Lawyers are quietly becoming builders
Ernie Svenson · email · 2026-07-10 source ↗ Reader 🖥️
AI Labor DisruptionCitizen DeveloperAgencyNon-Engineers BuildingTechnology Diffusion & Transformation CurvesVendor Research Conflict of InterestAI Workflow Patterns
AI has redefined 'building' from writing software to iteratively correcting and reusing AI workflows for repeatable tasks, so any lawyer — not just a technical few — can and should become a builder without learning to code.
practical · mixed (expository + argumentative + hortatory), built around a relayed narrative — A short newsletter riffing on a LawNext podcast interview (Bob Ambrogi interviewing Anthropic's Mark Pike). It second-hand relays Pike's claims, reframes what 'building' means, and converts that into a start-this-week instruction for a solo/small-firm-lawyer audience — closing with a paid-workshop upsell.
Structure
| Part | Role |
|---|---|
| Lede + Mark Pike origin story (Facebook/Slack background; 'as surprised as anyone' pushing his own code to GitHub; built tools for Anthropic's own legal team first, then productized them as Claude for Legal) | Hook and existence proof — the leader of a legal-AI product got there by building for himself first, not by setting out to build a product. |
| 'Pockets of builders exist everywhere' — the old way meant convincing IT or an engineer an idea was worth the trouble; now anyone with an idea can make progress themselves (rapid prototyping); a solo lawyer can test ideas in their own practice with no budget request and no outside developer | Establishes the enabling shift: AI removes the IT/engineer gatekeeper. |
| 'Why lawyers, of all people' — lawyers are the deepest, most-returning users of Anthropic's tools; Pike's 'law is code'; big messy document sets (deposition transcripts, statutes) are AI's strength; the document work lawyers dread is what AI is best at | Evidence-and-fit argument for why the reader's profession is especially well-suited. |
| 'You don't have to become a coder' — not every lawyer must build; a certain type will, everyone else must get good at the tools; can't ignore it. The reframe: 'building' = take a repeatable task, set up a faster way, correct it when wrong, reuse it. Pike's plugins: feed in your playbook/preferences, fix its mistakes, it remembers | The core conceptual move — redefines 'building' accessibly as iterative correction rather than software authoring. |
| 'What to do this week' — pick one repeatable task, build a simple reusable 'recipe' in whatever AI tool you already use, correct it a few times; 'that's the whole muscle. Start small.' | Concrete, low-stakes call to action. |
| P.S. — invitation to the Inner Circle's AI Lab / ChatGPT workshop | Commercial call-to-action. |
Key terms
- builder / building
Not writing software. Taking a repeatable task, setting up a faster reusable way to do it, correcting it when it is wrong, and reusing it — an iterative-refinement practice available to any professional. - law is code
Pike's phrase: legal material is large, messy, structured-enough document sets (transcripts, statutes) that current LLMs process well — law's substrate is document-shaped and machine-tractable. - rapid prototyping
An individual testing an idea inside their own practice with AI — cheaply, immediately, with no budget and no developer — and keeping the ones that work. - recipe / the muscle
A saved, reusable, corrected AI workflow for a repeatable task; 'the muscle' is the practiced habit of building such recipes.
Propositions
- explicitcorroborated'Building' in the AI era means turning a repeatable task into a corrected, reusable AI workflow, not writing software.conf: highsupportsFragments: February 18 · p13 — This doc's reframe — that building means iterating a reusable workflow, not writing code, so lawyers can build — is a concrete profession-level instance of AI lowering the barrier to software development and democratizing hard-to-acquire skills.supports ←How Anthropic Teams Use Claude Code · p2 — The report's Legal, Growth Marketing, and Finance examples are concrete instances of the broader claim that 'building' in the AI era means turning a repeatable task into a corrected, reusable AI workflow rather than writing software.
- explicitAI lets a domain expert prototype a solution inside their own practice without a budget request or an outside developer.conf: highsupports8) Christian Printzell Halvorsen om å lede på riktig nivå: Lokalt forankret, nordisk samlet · p20 — Lawyers testing ideas in their own practice with no budget and no outside developer is the same phenomenon as non-engineers doing rapid prototyping without consuming engineering-team capacity.
- explicitThe old barrier to acting on a software idea was convincing IT or an engineer it was worth the trouble; AI removes that gatekeeper so anyone with an idea can make progress themselves.conf: high
- explicitLawyers are the deepest and most-returning users among knowledge workers on Anthropic's tools.conf: medium
- explicitLegal work is well-suited to current AI because 'law is code' — it consists of large, messy document sets (transcripts, statutes) that these tools handle well.conf: medium
- explicitThe document work lawyers dread — reading, comparing, and summarizing piles of documents — is precisely what AI is currently best at.conf: high
- explicitNot every lawyer needs to become a builder, but every lawyer must at least become proficient at using the tools; the profession cannot ignore AI.conf: highsupports8) Christian Printzell Halvorsen om å lede på riktig nivå: Lokalt forankret, nordisk samlet · p19 — The claim that every lawyer must at least get good at the tools and cannot ignore AI supports the broader thesis that AI literacy must reach non-engineering functions.
- explicitdevelopedA reusable AI recipe is built by feeding in your playbook and preferences and correcting the tool's mistakes so it remembers — an iterative-correction skill, not a coding skill.conf: highsupportsAssumptions as code: SiriusXM's approach to platform prioritization · p7 — Pike's plugins — feed in your playbook/preferences and correct until it remembers — are exactly 'encoding expert reasoning as reusable skills' that operationalizes and democratizes institutional (here, a lawyer's) knowledge.extends ←How Anthropic Teams Use Claude Code · p6 — The report's Claude.md-maintenance and end-of-session-summary patterns, observed across many Anthropic teams, are a cross-team instance of the 'reusable AI recipe' pattern -- feeding in preferences and correcting mistakes so the tool remembers -- that the lawyers piece names as the decisive, non-coding skill of AI-era knowledge work.
- explicitAdoption should start small: pick one repeatable task and build a simple reusable recipe in whatever AI tool you already use.conf: high
- implicitThe decisive competence for AI-era knowledge work is the discipline of iterative correction and reuse of a workflow, not technical or coding ability.conf: high
- implicitA profession's rate of AI adoption tracks how closely its core work maps onto AI's current strengths — document-heavy professions adopt fastest.conf: medium
Arguments
Any lawyer can and should become a builder — practicing the turning of repeatable tasks into corrected, reusable AI recipes. strength: medium
Explicit premises:
- Building no longer requires writing software; it means iterating a reusable workflow (p1, p8).
- AI removes the IT/engineer gatekeeper, so an individual can prototype in their own practice with no budget or developer (p2, p3).
- Lawyers' document-heavy work is exactly what AI is currently best at, and lawyers are already its deepest users (p4, p5, p6).
- The leader of Claude for Legal reached his position by building internal tools for himself first — an existence proof.
Implicit premises:
- The 'lawyers are the deepest users' evidence, though it comes from the vendor's own product lead, is reliable and generalizable.
- Iterative correction of an AI on legal tasks reaches a quality bar adequate for real legal work — 'close enough' is safe enough.
- 'Law is code' — legal work reduces to large-document processing, which is AI's sweet spot.
- One well-resourced practitioner's success (Pike, inside Anthropic) generalizes to solo and small-firm lawyers with far fewer resources.
Inferences:
- The same fit that makes lawyers eager adopters also makes their billable document work the most exposed to automation and commoditization — adoption and disruption are two sides of one coin, but the author frames only the upside.
- Someone must still own, verify, and be accountable for a lawyer-built recipe's errors — the operational and liability question is never raised.
- If a 'certain type' builds while everyone else only uses tools, a new competence hierarchy forms inside the profession.
Lawyers, of all knowledge workers, are especially well-positioned to benefit from current AI. strength: medium
Explicit premises:
- Lawyers are the deepest and most-returning users of Anthropic's tools (p4).
- 'Law is code' — big, messy document sets are what these tools handle best (p5).
- The document work lawyers dread is exactly what AI is best at right now (p6).
Implicit premises:
- Depth of usage is a proxy for genuine benefit, not merely novelty or experimentation.
- Document processing is the core of legal value, rather than judgment, interpretation, and advocacy.
Inferences:
- High exposure of document work cuts both ways: the more AI can do it, the less of it lawyers can bill for.
Narratives
- Mark Pike's origin story: he came up through Facebook and Slack, was 'as surprised as anyone' when he started pushing his own code into GitHub, built tools for Anthropic's own legal team, saw how much they helped, and only then turned them into Claude for Legal.
→ Existence proof and role model — legitimizes internal-first, almost-accidental building and the claim that you don't set out to build a product, you build for yourself first. Implicitly argues that even a lawyer/AGC, not a career engineer, can build. - The LawNext interview itself: Bob Ambrogi, 'the veteran legal tech journalist,' interviews Pike, 'associate general counsel at Anthropic and the product lead for Claude for Legal.'
→ Credentialing device — a respected journalist plus Anthropic's own AGC lend authority to the newsletter's claims; also flags that the specifics are relayed second-hand from a podcast the author heard. - Pike's plugins: you feed in your playbook and your preferences, and when the tool gets something wrong, you fix it and it remembers.
→ Concretizes the reframe of 'building' as iterative correction and models the recipe practice the reader is told to adopt. - 'Pockets of builders exist everywhere' — Pike's description of latent builders he meets among customers who always had solutions they wanted but were blocked by the old gatekeeping.
→ Normalizes the idea that ordinary professionals are already becoming builders and positions the reader as one of these latent pockets.
Questions addressed
- ✅ What does 'becoming a builder' actually require of a lawyer who is not a programmer?
Iterative correction and reuse of an AI 'recipe' for a repeatable task — feeding in a playbook, fixing mistakes so the tool remembers — rather than writing code. - ✅ Why are lawyers, specifically, adopting AI so heavily?
Per Pike, their document-heavy work ('law is code') maps onto AI's current strengths, so lawyers are the deepest users; but the supporting evidence is a vendor-sourced usage claim the author does not scrutinize. - ✅ What concretely should a lawyer do to start?
This week, pick one repeatable task, build a simple reusable recipe in an AI tool you already use, and correct it a few times. Start small. - ⭕ Who verifies and is accountable for AI-produced legal work, and what quality bar counts as 'close enough'?
Not addressed. 'Correct it a few times, see how close you can get' is offered for legal work product with no treatment of malpractice/liability, error ownership, or the quality threshold for real practice. - ⭕ Does heavy adoption of document-AI also expose lawyers' own billable work to disruption?
Not raised. The piece frames document-heavy exposure only as an adoption opportunity, never as a commoditization risk to the same work.
Gaps
- incompleteThe central 'why lawyers' evidence — that lawyers are the deepest, most-returning users among knowledge workers — comes solely from Anthropic's own product lead (Pike), relayed second-hand in a newsletter that closes with a paid-workshop upsell. No independent usage data or cross-profession comparison is offered, so the load-bearing claim is an unverified vendor assertion.
Read next: Independent, third-party AI-usage studies across professions; legal-AI adoption surveys not published by a model vendor; measured outcomes rather than depth-of-use anecdotes. - incompleteVerification and accountability of AI-produced legal work are elided. 'Correct it a few times, see how close you can get' is prescribed for legal work product, but the piece never treats what 'close enough' means when errors carry malpractice or liability, who owns the mistake, or how a solo builder without an engineer maintains and QA's the recipe over time.
Read next: The operational-ownership concern in the FOSE 'they build it, you run it' discussion; malpractice/duty-of-care standards for AI-assisted legal work; human-in-the-loop verification patterns for high-stakes domains.closed byWhat's an Orchestrator—and Why Does Software Need One? — That gap asks what mechanism establishes accountability and 'close enough' for AI-produced professional work when errors carry malpractice/liability exposure. O'Brien answers with a concrete (if unproven) institutional proposal from the software side: multi-year apprenticeship modeled on medicine/law, plus eventual Professional-Engineer-style licensing and rising malpractice-analogous insurance requirements as productivity (and therefore responsibility) scales — directly the 'malpractice/duty-of-care standards' the gap's reading direction points toward, generalized from law to software orchestration. - incompleteThe 'law is code' analogy is accepted uncritically. It reduces legal work to large-document pattern-matching and never examines where legal reasoning — interpretation, advocacy, novel argument, judgment under adversarial conditions — diverges from what document-summarization AI actually does, which is exactly where the analogy would break.
Read next: Critiques of legal formalism and 'law is code'; the gap between document processing/retrieval and genuine legal reasoning; LLM limitations on reasoning versus summarization.
Closes gaps elsewhere
- No, AI Did Not Kill Agile · g4 — Provides a concrete profession-level instance of exactly the collaboration mode that gap says its source left unexplored: non-technical stakeholders (lawyers) prototyping directly with AI, bypassing the 'convince IT or an engineer' gatekeeping and collapsing the idea-to-implementation gap. It illustrates the phenomenon rather than fully resolving the open question of whether this transcends the management-vs-engineering power dynamic.
Related documents
- Assumptions as code: SiriusXM's approach to platform prioritization
- 8) Christian Printzell Halvorsen om å lede på riktig nivå: Lokalt forankret, nordisk samlet
- Fragments: February 18
- Why did Martin Fowler invite 60 engineers to a ski resort?
- Why cultivating agency matters more than cultivating skills in the AI era | Max Schoening (Head of Product, Notion)
- The "funhouse mirror": How AI reflects the hidden truths of your software pipeline
- Presentation: A Few Predicted Talks From QConAI 2030
- Arvind Narayanan on Why AI Isn't All That Revolutionary
- Travel Through the Lens of AI with Booking.com CEO Glenn Fogel
- Én enkelt behandling kan gi svært langvarig kolesterolsenkning: – Noe radikalt nytt