Skip to content

    AI & automation

    AI e-commerce automation specialist

    Pipelines that take over repetitive marketplace work, with a human on the sensitive decisions

    I'm Hans van Leeuwen, a freelance e-commerce manager who has automated most of his own marketplace operation. At ABS All Brake Systems I automated the VIN-based parts lookup, a first in the industry, and connected eBay, Amazon and Bol.com into a single operation; at Alpine Hearing Protection I automated marketplace content through Channable and automated demand forecasting brought out-of-stock below 2%. The stack is n8n for orchestration, Supabase as the data layer, Claude for content and analysis and Channable for feeds. What I deliver is not 'an AI that runs your store' but concrete, monitored workflows where every success message can be proven.

    Hans van Leeuwen

    Hans van Leeuwen

    Freelance e-commerce & marketplace manager, Amersfoort (NL)

    Last updated: About Hans and his track record · LinkedIn profile

    Who automation pays off for now

    • Teams that build the same exports, checks and reports by hand every week
    • Brands with large catalogues where product data is maintained manually per channel
    • Multi-marketplace operations where stock and pricing drift out of sync
    • Companies that want AI for content but refuse to publish errors
    • Founders who already have agents and dashboards and aren't sure they tell the truth
    • Teams that want a competitor and price radar per channel without staffing it

    What I automate

    • Product data: feed normalisation, enrichment and nightly validation per channel (Channable)
    • Orders and stock: synchronisation between Amazon, Bol.com, eBay and the webstore, with alerts on failure
    • Content pipelines: AI-generated titles and descriptions with source and confidence labels and human approval
    • Advertising: automated ROAS/ACOS reporting, budget alerts and negative keyword lists
    • Competitor and price radar per marketplace, weekly or daily
    • KPI reporting to Slack or email, with evidence (URL, database row, status) per message
    • Meta-monitoring: dead man's switches so the safety net itself is watched
    • Governance: who may do what, where the human decides, and what your audit trail looks like

    Results from my own operations

    1st

    automated VIN-based parts lookup in the industry

    ABS All Brake Systems · 2026

    3 → 1

    eBay, Amazon and Bol.com connected into one marketplace operation

    ABS All Brake Systems · 2026

    <2%

    out-of-stock through automated demand forecasting

    Alpine Hearing Protection · 2021–2022

    Experience at

    • ABS All Brake Systems

      E-commerce manager · 2025–present

    • Alpine Hearing Protection

      Marketplace manager · 2021–2025

    • Intergamma (Karwei & Gamma)

      E-commerce manager · 2017–2020

    • IGM Badkamerwinkel

      E-commerce manager · 2019–2020

    • Webhelp (GGD GHOR)

      Team coach · 2020–2022

    • Talpa · Edelman

      Internships: online marketing · UX design · 2013–2015

    References from clients and managers available on request. Recommendations on LinkedIn

    From the field

    Automation I built myself, including the mistakes I made along the way:

    My AI assistant reported success. Half of it was invented.

    An audit of my own stack (spring 2026) showed that roughly half of the success messages could not be traced to a real action. A 300-second timeout was logged as 'success'. Three of the four causes were my own build errors; this is what I rebuilt.

    Read the audit (Dutch)

    Everything was green. Except the watchman.

    The safety net that should report failing crons had itself been broken for two weeks: 258 errors, zero alerts, 34 days without a backup. On meta-monitoring, dead man's switches and why an UNKNOWN in a dashboard is usually an untested assumption.

    Read about meta-monitoring (Dutch)

    Codex is no longer a copywriter. It's an operator.

    Agentic AI completes full competitor research from a single prompt in nine minutes. What that means for your Bol.com and Amazon operation, and why every agent owes you proof.

    Read about agentic AI (Dutch)

    How an automation engagement runs

    An engagement starts with a free automation audit: I map the manual workload and deliver, within 48 hours, a prioritised plan with the three workflows to automate first, calculated on hours per week and cost of errors.

    Every workflow is built on your infrastructure and accounts, with a human in the loop on price limits, brand voice and publishing. Every success message must trace back to an artefact: a URL, a database row, an HTTP status. That is the lesson from my own audit and it's in every workflow I deliver.

    After delivery the operation stays yours: documentation, access, monitoring and a handover your own team can maintain. Where needed I stay available on retainer for extensions and maintenance.

    Rates and ways of working

    No agency margin, no six-month minimum, cancellable monthly. Three models, with indicative rates (excl. VAT):

    Day rate (interim) · €760 per day
    About €95 per hour. For temporary leadership, 2 to 5 days a week, on-site in the Utrecht region or remote across NL/EU.
    Project fee · from €2,500
    For a scoped engagement with a fixed outcome: an audit with plan, a Bol.com launch, an Amazon DE roll-out or a feed automation.
    Retainer · from €1,750 per month
    Ongoing Amazon and Bol.com account management: two fixed days a month, weekly trading review and monthly reporting.

    After a 30-minute intro call you receive a written quote within one working day: scope, cadence and final rate. For context: a permanent e-commerce manager, including employer costs, quickly runs to €6,000 or more per month, without the option to stop monthly. An interim engagement of three days a week usually stays below that. All rates and what is included →

    Frequently asked questions about AI automation in e-commerce

    What is AI e-commerce automation?

    A set of pipelines for product data, operations, advertising and reporting that take over repetitive marketplace work and speed up decisions, with a human in the loop on strategy, pricing and brand voice. Concrete, monitored workflows on n8n, Supabase and Claude; not 'an AI that runs your store'.

    Which platforms and tools do you automate?

    Amazon (NL, DE and other EU marketplaces), Bol.com, eBay and webstore platforms such as Magento and WooCommerce, plus the layer in between: Channable for feeds, n8n for orchestration, Supabase as the data layer and Claude for content and analysis.

    Do I need to replace my team?

    No. Automation takes over the repetitive 80% so your team focuses on strategy, assortment and brand. The human stays in the loop on price limits, brand voice and final decisions.

    How do I know the automation tells the truth?

    Because every message ships with evidence. My own audit showed roughly half of success messages were untraceable; since then I build every workflow with an artefact per step and a dead man's switch on the safety net itself.

    How does an engagement start?

    With a free automation audit. I map your current manual workload and deliver a prioritised plan within 48 hours with the three highest-ROI workflows.

    Are my data and account access safe?

    Pipelines run on your own infrastructure and accounts; marketplace credentials are never shared in plain text. The automation layer is auditable and you keep full control of access.

    What should you automate first?

    Request the free automation audit. Within 48 hours you have a plan with the three highest-ROI workflows.

    Request the audit

    Amazon NL specialist · Bol.com consultant · Interim e-commerce manager · Case studies