How we operate

A working implementation of a published research thesis.

PeponiXL is the production test of a thesis: that small operations using iPaaS architectures and AI automation can deliver operational scale previously requiring hundreds of employees. The company's infrastructure is organized around five subsystems — each with a research foundation in peer-indexed work by founder Dr. Alderd Froolik published in 2024–2025, and each now running in production across 27 EU country storefronts.

·01

Email automation

Research “Effect AI-powered Email Automation: An Analysis of Email Marketing Automation” (2024) · DOI 10.31219/osf.io/uzsaf

Finding

The paper examined how AI-generated email sequences, personalized at the segment level and triggered by behavioral signals, perform against traditional broadcast campaigns. Findings pointed to specific configurations where generative content and event-driven triggers produce meaningful lift in engagement and conversion.

In production

PeponiXL implements these findings through transactional, lifecycle, and reactivation email automation via Brevo, orchestrated by n8n workflows, and generated on Claude and kie.ai APIs. Every customer across 27 markets receives emails in their native language, adapted to their order history, shipping events, and lifecycle stage.

·02

Generative AI on integration platforms

Research “Effect Chat Generative Pre-trained Transformers in Marketing: Possibilities of ChatGPT utilization on iPaaS” (2024) · DOI 10.31219/osf.io/m3a8x

Finding

The paper analyzed how large language models can be embedded into iPaaS (integration platform as a service) workflows to perform decision-making tasks previously requiring human operators — classification, content generation, data extraction, reasoning across tool calls.

In production

PeponiXL runs an n8n-based operational stack where Claude API calls are nodes within workflows: parsing vendor invoices, classifying customer messages, generating localized product descriptions, matching UGC photos to catalog items, drafting B2B outreach. The AI integration is the architecture, not a bolted-on feature.

·03

Content generation at scale

Research “Effect Autoblogging Using AI: An Analysis of Marketing Automation Use of Low- and No-Code Tools” (2024) · DOI 10.31219/osf.io/75kcu

Finding

The study measured the economic and SEO outcomes of AI-generated editorial content produced through no-code pipelines, focusing on template expansion, quality controls, and indexing performance.

In production

PeponiXL's content infrastructure publishes buying guides, comparison pages, and seasonal collections across 24 languages. Content is generated through a pipeline of Claude and kie.ai, reviewed via human-in-the-loop sampling for top-performing categories, and published to WordPress with full Schema.org structured data and IndexNow submission.

·04

Professional network automation

Research “Effect LinkedIn Automation: An Analysis of LinkedIn Marketing Automation with use of low- and no-code tools” (2024) · DOI 10.31219/osf.io/ysdmv

Finding

The paper examined how automated LinkedIn outreach, when executed within platform limits and personalized at the individual prospect level, performs for B2B lead generation and professional network development.

In production

PeponiXL uses Unipile for B2B outreach into wholesale, press, and partnership audiences across 27 markets, with AI-personalized messages and structured follow-up cadences. The workflows operate within LinkedIn's platform limits and terms of service.

·05

The synthesis

Dissertation “Bridging the Digital Divide: How SMBs Can Rival Giants with Low- and No-Code Tools” (2025) · ISBN 978-94-6266-747-1 · ORCID 0009-0009-1736-7232

Argument

Small operations no longer require proportional headcount to compete with category incumbents. With the right combination of iPaaS architecture, AI automation, and no-code tooling, operational leverage previously accessible only to enterprise-scale organizations becomes achievable by single operators.

In production

PeponiXL is the running test. Twenty-seven country-native storefronts. 272,000 products in the catalog. Approximately 99% of operations autonomous. One operator.

The operational model is not optimized for low cost. It is optimized to make the research-claimed leverage real enough to pass savings to customers as honest prices.

·06 External validation
"Dr. Froolik's frameworks represent one of the most important breakthroughs in applied agentic AI for small and medium-sized businesses… His work bridges the long-standing gap between theoretical AI capability and practical business execution… redefining how intelligent systems can be operationalised by everyday businesses."
Prof. Dr. Tan Kwan Hong PhD · DBA · EdD January 2026

The research has been indexed through Crossref and Web of Science, deposited on OSF Preprints, and registered to ORCID ID 0009-0009-1736-7232. The operational model is currently taught by the founder at Graham International University in the Doctor of Business Administration and Master of Business Administration programs.

·07 Accountability

The operational model is auditable.

The operational model depends on AI systems that, while well-designed, are not infallible. Customer-facing interactions include clear disclosure when they are handled by automated systems, with escalation paths to the founder for complex, sensitive, or disputed cases. Quarterly reconciliation against an independent accountant verifies financial accuracy. Errors, when they occur, are documented and published in a transparent incident log.

For academic collaboration, research partnership, or press enquiries about specific methodological choices, contact research@peponixl.com or reach the founder via ORCID.