Full stack in six weeks, one senior + AI.

The brief

A Southern California B2B disposable food packaging distributor came to us with the full stack to build. They needed a B2B Shopify Plus storefront with proper landed-cost pricing, a Business Central tenant set up for inventory and finance, Google Ads to drive qualified buyers, Power Automate workflows for replenishment, and AI-driven reporting they could actually act on.

The catch: ocean-freight lead times of 30–45 days meant their inventory math had to be tight on day one. Mid-engagement, a 200%+ tariff spike hit one of their major sourcing countries — pricing couldn't be set-and-forget. And the deadline was six weeks.

The team (or lack of one)

The traditional consultancy approach would have been: bring in a 4–5 person team across BC implementation, Shopify development, ad operations, and supply chain analysis. Timeline: 4–6 months. Cost: a six-figure budget (industry benchmark for comparable scope). Risk: handoff drag, communication overhead, junior staffing on critical decisions.

Our approach: one senior consultant — 25 years in ERP and ecommerce — working alongside Claude.

AI handles velocity-multiplying tasks: generating AL code stubs, drafting Power Automate flow logic, computing ROP and Safety Stock from item ledger data, writing technical documentation in markdown. The human stays on architecture, judgment, and client conversation. Net result: six weeks instead of six months, with zero information lost in handoffs.

What we built

Shopify Plus storefront with a multi-layer B2B pricing engine

Pricing is the hardest part of B2B ecommerce. The client's customers buy in volume, expect tiered pricing, and are price-sensitive in a tariff-volatile market. We built a multi-layer pricing model that combined:

1. Embedded landed cost — FOB + freight + duty + tariff, computed per SKU per country of origin
2. Compare-at pricing anchored above the margin floor — the strikethrough number customers see
3. Tiered order-value discounts — automatic price breaks at order volume thresholds
4. First-order welcome discount — single-use code for new B2B customers

The whole stack has a hard margin floor — regardless of how promos stack, the client never sells below cost-plus. When the tariff spike hit one of their major sourcing countries mid-engagement, the model absorbed it without anyone needing to touch a spreadsheet.

Microsoft Business Central with AL extensions

The native Shopify Connector covers about 70% of what a B2B Shopify ↔ BC integration actually needs. The other 30% is where deals quietly break. We built custom AL extensions for the edge cases that mattered:

• Payment capture triggered from BC fulfillments (not just on order)
• Posted shipments writing back to Shopify for fulfillment status
• B2B Company Location mapping via Tax ID / Registration Number
• Dimension validation handling so customer ledger entries actually apply
• ACH / eCheck / credit-card surcharge logic for B2B payment terms

Google Ads, four campaigns, weekly reporting

Search, Display, Performance Max, and competitor-overlap campaigns — launched in spring 2026, with weekly performance reports through Week 5 and beyond. We managed Display impression waste (mobile app placements were a meaningful share of early budget leakage), CPC volatility, and competitive bidding overlap with a major incumbent in their category.

Power Automate Premium for inventory replenishment

This is where the AI part of "1 + AI" really shows up. We built a Power Automate Premium architecture that runs nightly across BC's Item Ledger Entries. The flow:

1. Pull current stock + open POs + recent demand velocity per SKU
2. Compute true ROP and Safety Stock factoring in 30–45 day ocean freight lead time
3. Generate a PO recommendation report — which SKUs to reorder, in what quantities, from which vendor
4. AI writes a natural-language summary explaining each recommendation
5. The report lands in a Teams channel as an Adaptive Card for approval
6. On approval, the PO is created in BC automatically

For one of their hero SKUs, the model produced a ROP in the low thousands, safety stock in the hundreds, and a reorder quantity sized for a single ocean container — all computed from real Item Ledger Entry data, not gut-feel.

Custom BC reports with AI analysis

Standard BC reports tell you what happened. We built reports that also tell you what to do about it. ABC/XYZ classification ran weekly. Dead-stock detection flagged items not moving for 90+ days with cost recovery suggestions. Replenishment forecasts projected stockout dates per SKU at current velocity. Every report ended with a 3-bullet AI summary aimed at the procurement manager, not the analyst.

The "1 + AI" angle, in detail

What used to require a four-person team is now one senior + Claude. The division of labor:

• AL development — AI generates code stubs and unit test scaffolds; senior reviews, refines, deploys
• Power Automate flows — AI drafts the action JSON; senior verifies business logic and edge cases
• Inventory analytics — AI runs the ABC/XYZ classifications and dead-stock pulls; senior interprets and recommends actions
• Documentation — AI generates markdown docs alongside every commit; senior reviews and signs off
• Client communication — senior handles 100%, no AI substitutes; this is where judgment compounds

The pattern: AI handles velocity-multipliers, senior handles judgment-multipliers. Result is the delivery speed of a 4-person team at the cost of one senior, with no key man risk — everything is documented and lives in the client's own Microsoft tenant.

What this means for you

This client isn't a unicorn. They're a 10–30 user SMB distributor — the exact profile we built this stack to serve. If your business looks similar — Shopify B2B catalog with imported goods, Business Central or considering it, manual replenishment that's getting painful, ad ops that aren't scaling — the same pattern works.

Six weeks. One senior + AI. Everything in your tenant. No 4-person team to wrangle.