Ramona Furter
See the agent
Application · Marcura · AI Agent Manager, Operations

I build AI agents that take the manual work off operations.

I lead AI business models at Swiss Post and build my own products end to end with agent workflows. I'm hands-on with Claude every day, the CLI, tool use and prompt engineering. This very site was built with Claude Code. This is my application for AI Agent Manager, Operations at Marcura.

See an agent I'd deploy Read the CV
AI leadAI business models, Swiss Post
Claude Code · agentsbuilt with daily, hands-on
+9% / +15%lift, via testing & analytics
Ops & process20+ yrs in regulated businesses
Portrait of Ramona Furter
Why I fit

What this role needs, and where I've done it

This role is about finding the operational busywork worth automating, building and deploying the agent end to end, getting the team to actually use it, and proving it saved real hours. That's the work I do now. Here's how it lines up with what you're after.

01

Spotting the automation that pays

Use-case discovery is the part I'm best at: looking across a workflow, finding where manual effort and error cost actually sit, and ranking by value rather than novelty. I've done it as a venture builder and a product lead, picking the smallest thing that proves the win.

Proof: found and prioritised the e-commerce work that lifted a CHF 100M+ business; ran pilots from MVP to launch at Die Mobiliar.

02

Building and shipping the agent end to end

I design multi-step agents and automations and take them from prototype to running, not just a demo. Document extraction, data validation, routing and exception handling are exactly the shapes I build with LLM tool use and workflow tools today.

Proof: built Pedal Peak end to end with AI agent workflows; build automations with Claude tool use and n8n.

03

The Claude toolchain, hands-on

The posting asks for Claude for Enterprise and Teams, Claude Code CLI, Cowork-style desktop automation, multi-step agent and tool-use patterns, and advanced prompt engineering. This is my daily practice, not a line on a course certificate. I built this site with Claude Code.

Proof: work in Claude Code CLI daily; design agent and tool-use workflows; comfortable with the Anthropic API and AI safety and governance.

04

Driving adoption, not just shipping

A great agent nobody uses is worthless. I drive bottom-up adoption: sit with the people doing the work, run the training and workshops, write the playbook, and make the new way easier than the old one. Change management is half the job and the half most automation misses.

Proof: drove adoption across cross-functional teams and agencies at ifolor; built and trained a go-to-market team as WePractice scaled to 23 people.

05

Measuring impact: accuracy, ROI, KPIs

I define success up front and let the numbers decide: adoption rate, accuracy, hours saved, ROI. I'm comfortable in analytics and SQL-level questions and I report honestly, the misses as well as the wins. No "it feels faster".

Proof: +9% conversion and +15% checkout lift through A/B testing and analytics on a CHF 100M+ business; owned budgets and KPIs.

06

Operations in regulated, real businesses

Much of my career has sat inside process-heavy, regulated organisations, insurance, financial-services partnerships, e-commerce operations, where data quality and clean handoffs decide whether anything works. I partner naturally with product, engineering and data to fix the messy parts.

Proof: owned UBS and Baloise partnerships at Brixel; ran innovation inside Die Mobiliar; led the e-commerce ecosystem at ifolor.

Curriculum vitae

Ramona Furter

AI and operations lead in Zurich with over twenty years of experience, open to relocating to Dubai. I turn ambiguous, manual processes into shipped automation and measured results, increasingly with AI agents at the core. German and Swiss German native, English fluent, French conversational.

Jan 2026 to present

AI Project Lead, Business Development

Swiss Post, Advertising · Zurich

  • Lead AI-driven business models for Swiss Post Advertising, from sizing the opportunity to building and running the roadmap.
  • Design AI agent and automation workflows, and bring them from prototype to live use with clear KPIs.
  • Run cross-functional work from concept to launch across product, tech, data and commercial teams.

Oct 2024 to Jul 2025

Senior Product Manager, Lead E-Commerce

Ifolor Group · Zurich

  • Owned the e-commerce ecosystem and operations for a CHF 100M+ business, reporting to C-level.
  • Lifted conversion 9% and the checkout step rate 15% through research, A/B testing and analytics.
  • Led a cross-functional team and external agencies, owning budget, resourcing and KPIs.

Jun 2023 to Sep 2024

Lead Project Manager

Brixel · Zurich

  • Owned the partnerships with financial institutions, UBS and Baloise, that drove growth.
  • Was the main bridge between senior client stakeholders and the internal product and delivery team.

Mar 2020 to May 2023

Marketing & Growth Lead, Founding Team

WePractice · Sparrow Ventures (Migros Group) · Zurich

  • Founding team of a mental-health venture. Closed two funding rounds and grew it to 10 locations, 23 people and 170+ customers.
  • Generated 1000+ client matches in year one and built the full go-to-market on a hypothesis-and-data approach.
  • Built, trained and led the marketing and sales team after Series B, owning budget, KPIs and growth.

Sep 2019 to Sep 2022

Growth & Venture Builder

Sparrow Ventures · Zurich

  • Built and ran growth and go-to-market for several internal startups, from early validation to scale-up.
  • Used research and experimentation to improve conversion, lower acquisition cost and raise customer lifetime value.

Jan 2017 to Aug 2019

Intrapreneur, Innovation

Die Mobiliar · Bern

  • Ran market pilots for new products (Smide, now BOND Mobility, plus XperCheck and Lizzy) from MVP to launch, inside one of Switzerland's largest insurers.
  • Coached cross-functional teams and explored new data and partnerships.
A worked example

An AI agent I'd deploy at Marcura

Not a slide of buzzwords. A real agent taken the way I'd actually run it: find the use case, design the multi-step workflow, decide how much it's allowed to do on its own, then roll it out and prove it saved hours. Click through the four steps, and try the agent pipeline.

Worked example · illustrative figures

Find the use case worth automating

Start where the manual effort and error cost actually sit, not where the AI is most fun. Score the candidates by value, volume and how cleanly an agent can do the work.

The pick

A document agent for port disbursement accounts: read each PDA and FDA, extract every line item, check it against the agreed tariff and history, and route only the exceptions to a human.

Candidate A · chosen
Disbursement account review

High volume, document-heavy, every line checked by hand today. Clear right answers to validate against, so accuracy is measurable and the hours saved are real.

Why first: biggest manual load, lowest ambiguity.
Candidate B
Vendor invoice matching

Real value, but more edge cases and messier source data. A strong second once the first agent proves the pattern and the guardrails.

Later: reuse the same extract-and-validate spine.
Candidate C
Inbox triage & routing

Tempting and visible, but lower error cost and harder to measure cleanly. Good for adoption buzz, weak as a first ROI proof.

Skip for v1: hard to prove the number.
From day one

My first 90 days

Roughly how I'd spend my first three months as AI Agent Manager at Marcura: learn the ground truth, pick the highest-value safe bet, and ship an agent that saves measurable hours.

Phase 1 Days 1 to 30

Learn the ground truth

  • Meet the operations teams, product, engineering and data, and the people doing the manual work today.
  • Map the document-heavy, repetitive, error-prone workflows and where the hours and mistakes really go.
  • Learn the Claude estate cold: what's deployed, the Enterprise and Teams setup, and the safety and governance rules.
Phase 2 Days 31 to 60

Pick the bet, design the agent

  • Choose one high-value, low-ambiguity workflow and design the agent, its steps and its guardrails.
  • Stand up the eval set and the accuracy bar before the agent touches a live account.
  • Align operations, engineering and governance on scope, autonomy level and the phased rollout.
Phase 3 Days 61 to 90

Ship, adopt and measure

  • Get the agent live in shadow then Assist mode, behind the right approvals.
  • Run the first training and workshops, write the playbook, and report hours saved and accuracy honestly.
  • Turn what we learned into the next iteration and the start of the reusable agent-pattern library.