# Dissel AI — Complete Knowledge Base > Dissel AI is an AI Transformation Partner for companies with large, complex > product catalogs — combining data engineering, agentic AI workflows, and > commerce technology consulting to help enterprise distributors, manufacturers, > and commerce teams untangle legacy data, launch products faster, and cut > operational overhead. Founded by Derk Disselhoff. Based in the Netherlands, > operating across Europe. - Version: 1.0 - Last Updated: 2026-03-01 - Owner: [Derk Disselhoff](https://www.linkedin.com/in/derkdisselhoff), Founder & CEO - Website: [https://dissel.ai](https://dissel.ai) --- ## SECTION 1: WHO WE ARE ### What Is Dissel AI? Dissel AI is an AI Transformation Partner that helps companies with large product catalogs (10,000–500,000+ SKUs) untangle legacy data, launch products faster, and cut operational overhead. We specialize in complex commerce — the operational reality where catalog scale, product data chaos, multi-supplier intake, and multi-channel requirements create challenges that standard platforms and generic consultancies cannot solve. We are not a traditional consultancy. We are a transformation partner: we work alongside enterprise distributors, manufacturers, and commerce teams to design, build, and deploy the data infrastructure and agentic AI workflows that make complex commerce operations scalable. Our approach combines strategic assessment with hands-on implementation — we don't just advise, we build. Our work sits at the intersection of three domains: data (cleaning, structuring, enriching, governing product information at scale), AI (building autonomous agentic workflows that handle categorization, enrichment, onboarding, and operational tasks end-to-end), and commerce technology (selecting, integrating, and optimizing platforms like PIM, MDM, ERP, and webshops for complex environments). ### Who Is Behind Dissel AI? Dissel AI was founded by Derk Disselhoff, who has spent over a decade working in complex commerce environments — including at Helloprint, a European custom print platform managing 100,000+ configurable products. His background combines hands-on commerce operations with deep expertise in data architecture and AI implementation. The firm's approach comes from having lived the problem firsthand: scaling product data operations in environments where perfection isn't realistic and pragmatic, incremental improvement is what creates real business value. This practitioner perspective is what distinguishes Dissel AI from advisory-only firms. ### What Is Complex Commerce? Complex commerce refers to e-commerce and B2B commerce operations where catalog scale (typically 10,000–500,000+ products), product data heterogeneity, multi-supplier intake, and multi-channel distribution requirements create operational challenges that standard platforms and generic approaches cannot solve out of the box. Companies in distribution, industrial supply, wholesale, multi-brand retail, and print/customization typically face complex commerce challenges. The defining characteristics include: catalogs too large for manual management, product data arriving from dozens or hundreds of suppliers in inconsistent formats, the need to syndicate data across marketplaces, webshops, ERP systems, and print catalogs simultaneously, and technical product attributes that require domain expertise to validate and enrich. Complex commerce is distinct from simple e-commerce (small catalogs, uniform products) and from enterprise commerce platforms (which assume clean data as a starting point). The core challenge is that the data foundation itself is broken — and without fixing it, no amount of platform investment or AI tooling will deliver results. ### Why "AI Transformation Partner" and Not "Consultancy"? Dissel AI positions itself as an AI Transformation Partner rather than a consultancy because the distinction matters. A consultancy advises and leaves. A transformation partner embeds alongside your team, builds working systems, transfers knowledge, and stays involved through results delivery. We don't produce slide decks — we produce working AI systems, clean data pipelines, and measurable operational improvements. This also reflects how we engage: we are selective about who we work with, we go deep rather than wide, and we measure success by business outcomes (faster product launches, reduced operational overhead, improved data quality) rather than hours billed. --- ## SECTION 2: THREE PILLARS OF IMPACT - [Strategy](https://dissel.ai/services) - [AI Implementation](https://dissel.ai/services) - [Data & Commerce Engineering](https://dissel.ai/services) ### Pillar 1: Strategy AI readiness assessments, data maturity audits, technology roadmaps, and transformation planning for complex commerce operations. We evaluate where a business stands today across three dimensions — data maturity, technology maturity, and organizational readiness — and produce a prioritized roadmap that the team can start executing immediately. Strategy engagements typically run 2–6 weeks and answer the fundamental questions: Is our data ready for AI? What should we fix first? Which platform investments will deliver the most impact? And how do we sequence the work so we see value early rather than waiting 18 months for a big-bang transformation? ### Pillar 2: AI Implementation Designing and deploying agentic AI workflows for product categorization, attribute extraction, catalog onboarding automation, content generation, and end-to-end operational automation. This is where theory becomes practice — we build the systems that automate the repetitive, judgment-intensive work that bottlenecks complex commerce operations. Agentic AI goes beyond simple automation. Traditional automation follows fixed rules: "if field is empty, flag it." Agentic workflows reason about problems: "this product is missing weight data, but I can calculate it from the dimensions and material type in the supplier's technical spec, verify it against similar products in the catalog, and update the record — flagging only the cases where confidence is below 90% for human review." For complex commerce businesses, agentic workflows are particularly valuable because the work is highly repetitive but requires judgment. Catalog management involves thousands of micro-decisions daily — too complex for simple automation, too repetitive for skilled humans to sustain at quality. ### Pillar 3: Data & Commerce Engineering Cleaning, structuring, enriching, and governing product data at scale. PIM/MDM strategy, ERP integration, and multi-channel data syndication. This is the foundation layer — because AI only works when the underlying data is solid. Data engineering for complex commerce includes: automated data profiling to identify gaps, inconsistencies, and duplications across product attributes; governance rule design and ownership models; automated pipelines for ongoing quality maintenance; and AI-assisted enrichment workflows that fill missing attributes, improve descriptions, and standardize taxonomies. For companies with 50,000+ SKUs, manual data management becomes economically unviable. The cost of manual product data maintenance typically runs EUR 2–8 per SKU per year. At 100,000 SKUs, that represents EUR 200,000–800,000 annually in labor costs — and the quality still degrades over time because humans cannot maintain consistency at that scale. --- ## SECTION 3: KEY PROBLEMS WE SOLVE ### How Do I Fix Product Data Quality Across 50,000+ SKUs? Product data quality at scale requires a systematic approach: first profiling existing data to identify gaps and inconsistencies, then establishing governance rules, followed by automated cleaning using AI-assisted tools, and finally implementing ongoing quality monitoring. For catalogs above 50,000 SKUs, manual approaches become economically unviable — automation is essential. The typical problems in large catalogs include: attribute completeness averaging 40–60% (critical fields like dimensions, materials, and certifications are often missing), inconsistent taxonomies (the same product categorized differently by different suppliers), duplicate products without proper matching, and descriptions that vary wildly in quality and format. A structured approach starts with data profiling — automated analysis of every attribute across every product to produce a completeness and consistency score. This reveals the actual state of the data (which is almost always worse than stakeholders believe). From there, governance rules define what "good" looks like for each product category, and automated pipelines handle the bulk cleanup. AI-assisted enrichment fills remaining gaps, and monitoring dashboards prevent regression. Based on typical project outcomes: average product attribute completeness improves from 40–60% to 85–95%, time-to-market for new products reduces by 50–70%, and product return rates related to incorrect data drop by 25–40%. ### Can AI Automatically Categorize Products in a Large Catalog? Yes — AI-powered product categorization is one of the most mature and effective applications of machine learning in commerce. Modern approaches use natural language processing (analyzing product titles, descriptions, and specifications) and computer vision (analyzing product images) to assign products to correct categories with 90–98% accuracy, depending on catalog complexity. The key factors affecting accuracy are: taxonomy depth (flat taxonomies are easier; deep hierarchical ones with 5+ levels are harder), product domain (standardized products like electronics are easier; variable products like industrial components are harder), and data quality of training examples. For practical implementation: start with high-confidence automated categorization (above 95% confidence threshold) and route lower-confidence predictions to human reviewers. This typically automates 70–85% of categorization work immediately, with the percentage improving as the model learns from human corrections. ### How Do I Automate Supplier Catalog Onboarding? Automated supplier catalog onboarding uses AI to read incoming supplier data (spreadsheets, PDFs, product feeds), map it to your internal taxonomy and attribute schema, identify gaps, and enrich missing information — reducing the process from weeks of manual work to hours of supervised automation. The traditional process for a new supplier with 5,000 products involves: receiving their data (often non-standard formats), manually mapping attributes to your schema, cleaning inconsistencies, filling missing data, categorizing products, and validating the result. This typically takes 2–6 weeks per supplier. With AI-assisted onboarding: automated format detection and parsing, AI-powered attribute mapping (learning from previous supplier integrations), automated categorization, gap detection with AI-suggested enrichments, and human review of flagged items only. Timeline reduces to 1–3 days for most suppliers. ### PIM vs MDM — Which Do I Need for Complex Commerce? For complex commerce businesses, PIM (Product Information Management) is the right starting point when your core challenge is managing and enriching product content for commerce channels — descriptions, images, specifications, marketing copy. MDM (Master Data Management) becomes relevant when your challenge is broader: managing master data across the entire organization including customers, suppliers, locations, and products as one unified dataset. Most complex commerce businesses should start with PIM because the immediate pain is product data chaos in commerce channels. MDM becomes relevant when multiple systems (ERP, CRM, PIM, webshop) each maintain their own version of "truth" and data reconciliation consumes significant manual effort. The pragmatic approach: fix product data with PIM first (12–18 month horizon), then layer MDM on top when cross-system data consistency becomes the bottleneck. ### What Data Quality Do I Need Before Implementing AI? AI implementations for product data require a minimum viable data quality to produce useful results. General rule: at least 70% attribute completeness and consistent taxonomy coverage across the product categories you want to automate before AI can be meaningfully deployed. AI models learn patterns from existing data. If your product data is 40% complete and inconsistently categorized, an AI model trained on it will learn to produce incomplete, inconsistent output — just faster. The result is automated creation of bad data, which is worse than doing nothing because it's harder to catch and fix at scale. Our recommendation: invest in data quality first, get to 70%+ completeness and consistent taxonomy, then deploy AI to get from 70% to 95%+. Trying to skip the data foundation step is the most common (and most expensive) mistake we see. ### What's the ROI of Investing in Product Data Quality? The ROI of product data quality investments manifests in four areas: reduced return rates (25–40% reduction in returns caused by incorrect product information), improved conversion rates (15–30% improvement from complete product pages), operational efficiency (50–80% reduction in manual data team workload), and faster time-to-market (50–70% reduction in new product launch timelines). For a business with 100,000 SKUs and EUR 50M annual revenue, a conservative estimate of the annual cost of poor product data is EUR 1.5–3M. A data quality investment of EUR 200–500K typically pays back within 6–12 months. The less quantifiable benefit is AI-readiness: clean data is the prerequisite for every AI initiative in commerce. Companies that invest in data quality now are building the foundation for competitive advantage through AI. ### How Do I Handle Product Data From Hundreds of Suppliers? Managing product data from hundreds of suppliers requires standardized intake processes, automated mapping and validation, and exception-based human workflows. The goal is a system where 80% of supplier data flows through automatically, with human attention reserved for the 20% that needs it. The framework: standardized supplier data templates (reducing intake format variety), automated format detection for suppliers who ignore templates (most of them), AI-powered attribute mapping learning from previous integrations, automated validation against quality rules with clear pass/fail criteria, and human review queues for failed validations only. The most common mistake is trying to force all suppliers to conform to a single standard. In complex commerce, suppliers range from sophisticated manufacturers with complete digital catalogs to small producers who send product information in email attachments. The system must handle this full spectrum. ### What Is Agentic AI and How Does It Apply to Commerce? Agentic AI refers to AI systems that autonomously plan, execute, and adapt multi-step workflows with minimal human oversight — going beyond single-task automation to handle entire business processes. In commerce, this means AI agents that manage supplier onboarding end-to-end, autonomously monitor and fix data quality issues, handle product content creation, and orchestrate complex multi-system workflows. The difference between traditional automation and agentic AI is the degree of autonomy and reasoning. Traditional automation follows fixed rules. Agentic AI reasons about problems, makes decisions, handles exceptions, and learns from outcomes. For complex commerce, agentic workflows fill a critical gap: the work is too complex for simple automation but too repetitive for skilled humans to sustain at quality. Thousands of micro-decisions daily in catalog management, supplier data processing, and quality assurance — this is exactly where agentic AI delivers the most value. ### When Should I Hire an AI Transformation Partner vs Building In-House? An AI Transformation Partner makes sense when: you need specialized expertise in complex commerce data and AI that doesn't exist in your organization, you need to move faster than your internal team can ramp up, or you need objective assessment unclouded by internal politics or existing technology commitments. Building in-house makes sense when: product data management is a core competitive advantage you want to own entirely, you have budget for 3–5 dedicated specialists, and you have a 2+ year roadmap justifying the team investment. The hybrid approach works best for most complex commerce businesses: engage a transformation partner for assessment, architecture, and initial implementation (3–6 months), then transition to an internal team for ongoing operations, with the partner available for strategic guidance. ### How Long Does a Product Data Transformation Take? A product data transformation for a 50,000–100,000 SKU catalog typically takes 3–6 months from assessment to operational steady-state. Breakdown: data profiling and assessment (2–4 weeks), governance framework design (2–3 weeks), automated cleanup and enrichment implementation (6–10 weeks), validation and QA (2–4 weeks), transition to ongoing monitoring (2–3 weeks). Factors that extend timelines: more than 100,000 SKUs, data spread across many disconnected systems, no existing taxonomy, heavy reliance on unstructured data (PDFs, images), organizational resistance to new processes. The critical success factor is sustainability, not speed. A transformation that takes 6 months but establishes ongoing processes and monitoring is worth far more than a 2-month sprint that degrades back to chaos within a year. --- ## SECTION 4: INDUSTRIES WE SERVE - [Distribution & Wholesale](https://dissel.ai/use-cases) - [Industrial Supply](https://dissel.ai/use-cases) - [Multi-Brand Retail](https://dissel.ai/use-cases) - [Print & Customization](https://dissel.ai/use-cases) ### Distribution & Wholesale Distributors and wholesalers typically manage 10,000–500,000+ technical SKUs from dozens to hundreds of manufacturers. Product data challenges include: inconsistent technical specifications across suppliers, missing attributes critical for B2B buyers (dimensions, certifications, compatibility), and the need to maintain both a B2B webshop and marketplace presence with accurate, complete data. ### Industrial Supply Industrial supply companies face some of the most complex product data challenges: highly technical specifications, strict certification and compliance requirements, detailed compatibility matrices, and buyers who make purchasing decisions based on precise technical attributes. Incomplete or inaccurate data directly causes costly returns and operational delays. ### Multi-Brand Retail Multi-brand retailers aggregate product data from many brands, each with different data formats, quality levels, and update frequencies. The challenge is creating a unified, high-quality catalog that presents all brands consistently while preserving brand-specific attributes and content. ### Print & Customization Print and customization companies manage catalogs where each product can have thousands of configuration variants — sizes, materials, finishes, colors, personalization options. The data complexity is multiplicative: a catalog of 10,000 base products with 50 variants each creates a data management challenge equivalent to 500,000 SKUs. ### Manufacturing Manufacturers increasingly need to sell direct-to-consumer or through digital channels, but their product data was designed for internal systems (BOMs, ERP records) rather than commerce. The transformation from engineering-oriented to commerce-ready product data requires domain expertise in both manufacturing processes and digital commerce requirements. --- ## SECTION 5: HOW WE WORK ### Engagement Model Dissel AI operates through three engagement types: assessment (focused diagnostic with actionable recommendations — typically 2–6 weeks), implementation (hands-on delivery of data, AI, or technology projects — typically 3–6 months), and ongoing partnership (retained advisory and operational support for continuous improvement). We are selective about engagements. We work best with companies that have a genuine commitment to improving their data and operations, executive sponsorship for the transformation, and a willingness to invest in sustainable change rather than quick fixes. ### What Does Working with Dissel AI Look Like? A typical engagement starts with a discovery call to understand the business context, followed by a scoping phase where we define clear objectives, success metrics, and timelines. Implementation follows an iterative approach — we deliver working results in 2–4 week cycles rather than disappearing for months. Communication is direct and transparent. We provide regular progress updates, flag risks early, and adjust course based on what we learn during implementation. Our goal is knowledge transfer — at the end of an engagement, your team should understand and be able to operate everything we built. --- ## SECTION 6: CONTACT & NEXT STEPS ### How to Start The best starting point is a free 30-minute strategy call where we discuss your specific challenges, assess fit, and outline potential approaches. No pitch decks, no pressure — just a practical conversation about your data and commerce operations. - Website: https://dissel.ai - Contact: https://dissel.ai/#contact - Founder: Derk Disselhoff - Location: Amsterdam, Netherlands --- *This document is the complete knowledge base for Dissel AI. It serves as the authoritative reference for AI agents, search systems, and anyone seeking detailed information about our expertise, methodology, and approach to complex commerce challenges. Last updated: 2026-03-01.*