- type
- concept
- created
- Tue Apr 07 2026 02:00:00 GMT+0200 (Central European Summer Time)
- updated
- Tue Apr 07 2026 02:00:00 GMT+0200 (Central European Summer Time)
- sources
- raw/articles/SERVICE-AS-SOFTWARE-PLAN
- tags
- strategy pivot ai-broker commission revenue-model
Service-as-Software Pivot
The Shift
The original B2BPaper concept was a traditional marketplace: mills list their surplus paper, buyers browse and buy. The "Service as Software" pivot reframes the platform as an AI broker that does the work a human broker does, but faster and cheaper.
Old Model (Marketplace)
- Mill lists surplus inventory on the platform
- Buyer searches, filters, browses listings
- Buyer contacts mill or requests a quote
- Human negotiation, manual paperwork
- No commission or subscription-only revenue
New Model (AI Broker)
- Buyer describes what they need in natural language
- AI agents search supply database + live-scrape external marketplaces
- Agents present 2-3 best matches with pricing
- Agent handles negotiation, paperwork, and deal closing
- B2BPaper takes 2-4% commission per closed deal
Revenue Economics
| Metric | Value |
|---|---|
| Commission rate | 2-4% per deal |
| Human broker rate (competitor) | 3-8% per deal |
| Target volume | 50-100 deals/month |
| Average deal size | EUR 100k |
| Monthly revenue target | EUR 150k-400k |
| Ramp: Month 3 | EUR 50k/month |
| Ramp: Month 6 | EUR 150k/month |
The commission undercuts human brokers while providing faster turnaround. At scale, the AI broker handles 80% of deals autonomously, with Thierry personally managing the remaining 20% of high-value (EUR 50k+) deals requiring human trust.
Why This Works for Paper
Paper brokerage is relationship-heavy but information-simple. A deal is defined by a small set of parameters: paper type, GSM, width, volume, quality grade, origin, and price. This makes it highly amenable to algorithmic matching. The complexity lies in:
- Trust -- mills need to know the buyer is real and creditworthy
- Logistics -- container filling, freight rates, incoterms
- Visibility -- mills controlling who sees their surplus to avoid channel conflict
- Speed -- surplus has short shelf life (30-90 days before obsolescence)
An AI broker can handle items 2-4 natively. Item 1 is addressed by wiki/concepts/kyb-upfront and Thierry's personal relationships.
The Five AI Agents
The plan calls for five specialized agents in Phase 7 of the SaaS build:
- Buyer Qualification Agent -- automatically qualifies inbound buyer requests
- Supply Matching Agent -- triggers matching when new supply or demand enters
- Outreach Drafting Agent -- writes personalized offer emails to qualified buyers
- Deal Assistant Agent -- provides negotiation support during active deals
- Daily Digest Agent -- generates Thierry's morning briefing with pipeline status
Thierry's Hybrid Role
Thierry is not replaced by AI; he is amplified by it. His role:
- Opens doors at mills using his personal network
- Validates pricing and grade legitimacy (industry expertise)
- Closes the 20% of deals requiring human trust (high-value, first-time buyers)
- Provides the "anchor of trust" that makes mills willing to engage with the platform
Relationship to What Was Built
The actual MVP (phases 0-13) built the marketplace foundation: models, APIs, matching algorithm, newsletters, container assembly, frontend. The SaaS/broker pivot described in this plan represents the next evolution. The scraper component was partially built in /home/claude/customers/b2bpaper-scraper/ (24/56 stories) before being paused.
Sources
- raw/articles/SERVICE-AS-SOFTWARE-PLAN -- full strategic plan
Related
- wiki/summaries/saas-plan-summary -- complete summary of the plan document
- wiki/concepts/make-to-order-marketplace -- evolution from surplus-only to production listings
- wiki/concepts/spec-based-matching -- the matching engine core to the AI broker
- wiki/entities/b2bpaper -- the platform entity
- wiki/concepts/kyb-upfront -- trust mechanism enabling automated deals