or out?” Almost every group landed on “in,” reasoning that it’s here anyway, there’s no choice, so let’s make it as clean as possible. One speaker was troubled by exactly that phrasing. **”We have no choice” is not what agreement sounds like, it’s what resignation sounds like.** If your team adopts AI in that spirit, you didn’t win them. They just stopped telling you things.

There’s a second reason the debate is rigged before you open your mouth. **Your team members already met AI adoption as users, and it wasn’t polite.** The design researchers at Limites Numériques [documented the pattern](https://limitesnumeriques.fr/travaux-productions/ai-forcing): AI buttons pushed front and center, features switched on by default like Strava’s Athlete Intelligence, dialogs that offer “try it” and “not now” but never “no.” Sparkles and purple everywhere, the visual vocabulary of magic. **When you pitch AI to your team with vendor enthusiasm, you pattern-match to the forced adoption they already resent.** They’ve heard “this will make everything better” before, from a button they couldn’t refuse.

Organizational research has said this plainly for years. In their 2008 Academy of Management Review paper, [Ford, Ford and D’Amelio](https://journals.aom.org/doi/abs/10.5465/amr.2008.31193235) argued that r**esistance to change isn’t a defect in the resisters.** Change agents cause a good share of it themselves, and **resistance is better treated as a resource:** engagement, feedback, proof that people take the change seriously. **The person pushing back is often the person paying the most attention.**

If you read [How to lead a tech team through the AI shift](https://www.theaithinker.com/p/how-to-lead-a-tech-team-through-the), you might spot a tension here. I argued there for the 20-60-20 rule: **pour your energy into your champions and the watching middle, and stop exhausting yourself arguing with the resistors.** I stand by every word, and this article doesn’t change the math. Sorting is not arguing, and what follows is not a conversion campaign. It’s for the conversations that find you anyway: the 1:1 where a concern lands on the table, the team meeting where a hand goes up. **You still don’t chase your skeptics; you answer well when they’re in front of you, because the watchers are scoring how you do it.** One honest answer to a skeptic moves ten watchers.

And there’s simple arithmetic about credibility. **Argue once against something that turns out to be true, say the data-center water numbers, and you lose the room for a quarter.** The skeptic came with figures. You came with talking points. So the job isn’t to win. It’s to sort.

## Sort the concern before you answer it

Here’s the move that changes the conversation: **treat each concern as a claim to classify, not an attack to parry.** A triage nurse doesn’t argue with symptoms. She figures out what kind of problem she’s looking at, because the kind determines the response. Four kinds cover almost everything your team will raise.

**Kind one: they’re right.** The data-center buildout really is doubling electricity demand. When a concern is true, the honest answer starts with “you’re right,” and the follow-up is a change in what you adopt.**Nothing builds credibility faster than a concession nobody expected.****Kind two: a real risk you control.** Juniors losing the review and mentoring that builds expertise. This one isn’t about the industry, it’s about your team, which means your working agreements decide whether it comes true.**The answer is a rule you set together, not a rebuttal.****Kind three: a misconception.**“Every prompt is like pouring out a bottle of water.” The per-prompt numbers are knowable, and they don’t say that.** Correct the arithmetic gently, with sources, and without dismissing the worry underneath it.**The worry usually points at something real at a different scale.** Kind four: a value.**“I didn’t get into this craft to supervise a machine.”** You can’t argue someone out of a value, and trying reads as disrespect.**What you can do is design roles so the value survives, and be honest about where the boundary sits.

One subtlety makes the whole sort work: the same concern can live in two kinds at once, at different scales. The environmental objection is true at industry scale and misread at the single-prompt scale. The sort forces you to say which scale you’re answering, out loud. **Sorting is your actual job in this conversation; the debate is optional.**

Here’s what the sort sounds like in an actual 1:1, in three moves.

**First, ask for the strongest version:**“Give me the strongest version of this worry. Convince me.” That one question replaces the debate with respect, and it surfaces the real concern instead of the polite one.**Second, name the kind out loud in plain words:**“I think you’re right about this one,” or “this one is ours to fix,” or “I think the numbers say otherwise, let me show you,” or “that sounds like a value, and I won’t argue with a value.”**Third, before the conversation ends, commit to one follow-up in writing:** the rule you’ll add, the number you’ll check, the thing that goes out-by-default. A sort that doesn’t end in a written commitment is just a nicer way of nodding.

So let’s put the framework to work, one conversation at a time.

## Answer the seven concerns, one by one

The map below covers the seven objections I keep hearing. Six of them the talk maps better than anything else I’ve found; the seventh, what AI does to our thinking, has grown loud enough since to earn its own place. Your team may add others: discrimination baked into training data, deepfakes, privacy. **The sort handles those the same way.**

### The planet

Give this concern its strongest version, with numbers that survive checking, because the talk’s own figures were compressed in places. The [IEA’s 2025 Energy and AI report](https://www.iea.org/news/ai-is-set-to-drive-surging-electricity-demand-from-data-centres-while-offering-the-potential-to-transform-how-the-energy-sector-works) projects **data-center electricity demand will more than double by 2030**, to around 945 TWh, slightly more than Japan’s entire consumption today. [Alex de Vries](https://doi.org/10.1016/j.joule.2023.09.004) projected in Joule that **AI-specific consumption alone could reach the scale of the Netherlands or Argentina by 2027.** Water follows the same curve: [Microsoft’s use jumped a third in 2022 while Google’s rose a fifth](https://www.datacenterdynamics.com/en/news/microsofts-water-consumption-jumps-34-percent-amid-ai-boom/), and researchers behind the [Making AI Less “Thirsty” paper](https://arxiv.org/abs/2304.03271) project **AI water withdrawal reaching half of the UK’s annual total by 2027. The verdict: at industry scale, your skeptic is right.**

There’s a sharper twist worth conceding too. A [Guardian analysis](https://www.theguardian.com/technology/2024/sep/15/data-center-gas-emissions-tech) found **the real emissions of the big providers’ own data centers ran about 7.6 times higher than officially reported.** The trick is accounting: renewable-energy certificates let a company report clean power it never actually consumed at the site. **Notice what that is: not a technology problem, an honesty problem.**

Now the misread half. Per-prompt costs are measurable, and they’re small. [Google’s published figure](https://cloud.google.com/blog/products/infrastructure/measuring-the-environmental-impact-of-ai-inference) for a median Gemini text prompt: **0.24 Wh of energy and 0.26 mL of water, counting on-site water only.** [Mistral’s lifecycle study](https://mistral.ai/news/our-contribution-to-a-global-environmental-standard-for-ai/) counts everything upstream, training included, and lands at 45 mL per response. The gap is scope, not virtue, and either way one prompt sits far below one kilometer of driving or one beef meal. **Correcting this arithmetic matters, because a team member who quietly believes each prompt burns a lake will never use the tools well.** The obvious comeback is fair: small times billions of prompts equals the buildout we just conceded. Both things are true. Your prompt is cheap; the trajectory is not; your subscription is a vote for it.

**The move is to stop pretending it’s free, and to put that in writing.** Three rules are worth adding to your team’s working agreements.

**Default to the small model, escalate on need:** most tasks don’t need the frontier model, and the price gap is the energy gap.**No AI where a script does the job:** a regex, a spreadsheet formula, or a cron job costs nothing and never hallucinates.**And one screening question before any new AI use ships:** what is this for, and would we defend it at scale?

Expect efficiency gains to get eaten unless you watch for them; the talk calls this the rebound effect, where every drop in cost per task invites more tasks.

When the conversation deserves more than a meeting slot, forward ** Andy Masley’s cheat sheet on AI and the environment.** It’s the single best document I know for exactly this conversation: it runs the per-prompt arithmetic against everyday life (about 1,000 prompts to move your daily energy use by 1%) and still concedes the industry-scale buildout is a real question.

**Handing your skeptic the strongest version of the counter-arithmetic beats paraphrasing it in a meeting.** What you can say: “You’re right about the trajectory. Here’s what our use actually costs, and here’s what we choose not to use it for.”

That answer covers the machines. The people who made the models possible are a harder conversation.

### The people behind the data

The concern: **the models were trained on work scraped from people who never consented, and the labor that made them safe was outsourced at poverty wages.** The record backs it. [TIME’s investigation](https://time.com/6247678/openai-chatgpt-kenya-workers/) documented Kenyan workers labeling toxic content for less than $2 an hour. And the talk’s own example has a scholarly source: [Le Ludec, Cornet and Casilli](https://journals.sagepub.com/doi/10.1177/20539517231188723) documented French AI firms outsourcing annotation to Madagascar at poverty wages. This isn’t someone else’s supply chain. **The verdict: they’re right, and no arithmetic rescues it.**

So don’t reach for a rebuttal, because there isn’t one. What a product team controls is procurement and posture, and you can make both concrete. **Write three questions into your next tool evaluation:** where does the training data come from, what licensing deals exist, and what are the annotation labor practices. The differences between providers are inspectable: **Mistral publishes a full lifecycle audit** with the French environment agency, and some providers sign licensing deals with the newsrooms they train on, the way OpenAI did with the

[Associated Press](https://www.axios.com/2023/07/13/ap-openai-news-sharing-tech-deal)and

[Le Monde](https://openai.com/index/global-news-partnerships-le-monde-and-prisa-media/), instead of scraping them. Your team can prefer the better ones and say why. For the teammate who wants the full picture,

**the book to pass around is** written by three Oxford Internet Institute researchers from years of interviews with the annotators and moderators themselves.

[Feeding the Machine](https://canongate.co.uk/books/5234-feeding-the-machine-the-hidden-human-labour-powering-ai/),The other half of the move is harder: don’t launder the discomfort. **Deciding to use AI with open eyes is more respectable than pretending the problem is solved.** Some weeks the honest position is “this bothers me too, and here’s why I still think our use is defensible.” What you can say: “You’re right, and we can’t fix it from here. Here are the three questions we now ask before picking a provider, and here’s where I still think it’s worth it.”

The next objection sits closer to home: what the shortcut does to your own team’s craft.

### The craft and the shortcut

This is the worry I hear most from the best people. Expertise, the ergonomist in the talk explains, gets built four ways: **learning, experimentation, collaboration with peers, and teaching.** Then he asks the question that lands hardest: when you hit a problem, what’s more effective, **asking your colleagues or asking the AI?** In the Q&A, someone described a senior developer who used to review a junior’s code and now just asks the AI instead. **The transmission chain broke, quietly, without anyone deciding it should.**

Lisanne Bainbridge named this pattern in 1983. In [Ironies of Automation](https://www.sciencedirect.com/science/article/abs/pii/0005109883900468), the paper to forward to the teammate who wants the lineage, she showed that **automating the easy parts of a job leaves the human with the hard parts, minus the daily practice that kept them sharp.** A [field study of AI in a radiology practice](https://link.springer.com/article/10.1007/s00146-024-01951-x) by Gamkrelidze, Zouinar and Barcellini found exactly that shape: the fracture-detection AI worked as a genuine safety net, and it also bred overconfidence risk, while the transcription AI quietly deleted the verification step the medical secretaries used to perform. **The verdict: this risk is real, and it’s the one you control most directly.**

Which is good news, because **the fix is working agreements,** not philosophy, and they’re short enough to write down:

**Rule one:** anyone still learning a domain gets human review on their work, full stop, and the reviewer is a person with a name.**Rule two:** AI explains before it produces; used as a tutor it strengthens the learning station instead of skipping it.**Rule three:** seniors keep protected teaching time (a weekly pairing hour, a monthly brown-bag), because teaching is their own expertise engine too.

**The shortcut only breaks the loop if you let it become the default path.** What you can say: “You’re right that the shortcut is real. So we keep the loop: **review stays, teaching stays, and AI joins the loop instead of replacing it.**”

Working agreements protect the team’s craft. The newer worry is what the tools do to each person’s head.

### The thinking muscle

This is the objection that grew loudest over the past year, and it usually arrives as a sentence about the next generation: **AI is making us dumber, and kids who grow up on it will never learn to think for themselves.** It now arrives with a study attached. An [MIT Media Lab experiment](https://arxiv.org/abs/2506.08872) put EEG caps on 54 essay writers: the ChatGPT group showed the weakest brain connectivity, reported the least ownership of their work, and minutes after submitting, most couldn’t quote a line from their own essay. The authors named the lingering effect **“cognitive debt.”** A [survey of 666 people](https://www.mdpi.com/2075-4698/15/1/6) added that the heaviest AI users scored lowest on critical thinking, with the youngest participants the most dependent. **This worry has data now, and waving it off with “they said the same about calculators” doesn’t answer it.**

**The verdict: real at the learning edge, unproven as a sentence about humanity.** Read honestly, the evidence is narrower and more useful than the headlines. The MIT authors answer “is ChatGPT making us dumber?” with a flat no in their own FAQ: one narrow task, a small sample, engagement measured during the task rather than lasting change. The 666-person survey is correlational; maybe people who think less critically simply offload more. The result that should actually shape your team’s rules is causal: a [randomized trial with about a thousand students, published in PNAS](https://www.pnas.org/doi/10.1073/pnas.2422633122), found that **vanilla ChatGPT boosted practice performance by 48% and then cut unassisted exam scores by 17%.** Same trial, one design change: a tutor version that gave hints instead of answers erased the harm completely. **The variable isn’t AI. It’s whether the AI does the thinking or scaffolds it.** That’s also exactly where the calculator comparison holds and breaks: fine to offload arithmetic, dangerous to offload the reasoning itself.

**The move is to make “coach, not answer machine” a written rule for anything a person is still learning.** Three rules fit on a sticky note

**Think first, prompt second:** write your own answer, even five rough lines, before opening the chat.**Hints before answers while learning:** tell the model “ask me questions, don’t give me the solution” (the exact design that erased the harm in the trial).**Explain it back:** if you can’t restate what you shipped with the chat window closed, it isn’t yours yet.

For the teammate who wants the deeper story, **forward The Conversation’s walkthrough of the MIT study;** it takes the worry seriously and shows precisely what the study can and can’t say. What you can say: “You’re right to guard your thinking. So

**we use AI as a coach on anything we’re still learning, and we prove it by explaining our work with the chat closed.**”

Guarding heads is one half of consent. The other half is who signed up for any of this in the first place.

### Consent at work

The talk asks a question I’d put to any team verbatim: **do you consent to your work being organized by your team, or by an AI?** Then it sketches the scenario: a backlog where tasks are assigned by an AI, estimated by an AI, and timed by an AI that checks whether you matched its estimate. If that sounds like science fiction, the EU disagrees. The [EU AI Act’s Annex III](https://artificialintelligenceact.eu/annex/3/) classifies **workplace AI that allocates tasks or monitors and evaluates performance as high-risk**, with employer duties to inform workers before deploying it. The high-risk obligations start applying in August 2026. That’s next month. **The verdict: they’re right about the pattern, and whether it becomes true on your team is your call to make.**

Reality in the field is still mostly softer than the fear, and it’s worth being precise about that. The talk pointed at the French public sector, where the Lyon school district runs an AI copilot called [Cassandre](https://www.ekole.fr/blog/le-rectorat-de-lyon-utilise-lia-pour-aider-les-gestionnaires-rh-dans-leurs-reponses-aux-enseignants) that helps HR staff answer teachers’ questions about transfers. It doesn’t decide the transfers. T**he speaker’s warning is about where the road leads: to “it wasn’t a human who decided,” the sentence that ends accountability.**

Parts of the private sector are already further down that road, and the two clearest cases come from companies everyone knows. [Amazon’s own legal filings](https://www.theverge.com/2019/4/25/18516004/amazon-warehouse-fulfillment-centers-productivity-firing-terminations) describe warehouse systems that track each worker’s rate and auto-generate productivity warnings and termination paperwork, with Amazon noting supervisors can override. And in 2023, [an Amsterdam appeals court](https://www.fieldfisher.com/en/insights/amsterdam-court-of-appeal-rules-in-favour-of-uber-and-ola-cabs-drivers) ruled that Uber had deactivated drivers through solely automated decisions, the practice the drivers’ union called robo-firing and the exact thing the GDPR’s Article 22 exists to prevent. **The road the talk warns about doesn’t lead somewhere hypothetical; it leads to case law.**

The move is to **make adoption consent-real before anyone asks,** with three rules written into your working agreements:

Tools are offered, never mandated per person.

Every decision that touches a person carries a named human owner, and the name is written next to the decision.

Opting out is safe, with no career shadow, and you check twice a year that it’s still true in practice.

When legal wants the source, forward the [Annex III text itself](https://artificialintelligenceact.eu/annex/3/); it’s shorter than most policy memos. **Nobody who works for you should ever hear “the AI decided.”** What you can say: “A human owns every decision on this team, starting with me.”

And underneath every consent question sits the one your team member is actually asking.

### The jobs

The ergonomist tells a story from his early career. His mentor asked: if you design software that makes the work twice as fast, how many workers are left in six months? It’s the question underneath every other question, and your team member asking it deserves better than the two stock answers: the vendor’s “nothing changes for you” and the doomer’s certainty. The talk makes the hollowness concrete with translators, a craft told for years that the tools would free them for “more interesting work,” by people who never say what work is more interesting than the craft itself. You can’t answer the mentor’s question for the industry. Nobody can. **The verdict: partly true, partly unknowable, and the worst response is a promise you can’t keep.**

Start by questioning the premise. “Twice as fast” itself deserves doubt: [METR’s randomized trial](https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/) found **experienced developers were 19% slower with AI tools while believing they were 20% faster.** That perception gap is the strongest argument you’ll ever get for measuring your own team’s gains instead of quoting the pitch. And watch what the work becomes, not just whether it exists. When France’s tax authority deployed its pool-detection program [Foncier innovant](https://www.impots.gouv.fr/actualite/generalisation-du-foncier-innovant), field agents became desk agents. The job survived; the parts of it people liked didn’t all survive with it. **Even the jobs that stay change texture, and naming that out loud buys more trust than any reassurance.**

**The move is to commit only to what you control, and to make each commitment checkable.** Plans get heard from you first, not from a rumor: say it in the team meeting and repeat it every quarter. Skills investment happens on company time, with a number attached (a monthly learning day, a named course budget); an unnumbered commitment isn’t one. Roles get redesigned with the people in them, not around them: the person whose work is changing sits in the room where it’s redesigned. And when someone wants more than reassurance, **forward David Autor’s essay on AI and the middle class:** MIT’s leading labor economist, long a tech skeptic, arguing a future with neither doom nor denial in it. What you can say: “I can’t promise what the industry does. I can promise you’ll hear plans from me first, and that

**we invest in you while things change.**”

One objection left, and it aims at the output itself.

### The output you can’t trust

The talk’s most memorable stretch is about mediocrity. **Models weight toward the most probable output.** Ask for a pink elephant with wings and most image models fail, because the training data holds no pink elephants with wings. Ask for a wine glass filled to the brim; you get one filled halfway, over and over. The speaker’s conclusion: a designer who uses AI to generate ideas is stepping onto “a highway to the most probable solution,” and originality lives off that highway. Stereotypes ride along for the same statistical reason. **The verdict: half right, and the right half is a map of where to use the tools, not a case for refusing them.**

Because notice what the talk itself endorses: transcription, subtitles, handwriting recognition, classifying feedback. Small, checkable tasks with human expertise in the loop. That generalizes into the single most useful screen I know, and it’s worth writing exactly as a rule: **AI-assisted work ships with a named human checker, and if nobody can check it, it doesn’t ship.** Research gives the screen teeth. A [CHI 2025 study of 319 knowledge workers](https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/) by Microsoft Research and CMU found that **the more people trusted the AI, the less critically they thought; the more they trusted their own judgment, the more critically they engaged.** The check only happens when the human still believes their judgment beats the tool’s, so your job is to **keep that belief alive.** Taste, direction, and originality stay human, and they get more valuable, not less. For the teammate who wants the domain-by-domain version of “what AI can and can’t do,” **the book is AI Snake Oil,** by two Princeton computer scientists who respect skeptics enough to argue with evidence. What you can say: “You’re right about the median. That’s why

**we use it where we can verify, and why your judgment is the part that never gets delegated.**”

Seven concerns, seven verdicts. Put together, they stop being a list of complaints and start being something else: the specification for a practice.

## The objection to this whole approach

A sharp skeptic will see the framework coming and name it: “You’d already decided we’re adopting. The sorting is just a softer road to the same forced outcome.” **You might be silently raising the manager’s version of the same objection: there’s a mandate, there’s a quarter, and this looks like a philosophy seminar.** Both deserve a straight answer.

Take the skeptic’s version first, because it has a clean test. **The sort has real exits, or it’s theater.** If the “true” verdicts changed what you adopt, if something on your roadmap moved to out-by-default, if the person with the values objection still has their standing and their role, then the sorting was real. If you ran the conversations and nothing in your plan changed, the skeptic was right about you, and no framework fixes that.

Now the mandate pressure. Here’s the thing I’d want leadership to hear: **the triage is the fast path, not the slow one.** The alternative to honest sorting isn’t a converted team. It’s tools that get opened once, licenses that show up in the usage report as a flat line, and an adoption metric nobody believes. **Quiet non-use is the most expensive outcome an AI mandate can produce, and steamrolling buys it wholesale.** Sorting removes the actual reason people were stalling. That’s not a seminar. That’s unblocking.

And when the sorting is real, something better than compliance shows up: a posture you can write down.

## The practice you end up with

**Pull the seven moves together and you get a short team AI charter, written from your team’s concerns instead of a vendor deck.** Mine would fit on a page. Four blocks.

**Where AI is in:** tasks with verifiable outputs that a human reviews and owns. Transcription, classification, drafts that get read before they ship, code reviewed by a person who could have written it.**Where it’s out by default:** original ideation, unreviewed customer-facing output, and any decision about people.**Who owns what:** tools offered, never forced on a person; a named human accountable for every decision; an opt-out without career shadow.**How we pick tools:** providers that publish honest numbers on energy, water, and training data.

That last block has teeth because of the reporting gap from the planet section. A **provider that understates its emissions 7.6 times has already told you how it handles inconvenient numbers.** “Publishes honest figures” is a real screen, not a platitude. It costs nothing to apply and it sorts the market fast.

You can draft the whole thing in one team hour, and drafting it together is most of its power. **Fifteen minutes: everyone writes their concerns down, anonymously if that helps people say the real one.** Fifteen minutes: sort them into the four kinds together, out loud. Twenty minutes: draft the four blocks from what actually came up, not from a template. Ten minutes: put one name next to each block and a date three months out to revisit. **A charter the team wrote in an hour beats a policy nobody read.**

Two things make this charter travel further than your team. Legal will notice that the consent block is where EU compliance is already heading, with workplace AI obligations applying from August 2026. And the room where adoption metrics get discussed will notice that a written posture beats a usage dashboard nobody trusts. **The line that carries all of it: “You’re right about X. Here’s what we changed so our use doesn’t have that problem. And here’s where I still think it’s worth it.”** A manager who can complete that sentence for each concern doesn’t need to win any argument.

One reframe before you go write yours. I started this project wanting keys to change skeptical minds. I ended it convinced that’s the wrong target. **Mindsets move when the practice becomes trustworthy, not when the pitch improves.** The skeptic doesn’t need to love AI. **They need to see that their concern changed something real.**

## Trust your skeptics

Somewhere in your team is the person who read the water numbers, the one who worries about the junior who never gets reviewed anymore, the one who wonders what these tools are doing to how we think, the one who quietly decided prompting isn’t the craft they signed up for. They were never the obstacle. They were the free red team, stress-testing your AI practice before it shipped. The org-behavior researchers knew it in 2008; the talk’s speakers knew it on that stage in early 2025. **A team that sorted its concerns ships AI use it can defend out loud, starting with defending it to itself.**

Your turn. **Take the next 1:1 where a concern comes up and sort it out loud, all four kinds on the table.** Write the strongest version of your best skeptic’s objection before you meet them, and open with it. Draft the four-block charter with the team, not for them. Pick one thing that goes out-by-default this week, and say why. And [watch the talk](https://peertube.designersethiques.org/w/jhHv5GLE5iV5ccYgs8rWe9) that started this article, then forward it to the person on your team it will resonate with, with a note saying you want their read.

The speakers closed with a warning: **AI will only ever offer you a mediocre dream, so take the time to dream your own.** I’d turn it forward. Your team already has a dream about what their work should mean. **Make your AI practice serve that, on purpose, and the skeptics won’t need convincing, they’ll be the ones holding you to it.**