CHAPTERS
A “legal lamp” demo: using Claude Code to make a non-technical idea real
Mark Pike opens with a playful project: turning a desk lamp into a Morse-code messenger using Claude Code. The point is to show how a non-engineer can still build functional tools by describing what they want in plain language.
- •Built a desk “legal lamp” that blinks Morse code on command
- •Uses Claude Code as the bridge from idea to working implementation
- •Emphasizes he’s non-technical and not a programmer
- •Demonstrates “thinking” behavior and output (e.g., blinking “HI”)
Why legal teams need automation: repetitive work, errors, and wasted focus
He explains the day-to-day pain: legal work often involves repetitive tasks that are dull and error-prone. Claude helps reduce busy work so lawyers can spend time on higher-value judgment calls.
- •Repetitive legal tasks are tedious and easy to mess up
- •Busy work tends to get deferred because it consumes time without using core skills
- •Claude enables natural-language workflow building
- •Goal: redirect effort toward “best work” rather than admin tasks
The marketing crunch problem: last-minute launch reviews
Mark describes a common scenario where marketing requests legal review right before a product launch. The need is speed plus consistency—catching issues quickly without missing key risk areas.
- •Marketing often sends blog posts for review with very little lead time
- •Legal needs a reliable first-pass method to flag potential issues
- •Claude can be instructed with what legal cares about
- •Motivation: faster turnaround without sacrificing diligence
Creating a self-review workflow: Claude turns requirements into a tool
He asks Claude to build a workflow and is surprised by how fully it “runs with it.” This becomes a repeatable self-review tool marketing can use before escalating to legal.
- •Asked Claude to build a workflow based on legal review priorities
- •Claude set up an end-to-end process rather than a one-off answer
- •Outcome: a marketing material self-review tool linked/pinned for easy access
- •Shifts some review effort upstream, before legal is paged last-minute
Using the tool: paste content and run ‘Analyze Content’
Mark walks through the actual process: a marketer opens the tool, pastes a blog post, and triggers analysis. Claude then reviews the material and returns structured results.
- •Open the self-review link from the channel/pinned resource
- •Paste full marketing content into the tool
- •Click ‘Analyze Content’ to start Claude’s review
- •Produces a consolidated set of review results
Structured findings: issue categories, what looks good, and what needs review
Claude returns a list of issues and positive notes, helping the author understand both risks and strengths. The output is organized into specific legal concern areas that drive next actions.
- •Identifies issues to address (example: five issues)
- •Flags themes like accuracy and security-related claims
- •Checks for publicity rights/third-party content permissions
- •Notes partnership considerations
- •Also highlights what’s working well in the draft
Hand-off to legal: generating a Slack summary and filing a ticket
The workflow doesn’t stop at analysis—Claude prepares a concise Slack message and helps route the work into the legal ticketing system. This reduces back-and-forth and ensures legal gets a clear, prioritized request.
- •Generate a Slack message summarizing key issues for the legal team
- •Copy to clipboard for quick posting
- •Link out to file a legal ticket with context included
- •Creates a cleaner intake process for faster review
Prioritization and risk signaling: low/medium/high based on a legal framework
Mark explains that the tool highlights the most important issues and assigns risk levels using a framework he provided. Claude effectively acts as a first-pass “eyes and ears,” enabling smarter prioritization.
- •Surfaces and prioritizes the most important concerns
- •Provides low/medium/high risk signals
- •Risk scoring is based on a framework defined by legal
- •Acts as a first-pass reviewer rather than final decision-maker
Human-in-the-loop guardrails: speed gains without trusting blindly
He underscores responsible use: AI can hallucinate, so legal maintains oversight. Claude accelerates review and flags issues early, but humans remain accountable for final judgments.
- •AI systems can hallucinate; outputs must be checked
- •Legal retains human review and responsibility
- •Claude helps move faster while maintaining controls
- •Preemptive flagging reduces surprise risks near launch
Other legal use cases inside Anthropic: redlining, conflicts, and outside activities
Mark broadens the lens to show Claude’s role across the legal department. Beyond marketing review, Claude supports commercial contract work and compliance-style workflows.
- •Used for redlining exercises in commercial matters
- •Supports conflict of interest policy workflows
- •Helps review outside business activity requests
- •Demonstrates broad applicability across legal operations
Getting started advice: begin with the most routine work and experiment
He closes with a practical recommendation: start small by applying Claude to routine tasks. The key is to try—capabilities become clear through hands-on experimentation.
- •Identify the most routine/repetitive task first
- •Open Claude and try describing the task in plain language
- •Iterate toward a workflow that fits your needs
- •You learn what’s possible by experimenting
