What AI integration
looks like in practice.

Contract review was consuming disproportionate lawyer time on repetitive checklist tasks — identifying missing clauses, flagging unusual terms, cross-referencing standard positions. The firm needed this volume handled without sacrificing accuracy or increasing headcount.

Muljin integrated Claude via API into the firm's existing document management system. Contracts uploaded for review are automatically processed through a custom prompt pipeline, producing a structured issues report with clause-level annotations. Lawyers review the flagged items rather than the full document from scratch.

First-pass contract review time dropped from an average of 2.5 hours to under 35 minutes. Lawyers report higher confidence in their review thoroughness. The firm has since expanded the integration to cover NDA review and employment contract standardisation.

Content briefs were being manually interpreted differently by different writers, producing inconsistent output that required multiple rounds of revision. The agency had no systematic way to maintain brand voice, SEO requirements and client style guides simultaneously across a growing client base.

Muljin engineered a multi-stage content pipeline in which client briefs trigger an automated AI drafting process using custom system prompts encoding each client's brand voice, tone guidelines and SEO requirements. The draft is delivered directly into the agency's project management tool for human review and finalisation.

Writers shifted from producing 3–4 drafts per week to reviewing, refining and publishing 12–15. The agency onboarded four new clients within three months of deployment — capacity that would previously have required hiring three additional writers. Content revision cycles reduced by 70%.

Client reports required pulling data from multiple systems, writing plain-English portfolio commentary, applying brand formatting and conducting a compliance review — all within a tight monthly deadline. The process was entirely manual and had produced regulatory compliance issues on two prior occasions.

Muljin built a data pipeline that aggregates client portfolio data from the firm's systems and feeds it into a structured Azure OpenAI prompt that generates personalised commentary. Copilot was integrated for formatting and template population. A compliance-check stage flags any outputs requiring human review before dispatch.

The 40+ monthly analyst hours were reduced to under 8 — primarily spent on exception handling and sign-off. Report quality improved markedly: clients noted increased clarity and consistency. The firm has since extended the pipeline to quarterly strategic reviews and onboarding documentation.

Documentation existed but was scattered, outdated and hard to navigate. Engineers spent 2–3 hours per week fielding questions that should have been self-service. The onboarding process relied heavily on senior team members, creating bottlenecks and slowing new hire productivity.

Muljin built a retrieval-augmented generation (RAG) system over the company's Notion documentation, GitHub codebase and internal wikis. An AI assistant was deployed into Slack, allowing any team member to query the company's knowledge base in natural language. Answers are sourced from verified documents and include citations.

Internal knowledge queries via Slack dropped by 65% within 30 days. New hire ramp time fell from an average of 7 weeks to 3. Senior engineers reported reclaiming meaningful focus time. The company has since extended the system to cover customer-facing support documentation.