Trauma-Informed Content Consulting

Anthropic released Claude Opus 4.8 on Thursday. The headline claims more conservative scoping, more honest acknowledgement of uncertainty and higher scores on alignment with user interests. It supposedly won’t make claims when it’s not sire of the facts. Sounds great in theory.

Rembrandt Editor, the trauma-informed content review tool I’ve been building, currently runs on Claude Sonnet 4.6, an older model from the same family. When the model maker releases an upgrade claiming exactly the qualities a content review tool depends on, including careful scoping, honest acknowledgement of limits and prosocial alignment, you have to investigate. So I did.

Yesterday I ran the same three pieces of content through both models: a police survivor page, a contractor timesheet email and an NHS-style breast cancer information page.

The result was clearer than I expected. I’m staying on Sonnet 4.6.

Opus 4.8 is genuinely more honest. It scopes regulatory and accessibility flags more conservatively, sticking to the criteria that actually apply rather than stretching to fit. It models the reader’s emotional state more sharply on stakes-high content: “a survivor reading this in distress is likely to be running a parallel internal process: was what happened to me real enough?”

Sonnet 4.6 catches more specific structural things. A naming collision where “invasive breast cancer” is used as both general category and specific subtype on the same page. A 43-word sentence that buries the action under escalation language. Two date references the reader has to reconcile in opening sentences.

They’re complementary, not interchangeable. Opus is stronger on psychological reader-state observation. Sonnet is stronger on structural analytical pattern detection. Across the mix of content Rembrandt actually reviews, Sonnet’s pattern detection earns its place more reliably.

There’s also a 67% cost premium on Opus 4.8 at standard pricing. That matters less if you’re running a few reviews a day, more if you’re operating a tool at any kind of scale.

For trauma-informed content design specifically, I’d expected Opus to be the obvious upgrade. It does add something. Just not consistently enough across content types to justify the cost premium or strong enough to defend “Pro runs on Opus” as product substance rather than marketing positioning.

Rembrandt Editor was created with input from experienced content designers. The free version is available now, with a Pro version launching in June.

https://www.rembrandteditor.com

Rembrandt — Trauma-informed content review Trauma-informed content review for readers in living experience. UK lens: FCA Consumer Duty, ISO 22458, GDS, WCAG 2.2 AA. EU and US lenses available.

Thanks to the experienced content practitioners who've spent the past few weeks putting my new trauma-informed content review tool Rembrandt Editor through its paces. The feedback has been more useful than I had any right to expect.

Plenty of people will tell you that any Al writing tool is, by definition, another lump of slop that strips out the critical thinking, compassion and human element good content depends on. It's a reasonable suspicion. It's also why I cared what experienced practitioners thought once they'd used Rembrandt Editor on real work, rather than reacting to the category.

A selection of what came back, names off because the testers were generous enough already:
- "This is fabulous! Super useful, a really interesting and genuinely transformative tool."
- "I was really impressed by what it flagged and the suggested improvements. I think it is genuinely an incredibly useful tool."
- "The first Al writing tool that is genuinely helpful for a writer who is already a professional. Rembrandt revealed to me the handful of places where I had gone too far in trading clarity for reassurance. Also, the copy suggestions were more nuanced than other writing tools I've used."
- "Oh, it is GOOD! I really like how it tells you ways to tighten your copy and calls to action."
- "The 'specific issues' are well considered and offer useful tips. I especially like how the 'try instead' callout opens with the frontloaded instruction of how to improve the content."

That's the bit I find most encouraging, testers describing the tool as something that surfaces trade-offs and sharpens judgement, not something that does the thinking for them.

Which is the design intent. Rembrandt flags plausible concerns. It doesn't adjudicate compliance, replace a content designer or pretend to have judgement it doesn't have.

If you volunteered to test and haven't got round to it, your feedback is still much appreciated. Have a go. Tell me what's wrong with it. The next round of changes will be shaped by what testers find, not by what I think the tool should do next.

I have spent months trying to create a credible working version of my trauma-informed content analysis tool, Rembrandt Editor. It now reviews content, gives detailed feedback and suggests a rewrite as an experienced trauma-informed content designer.

I'm opening a closed beta to a small group of practitioners and would value your feedback.
DM me if you'd like to take part. Run a couple of pieces through (text only or PDF at this stage) and tell me whether it sounds like a trauma-informed content practitioner or a generic content tool, whether the framework is visible in the selected language and jurisdiction and anything that surprised you. You can send me feedback in the tool or reply here.

There is no deadline. Even a quick reaction is useful. Thanks in advance.

The UK government just published comprehensive AI skills research. You may have seen headlines about a related government announcement offering everyone a 20-minute training session to learn how to write basic AI prompts. But there's a lot more to it.

The research covered 139 literature reviews, surveys of 1,189 people and 801 employers, and multiple work packages. This required serious work and serious money.

Their findings are that 73% of people think they've used AI, but only 17% can explain it. Two-thirds distrust AI products. Women are less confident and get less training. Employers can't identify use cases, but are deploying systems anyway.

Their conclusion is that workers need more AI skills training. I think that this is fundamentally wrong. The distrust isn't a deficit, but a rational response to observable harm. When people see AI hallucinating medical advice and biased hiring tools, it's not a failure to understand AI. People understand it perfectly.

The real problem isn't that workers lack skills. It's that organisations are deploying systems without checking if they work, who they harm or what alternatives might serve people better.

Workers aren't the problem. Workers ARE the economy. But this research treats them as inputs to optimise rather than people whose lives matter. This is the classic move of parking moral questions in technology projects. Instead of asking "Is this good?", they ask "How do we skill up for this?"

My analysis is now available online.

The AI skills gap doesn't exist: why government research blames workers for organisational failures - Workplace Insights by Adrie van der Luijt Analysis of UK government AI skills research reveals how worker "skills gaps" frame masks organisational deployment failures.

I've just written about the government's new digital roadmap for modernising public services.

I worked on my first government digital transformation project in 1987, years before the internet. After all these years, I've learned to read these documents for what they're designed to deliver versus what they promise.

The roadmap has genuine strengths, including real transparency commitments, an honest assessment of outdated systems, and a focus on building expertise rather than relying on consultants. But there are structural blind spots that worry me. We're measuring efficiency and satisfaction, not outcomes for the people most likely to need help. We're treating digital inclusion as a training problem rather than a design principle.

My article breaks down what works, what's missing and what trauma-informed design would actually look like.

You can read my article at https://www.workplaceinsights.co.uk/government-digital-roadmap-2025-critique/

Reading the government's digital roadmap as someone who's seen this film before - Workplace Insights by Adrie van der Luijt A trauma-informed critique of the Government Digital Roadmap, including gaps around vulnerable users risk turning transformation into structural exclusion.

I've been thinking about accountability in government digital projects. GDS already requires testing with vulnerable users and documentation of design decisions. But there's a difference between following the process and being personally accountable for whether content works for people in crisis.

What would happen if government digital had to follow the same rules as financial services, where named individuals have to prove they took reasonable steps to prevent harm?

I've written about what that would look like in practice, drawing on three decades of content work - including major government programmes. It's not about adding bureaucracy. It's about making the people with power to change things actually accountable for the judgment calls that matter.

Here's my new article on what that would look like in practice and why it would actually help content practitioners do better work: https://www.workplaceinsights.co.uk/government-digital-accountability-fca-rules/

Government digital accountability and FCA accountability rules - Workplace Insights by Adrie van der Luijt What would happen if government digital projects had to follow the Senior Managers and Certification Regime, the same rules as financial services?

The poverty premium has gone digital. One million UK households cancelled their broadband because they couldn't afford it last year. Now McDonald's, Tesco and most retailers charge more to people without apps.

This isn't about technology. It's about structural digital exclusion.

When 6 million households struggle to afford communications services, "just use the app" isn't customer service. When 10% of people cut food spending to pay for internet, app-exclusive discounts aren't rewards.

When 69% of eligible households don't know social tariffs exist, the market isn't working.

Financial services: this is what FCA Consumer Duty should address. Not whether your app passes accessibility guidelines, but whether people who don't use it are paying more.

You can read my article at https://www.workplaceinsights.co.uk/app-tax-digital-exclusion/

The app tax: how digital exclusion creates a poverty premium - Workplace Insights by Adrie van der Luijt One million UK households cancelled broadband they couldn't afford. Now essential services charge more to people without apps, forcing digital exclusion.

I have just seen the Sky News investigation into government AI spending. Since 2018, we've spent £3.35bn on AI contracts. Want to know where your tax money's going?

The Met Office got a billion pounds for a weather forecasting supercomputer. Yes, you read that right. A billion. For weather predictions.

Meanwhile, the Department for Work and Pensions, the one that handles benefits for disabled people, unemployed people, people in absolute crisis, is in the bottom three for AI spending. So is HMRC, which you'd think might benefit from some help making tax less of a nightmare for ordinary people.

Transport for London got £259m. Because apparently optimising tube routes matters more than stopping the benefits system from driving people to breakdown.

I worked on billion-pound government projects since 2012. Even then, we knew the real challenge wasn't the technology. It was understanding why people in crisis give "wrong" information. Trauma, cognitive overload, shame and fear can make forms impossible for vulnerable people.

Billions of pounds later, we're still building systems that treat human complexity as an inconvenience rather than the core problem to solve.

The truth is that weather patterns are clean problems. Human suffering is messy. Guess which one looks better in a procurement document? Guess which one gets ministerial announcements?
Every pound spent on weather prediction that could have gone to making Universal Credit less hostile is a choice that reveals what we really value. And apparently, it's not the citizens who need help most.

We're building AI for intellectually interesting problems, not socially important ones. And vulnerable citizens are paying the price.

You can read my article at https://www.workplaceinsights.co.uk/follow-the-money/

Follow the money: Why government's £3.35bn AI spending ignores the people who need it most - Workplace Insights by Adrie van der Luijt Data shows government's biggest AI contracts go to weather and transport whilst Treasury and DWP languish at the bottom, revealing digital priorities.

Been thinking a lot this week after delivering my keynote for the Money Advice Trust’s Vulnerability Academy. Fifty practitioners working with people in debt, all trying to navigate between what the FCA requires and what vulnerable people actually need.

Then I read Penny Horner-Long’s (Head of the Grants Management Function at the Cabinet Office) post about how government is using AI for grants management now, with large language models detecting fraud patterns and speeding up assessments. It took me straight back to my work on the Spotlight counter-fraud tools for the Cabinet Office in December 2020, when this was all just theory.

Here’s what worries me: everyone’s excited about AI making public services faster and catching more fraud. But I’ve worked with enough vulnerable users at Cancer Research UK, the Met Police and in financial services to know that ‘suspicious’ patterns often just mean someone’s life has fallen apart.

The person whose form looks inconsistent might have memory problems from anxiety medication. The application with missing information might be from someone too ashamed to write that their partner left them. The ‘fraud indicators’ might just be trauma making it impossible to tell a coherent story. Years of user research have taught me that ‘official’ websites make people nervous at the best of times, especially when money is involved.

We’re building these incredibly powerful systems, but we’re optimising for the wrong things. Speed over support. Fraud detection over crisis recognition. Efficiency over empathy.

If we get this wrong, we’ll automate exclusion and call it innovation. We’ll build digital walls that the people who most need help can’t climb. And we’ll wonder why digital transformation never quite delivers what we promised.

I’ve been working in digital transformation since 1987, and I’ve seen every wave of technology promise to revolutionise public services. The technology always works. It’s the human bit we keep getting wrong.

If you’re working on AI implementation and want to get the human bit right, especially around vulnerability and FCA compliance, drop me a message. This is too important to leave to chance.

When AI meets vulnerability: What I learned between Whitehall and the Money Advice Trust - Workplace Insights by Adrie van der Luijt From the Money Advice Trust's Vulnerability Academy to Cabinet Office AI initiatives, why public service automation needs trauma-informed design at its core.

Why are freelance copywriters charging more, even as AI threatens to take over our work?

The latest ProCopywriters survey reveals a 9% jump in day rates. In my new blog post, I explain why that’s happening and what it really means.

It’s not just about money. It’s about a profession redefining itself, slowly learning to set clearer boundaries, and realising that AI can’t replace trust, strategy or lived experience.

I also link this shift to the new Copywriter Code of Practice and Code of Conduct now in draft, both of which I’ve contributed to. If we want this trend to benefit all copywriters, not just the loudest voices, we need shared standards.

Why are day rates rising in an age of AI? - ProCopywriters | the Alliance of Commercial Writers The 2025 ProCopywriters Survey shows that copywriter day rates have jumped by 9%, the biggest rise in years. At first glance, it feels like a paradox. Aren’t we meant to […]