design ops & localization leader · copenhagen, dk

    Building operational systems that help organizations scale.

    One example: Localization from a heavy and messy cost center to automated infrastructure.

    Six+ years owning end-to-end localization for an 18-language SaaS product at Planday (from Xero) — and two years turning it into an AI-driven pipeline that ships verified translations in minutes.

    impact.summary18 languages · verified in production
    impact.cost_reduction−80%cumulative translation cost vs. agency-only baseline¹
    impact.turnaround4 days → 10 mintranslation delivery, standard product keys
    impact.volume+30%word volume absorbed with no added headcount
    impact.automation100%product keys delivered automatically, gated publishing

    portfolio.journey

    From manual bottleneck to no-touch pipeline

    One system, rebuilt in deliberate stages — each stage validated before the next. Dates below are the actual rollout milestones.

    2026 →

    No-touch, org-wide

    As Sr. Design Ops Manager: fully automated translation pipeline with gated publishing, org-wide AI adoption across design & research, and metrics linking UX work to bottom-line outcomes.

    Sept 2025

    Measure the promise

    "Cheaper, faster, better" audited one year on: 57% direct cost savings at that checkpoint, automated staging deliveries every 24h, verified production uploads weekly, translation quality scoring in place.

    Jan 2025

    AI-exclusive

    All in-product languages switched to AI translation. Commercial and support teams onboarded to the same tooling — ending five years of parallel spreadsheet workflows across the org.

    Oct – Nov 2024

    Democratize the workflow

    Removed the content-design bottleneck: every product designer got a Lokalise seat and the Figma plugin. Key creation moved into the design stage, owned by each product trio. Gated releases via staging/production branches.

    Sept 2024

    Go live, low stakes first

    AI translations live for 11 lower-risk languages; agency retained for core markets and monthly human review of AI output. Deliberately phased — high-stakes languages held back until quality data justified the move.

    May 2024

    Test before believing

    Started controlled trials of AI translation + AI LQA in Lokalise. Ran a head-to-head quality benchmark: AI output vs. professional linguists, same content, same scoring rubric.

    2019 – 2023

    Own the foundation

    End-to-end in-product localization for 18 languages (~65,000 words/year). Agency-based weekly cycles; every request routed through two content designers.

    portfolio.cases

    Workflow case studies

    The AI localization pipeline, supporting a 45,000-word website relaunch, and establishing no-touch translation delivery (work in progress).

    open case studies →

    portfolio.field_notes

    What two years of AI ops taught me

    Principles I've presented publicly, earned the hard way — through bloopers as well as wins.

    01

    AI is very smart, but has no common sense. Design the system around that: it needs context, constraints and guardrails, not trust.

    02

    You get what you give. Bad glossaries, stale translation memory and unclear source text produce bad translations — no model fixes upstream neglect.

    03

    Speculate and test. Rinse, repeat. Every rollout stage was a hypothesis with a measurement plan, not a leap of faith.

    04

    Build a safe way to fail. Branches, tags and rollback paths are what make speed responsible. You can only move fast on live products if reversal is cheap.

    05

    Treat the root cause, not the symptom. Fixing one translation is a task; fixing the glossary term behind twenty bad translations is ops.

    06

    Know the point of diminishing returns. The goal is not perfect translations — it's the right quality at the right cost, per language, per audience.