AI may make legal organisations faster while quietly making them more similar.
The institution legal board members actually govern is the in-house team, the panel of firms, the company secretariat, compliance, and the board pack factory that feeds directors. When those people consult the same models, the organisation does not just get faster. It gets more alike.
In metallurgy, the risk is not one bad heat. It is running every line on the same recipe. You get interchangeable parts. You also get a shared failure mode. When thousands of lawyers consult the same frontier models, with the same “be concise and balanced” system prompts, you are running one alloy across the industry. Uniformity is not strength. It is correlated risk.
Optimizers Converge. Legal Markets Do Too.
Given similar inputs, optimization processes converge. That is not an AI-specific mystery; it is what “most likely” means at organisational scale. If the inputs are (a) the same public law, (b) the same vendor playbooks, (c) the same model family, and (d) the same management request for a two-page board paper by Thursday, the outputs will occupy a tight cluster.
A 2026 UI/UX study is a clean analogue for legal operations. Generative tools cut task time on the order of 40–50% for both novices and experts, and scored lower on creativity and innovation than the traditional workflow. Efficiency up; originality down—for people who already knew what they were doing, not only for beginners (IJIST).
Legal research and first drafting are the same shape of task: structured, time-measured, quality-rated by fluency. The study’s warning for a GC is precise. Your experts get faster and less distinctive at the same time. The junior gap closes on speed. It does not close on judgment. You may have paid for a flatter, faster, more average legal function.
“Workslop” Is a Legal-Operations Incident Waiting for a Name
Harvard Business Review, with BetterUp Labs and Stanford Social Media Lab, named a workplace pattern that legal departments will recognise immediately: workslop—AI-generated content that looks like good work but lacks the substance to advance the task. It shifts the burden downstream. Recipients spend on the order of two hours dealing with each instance; in their survey, about 40% of employees had received workslop in the previous month (HBR).
In a legal organisation, workslop is not a jokey LinkedIn carousel. It is:
- A “research memo” that restates the question in polished headings and never grapples with the messy fact.
- A board paper that is formatted, sourced, and empty of a decision.
- A mark-up that looks negotiated and has not engaged the other side’s actual incentive.
- A compliance assessment that maps every article and recommends “enhance documentation.”
- An outside-counsel email that summarises the client’s own AI draft back to them.
The productivity theft is double. The author saved an hour. The partner, the GC, or the legal director on the board spends two hours reconstructing the problem. Trust falls. HBR’s survey work also found that recipients mark down colleagues’ competence. In a profession that runs on reputational capital, that is not a soft cost.
Workslop is homogenization you can feel in the inbox. Everything is fluent. Nothing moves the matter.
What Correlated Legal Advice Does to a Board
Legal members of boards should picture three concentric failures.
Inside the company. If GC, privacy, compliance, and the business all prompt the same model, “three lines of defence” produce one line of prose. Challenge is cosmetic.
Across the panel. If every firm on the panel uses the same tools against the same “market standard” corpus, you are not buying independent advice. You are buying a second printout. Panel diversity was supposed to be a control. Model concentration undoes it.
Across the industry. Supervisors already see templated responses. When every institution’s AI-assisted DORA or AI-Act narrative converges, the supervisor learns nothing, and the first distinctive failure looks like an outlier instead of a forecast. Competitive legal strategy—the structure nobody else used—becomes rarer for the same reason breakthrough product ideas become rarer in the Sloan studies.
Thomson Reuters’ warning is the right one for boards that think this is “just a tooling issue”: deploying AI for efficiency alone leaves the creative upside on the table, and it turbocharges whatever template you already reward (Thomson Reuters Institute). If your legal function already rewarded safe, complete, non-surprising papers, the model will give you that at industrial scale.
Board-Level Risks That Do Not Show Up as Hallucinations
You already know the hallucination-and-sanctions story. Homogenization adds risks that look like success:
| Risk | What you see | What you miss |
|---|---|---|
| Strategy collapse | Fast consensus on “market approach” | The non-market structure that fitted the facts |
| False assurance | Consistent papers across functions | Correlated error, not independent confirmation |
| Training failure | Juniors shipping fluent work | No one who can try a case or a negotiation without a model |
| Downstream rework | Impressive drafts | Workslop tax on seniors and the board |
| Supervisory sameness | On-time, on-template filings | No distinctive risk story when you actually need one |
| Privilege and discovery | Convenient drafting in tools | A record of prompts that is not the legal analysis you thought you had |
The legal member’s job is to insist that consistency is not corroboration. Two models agreeing, or two teams prompting the same model, is not two opinions.
What to Ask Management and the Panel This Quarter
- Which legal artefacts are AI-assisted, by matter type, and who signs the thesis?
- How do we detect workslop? (Unowned recommendations, no discarded alternatives, generic issue lists, seniors redoing the analysis.)
- What is our model concentration? One vendor, one prompt library, one “house style” injected into every system prompt is a single point of intellectual failure.
- Do panel firms start from our draft? If yes, we have reduced independent counsel to copy-editing.
- Where did we last take a legal position that was not the industry mean—and can we prove it was considered on purpose?
I advise boards and GCs on legal-function AI operating models: concentration risk, panel independence, and quality bars that stop workslop before it reaches the pack. Contact me.
Relevant Sources
- Comparative Study of Generative AI and Traditional Tools for Creativity and Efficiency in UI/UX Design — International Journal of Innovations in Science & Technology (2026) — https://journal.50sea.com/index.php/IJIST/article/view/1876
- AI-Generated “Workslop” Is Destroying Productivity — Harvard Business Review (Sept 2025) — https://hbr.org/2025/09/ai-generated-workslop-is-destroying-productivity
- Organizations Are Misdiagnosing What’s Killing Their Innovation — Thomson Reuters Institute — https://www.thomsonreuters.com/en/institute/articles/feature-misdiagnosing-whats-killing-innovation
- The Hidden Cost of AI-Assisted Creativity — MIT Sloan Management Review — https://sloanreview.mit.edu/article/the-hidden-cost-of-ai-assisted-creativity/
- AI Legal Risk: Summary and Board Takeaways — Goldmanmalka — https://goldmanmalka.com/ai-legal-risk-summary/