YTS Consulting · AI Advisory & Automation
Confidential · Client Audit Report
Agentic AI Readiness Audit
Meridian Trade Solutions BV
A comprehensive multi-dimensional assessment of AI automation potential, Agentic AI agent specifications,
phased implementation roadmap, financial impact modelling, ROI projections, risk analysis,
and vendor technology recommendations for B2B Industrial Wholesale & Import/Export operations.
Distribution: Erik van den Berg (COO) · Sophie Janssen (CFO) · Pieter Loos (IT Manager) · Board of Directors
52
AI Readiness Score
/ 100 Points
"Emerging" Stage
01 · Executive Summary & Leadership Brief3
08 · ROI & Financial Impact Model14
02 · Company & Operations Profile4
09 · Phased Implementation Roadmap17
03 · AI Readiness Assessment (8 Pillars)6
10 · Risk Assessment & Mitigation19
04 · Operational Pain Point & Cost Audit8
11 · Data Governance & Quality Remediation21
05 · Agentic AI Use Case Catalog (7 Agents)9
12 · Change Management Strategy22
06 · Technology Stack & Vendor Analysis11
13 · Market Benchmarks & Competitive Context23
07 · Compliance, Security & EU AI Act13
14 · Strategic Recommendations & Next Steps24
Bottom Line Up Front: Meridian Trade Solutions BV is currently absorbing an estimated €1.0–€1.2 million per year in avoidable operational cost — driven primarily by 210 FTE hours of manual repetitive processing each week across order entry, accounts payable, export documentation, customs handling, and inventory management. A phased Agentic AI programme with a Year 1 investment of €150,000 is projected to deliver €217,000 in net savings in Year 1, growing to a cumulative €1.4 million by end of Year 3, representing a 3-year ROI of approximately 540%. Payback is projected at 12–16 months — well within the 18–24 month target stated by the CFO.
52/100
AI Readiness Score
"Emerging" — strong foundation
210 h/wk
Manual FTE Hours Lost
≈ 5.25 FTE equivalents absorbed
€1.18M
Total Addressable Cost
Annual recoverable via AI automation
€591K
Projected Year 2 Annual Saving
At 65% capture rate (conservative)
€14.50
Current AP Invoice Cost
vs. €2.78 best-in-class with AI
Level 2/5
Process Automation Maturity
7 of 10 processes still manual
12–16 mo
Projected Payback Period
Aligns with CFO 18–24mo target ✓
7 Agents
Priority AI Agents Identified
Across all operational domains
Key Strategic Findings
🔴 Critical Gaps (Act Now)
- 480 manual re-keying actions per day across 87 FTE — an extraordinary operational drag absorbing the equivalent of 5+ full-time employees in pure data-entry activity.
- 54% of 52 daily purchase orders are received by email and keyed manually into D365 BC — the single largest time sink at 90 h/week (€131,500/year).
- Customs & trade documents are at data quality 2/5 — critical data locked in paper-scanned PDFs with no structured extraction, creating compliance exposure for AEO-C status.
- ERP and CRM are completely siloed — sales team lacks real-time stock/order visibility, limiting cross-sell opportunity and forcing 4.2-hour customer query response times.
- €620,000/year in inventory losses (stock-out + overstock) attributable to Excel-based demand forecasting with ±12 day lead-time variance.
🟢 Strengths to Build Upon
- Microsoft Dynamics 365 BC (v22, cloud) with full REST API — the optimal foundation for AI agent integration without rip-and-replace.
- 3 years of clean transactional data (since 2023 ERP migration) — sufficient for demand forecasting ML model training on 2,400 SKUs.
- CEO as AI champion and €120K–€180K Y1 budget confirmed — governance and funding are the two most common failure modes for AI programmes; both are addressed here.
- Proven small-scale automation success — Power Automate PO acknowledgement flow saved 5h/week, demonstrating change acceptance.
- 9% CAGR revenue growth with 28% gross margin — financial capacity to absorb AI investment and scale operations without proportional headcount growth.
Leadership Alignment Summary
| Stakeholder | Primary Concern | What This Report Delivers |
|---|
| Erik van den Berg (COO) | Reducing manual processing time; change management risk; WMS compatibility | Phased roadmap with WMS-agnostic Phase 1; change management strategy in §12; 210h/week savings mapped to specific agents |
| Sophie Janssen (CFO) | Clear ROI; payback ≤24 months; avoid vendor lock-in | Detailed 3-year financial model with conservative/base/optimistic scenarios; 12–16mo payback; open-API architecture recommendation |
| Pieter Loos (IT Manager) | Limited IT bandwidth (2 FTE); integration complexity; EU data residency | Microsoft-native stack recommendation minimising custom dev; EU-hosted vendor shortlist; YTS Consulting advisory support reducing IT burden |
€18.4M
Annual Revenue (FY2025)
47 days
Days Sales Outstanding
€4,200
Average Order Value
84%
OTIF Rate (target: 92%)
Headcount by Department (87 FTE)
Import Source Countries (%)
Order Intake Channels (%)
Operations at a Glance
| Operational Metric | Current Value |
|---|
| Daily purchase orders received | 52 orders/day |
| Daily outbound shipments | 85 shipments/day |
| Avg. order processing time | 3.2 hours / order |
| Avg. order lines per PO | 7.4 lines |
| Order error rate | 4.1% (22 credit notes/month) |
| Customer query response time | 4.2 hours (business hours) |
| Customer churn rate | 9% per year |
| Quote-to-order conversion | 63% |
| Top 10 customer revenue share | 38% (moderate concentration) |
Warehouse Operations
| Warehouse Metric | Current Value |
|---|
| Total warehouse area | 8,400 m² (Rotterdam + Łódź) |
| Storage bin locations | 6,200 bin locations |
| Inventory accuracy rate | 91% (9% GRN-driven error) |
| Order picking accuracy | 97.2% |
| Avg. put-away time / pallet | 14 minutes |
| Avg. pick & pack time / line | 4.5 minutes |
| Monthly returns volume | 60–80 return lines |
| WMS system | D365 BC basic WMS module |
| Current automation | Barcode + conveyor; no AMR/RFID |
Financial Profile Summary
Revenue & Trade Flow Profile
Customer Complaint Categories (%)
Strategic Context: Meridian is operating a healthy, growing B2B wholesale business with 9% CAGR and €18.4M revenue. However, operational leverage is constrained by manual processes that grow linearly with volume. Every additional 10% revenue growth currently requires proportional headcount additions — precisely the scalability trap that Agentic AI is designed to break. The company's stated strategic driver — "scale revenue without proportional headcount growth" — is the textbook use case for AI-driven operations.
The YTS Consulting AI Readiness Framework evaluates organisational readiness across eight dimensions, each scored 0–100. An organisation scoring below 40 on any pillar faces a blocking risk to AI deployment. Meridian has two blocking risks (Process Maturity: 40, Integration Maturity: 35) which must be addressed in the programme foundation phase before production AI agents are deployed.
AI Readiness Radar — Meridian vs. Best-in-Class
Technology Stack
68
D365 BC REST API · M365 E3 · Power BI · Strong platform foundation
Data Quality
48
CRITICAL GAP: Customs docs 2/5 · Siloed unstructured data
Process Maturity
40
⚠ BLOCKING: Level 2/5 · 7 of 10 processes manual
Digital Skills
60
Finance strong (4/5) · Warehouse gap (2/5) · Mixed team
Leadership Buy-In
65
CEO champion · CFO ROI-gated · COO change-cautious
Integration Maturity
35
⚠ BLOCKING: ERP–CRM silo · CSV exports only · No middleware
Budget Readiness
72
€120K–€180K committed Y1 · Phased approach approved
Compliance & Security
60
GDPR · AEO-C · M365 E3 DLP · ISO 27001 in planning
Data Quality by Domain
Data Quality Scores by Domain (Current vs. Required for AI)
Current vs. Target Automation Level by Process
| Data Domain | Current Score | Min. Required for AI | Gap | Key Issue | Remediation Action |
|---|
| Financial & Accounting | 4/5 | 3/5 | ✓ Met | Inter-company recon. occasionally delayed | Automate inter-company posting rules in D365 BC |
| Sales & Order History | 4/5 | 4/5 | ✓ Met | Pre-migration data not fully cleansed | Archive pre-2023 data; flag in forecasting model |
| Inventory & Stock Levels | 4/5 | 3/5 | ✓ Met | 9% inaccuracy from GRN delays | Digitise GRN process (Phase 2) |
| Logistics & Shipping | 3/5 | 3/5 | △ Marginal | Carrier tracking data not systematically captured | Shippypro API → D365 BC structured capture |
| Customer Master | 3/5 | 4/5 | ✗ Gap | Duplicate accounts; outdated contacts | Data dedup sprint; HubSpot–D365 BC sync |
| Product & HS Codes | 3/5 | 4/5 | ✗ Gap | HS codes not systematically reviewed; inconsistent descriptions | AI HS Classification Agent (Phase 3); manual review sprint for top 500 SKUs |
| Supplier / Vendor Master | 3/5 | 3/5 | △ Marginal | Banking details and lead times not updated | Automated supplier data refresh workflow |
| Customs & Trade Documents | 2/5 | 4/5 | ✗ CRITICAL | Paper PDFs; no structured extraction; 30% arrive incomplete | AI Document Extraction Agent — Phase 1 priority |
Weekly FTE Hours Lost by Process
Annual Quantified Cost by Pain Point (€)
Detailed Pain Point Register
| Process / Pain Point | Current State | FTE hrs/wk |
Annual Labour Cost | Error/Penalty Cost | Total Annual Impact | Priority |
Email PO Order Entry
54% of 52 daily orders |
Manual re-key from email into D365 BC by CSR; 3.2h avg per order cycle; 2–3 FTE absorbed daily |
90 h/wk | €131,546 |
4.1% error → 22 credit notes/mo × €380 = €100,320/yr |
€231,866 |
P1 — Critical |
Export Document Preparation
~85 shipments/day, 6 markets |
2.8h/shipment; Word/Excel templates; copy-paste from ERP; country requirements in manual spreadsheet |
48 h/wk | €70,158 |
3–5 L/C discrepancies/yr; document errors → 27% of complaints |
€80,000+ |
P1 — Critical |
AP Invoice Processing
~304 invoices/month est. |
Manual 3-way match (PO/GRN/invoice) by AP clerk; €14.50/invoice vs. €2.78 benchmark; 37h/wk absorbed |
37 h/wk | €52,896 |
Late payment penalties; missed early payment discounts est. €8,000/yr |
€60,896 |
P1 — Critical |
Import Document Processing
~200 emails/day with PDFs |
Manual collection from supplier emails; re-keying into ERP; 30% arrive incomplete; 3.5h/shipment customs prep |
~30 h/wk | €43,848 |
8–12 customs errors/yr → €28,000 penalties; AEO-C risk |
€71,848 |
P1 — Critical |
Inventory Management (Stock-out + Overstock)
2,400 SKUs · ±12 day lead variance |
Buyer intuition + Excel rolling 3-month forecast; no statistical model; manual weekly stock review |
12 h/wk | €17,534 |
Stock-outs: €390K; Overstock carrying cost: €230K |
€637,534 |
P2 — High |
Customer Service & Order Queries
340 active accounts |
4.2h avg response; tracking by email; no portal; 9% annual churn rate |
~15 h/wk | €21,918 |
9% churn × avg customer value = est. €165,600 lost revenue/yr |
€187,518 |
P2 — High |
Supplier Performance Monitoring
143 suppliers · Basic spreadsheet |
Monthly spreadsheet compiled manually; no SLA scorecard; no automated supplier alerts |
8 h/wk | €11,683 |
Poor supplier OTD visibility; reactive rather than proactive management |
€15,000+ |
P3 — Medium |
Sanctions & Compliance Screening
118 export customers · 6 markets |
Manual spot-checks against EU/OFAC lists; not systematic; compliance officer time-constrained |
~4 h/wk | €5,842 |
Regulatory risk: potential AEO-C revocation; OFAC penalties |
HIGH RISK |
P2 — Compliance |
| TOTAL QUANTIFIED OPERATIONAL COST |
~210 h/wk |
€355,225 |
€831,920 |
€1,187,218 |
|
Critical Finding for CFO: The €1.187M total annual impact figure understates the true cost because it excludes opportunity cost — the strategic revenue that cannot be pursued while 5.25 FTE equivalents are absorbed in manual data entry. At Meridian's blended revenue-per-employee of €211,500 (€18.4M / 87 FTE), the opportunity cost of 5.25 trapped FTEs represents a notional €1.11M in unrealised revenue capacity.
What is an Agentic AI? Unlike traditional automation (rule-based, brittle, requires human triggers), an Agentic AI system uses a Large Language Model as its reasoning engine, combined with tool-calling capabilities (APIs, databases, document processors) to autonomously perceive inputs, plan a sequence of actions, execute those actions across multiple systems, and handle exceptions — all with configurable human-in-the-loop escalation thresholds. Agents do not just extract data; they act on it.
Priority 1 · Phase 1 · Quick Win
🤖 Agent 1: AI Order Intake & PO Processing Agent
This agent monitors the shared orders@meridian-ts.com inbox in real time. Upon receiving an email with or without PDF attachments, it classifies the email intent, extracts all purchase order fields (customer, SKUs, quantities, requested delivery dates, prices) using a multimodal LLM-powered extraction engine, validates the extracted data against D365 BC customer master, product catalogue, and live pricing, then creates a confirmed draft sales order in D365 BC — all within seconds. For unrecognised formats or ambiguous data, it routes to a human reviewer with pre-populated context, dramatically reducing resolution time from 6.5 hours to under 15 minutes.
Trigger
Inbound email to orders@ inbox
Processing Time
<30 seconds per order
Complexity
Medium — 8–10 weeks to deploy
Confidence Threshold
≥95% auto-confirm; <95% → human review
INPUTEmails + PDF/Word/Excel attachments · Customer PO PDFs · EDI fallback · WhatsApp forwarded confirmations
OUTPUTDraft Sales Order in D365 BC · Auto-confirmation email to customer · Exception alert to CSR · HubSpot deal/activity update
Microsoft Copilot StudioAI Builder (D365)
D365 BC REST APIHubSpot API
Power Automate PremiumAzure OpenAI (EU region)
Priority 1 · Phase 1 · Quick Win
📄 Agent 2: Import Document Intelligence Agent
Processes the ~200 daily supplier emails containing commercial invoices, packing lists, bills of lading, certificates of origin, EUR.1 forms, and customs entry documents. Using a purpose-built Intelligent Document Processing (IDP) engine, the agent extracts all structured fields from multi-format documents (PDF, scanned image, mixed-content), cross-references against open POs in D365 BC, flags discrepancies (missing fields, value mismatches, HS code anomalies), auto-populates GRN data into D365 BC, and triggers alerts to the customs broker (Van der Berg Logistics BV) with pre-validated document packages. Eliminates the 30% incomplete-document problem at source.
Daily Document Volume
~200 emails · ~400–600 PDF pages
Extraction Accuracy
≥98% with human review loop
Complexity
Medium-High — 10–12 weeks
Languages Supported
EN, ZH, DE, TA (supplier markets)
INPUTSupplier emails · PDF commercial invoices · Packing lists · B/Ls · Certificates of origin · EUR.1 forms · Scanned stamps/signatures
OUTPUTStructured data → D365 BC GRN · Auto-validated customs dossier · Discrepancy alert to trade compliance officer · Structured customs doc repository
80%
Customs Error Reduction
Rossum / Docsumo (EU-hosted IDP)Azure Document Intelligence
D365 BC APIMultimodal LLMPower Automate
Priority 1 · Phase 1 · Quick Win
💼 Agent 3: AP Invoice Automation Agent
Delivers end-to-end accounts payable automation for Meridian's ~3,648 annual supplier invoices. The agent ingests invoices from supplier emails, extracts all fields via IDP, performs automated three-way matching (Purchase Order / Goods Receipt Note / Invoice) against D365 BC records, applies configurable matching tolerance rules (e.g., ±2% on line amounts), routes matched invoices to automatic payment queue, and routes exceptions with full context to the AP clerk. Targets reduction in human review from 100% to <15% of invoices. Reduces cost per invoice from €14.50 to ~€3.50 and shrinks month-end close from 8 to 5 business days.
Annual Invoice Volume
~3,648 invoices/year est.
Auto-match Rate Target
>85% straight-through processing
Complexity
Medium — 8–10 weeks to deploy
Compliance
Full audit trail · GDPR · EU VAT rules
INPUTSupplier email invoices (PDF/XML/EDI) · D365 BC open POs · GRN records · Vendor master data
OUTPUTAuto-posted AP entries in D365 BC · Payment batch proposals · Exception queue for AP clerk · Audit trail log · Month-end accruals
D365 BC AI BuilderRossum / Stampli
Power AutomateExactOnline Bridge API3-Way Match Engine
Priority 2 · Phase 2 · Q1 2027
🚢 Agent 4: Export Document Generation Agent
Automatically generates complete, destination-compliant export document sets for each of Meridian's 6 export markets (Belgium, UK, France, Poland, UAE, Turkey) by reading confirmed shipment data from D365 BC and Shippypro. The agent maintains a live regulatory knowledge base for each destination (incoterms, certificate requirements, EUR.1/ATR eligibility, sanitary certificates, ATA carnets). Documents generated include: commercial invoice, packing list, CMR road consignment note, EUR.1 or ATR movement certificate, and EX-A customs declaration pre-fill. Replaces the manually updated spreadsheet and 2.8h/shipment document preparation time.
Daily Shipment Volume
~85 shipments/day
Markets Covered
BE, UK, FR, PL, UAE, TR + extensible
Complexity
High — 14–16 weeks to deploy
Regulatory Updates
Agent monitors EU trade regulation feeds
INPUTD365 BC confirmed sales orders · Customer & destination data · Shippypro carrier booking data · Incoterm & payment term settings · Product HS codes
OUTPUTComplete PDF document set per shipment · Pre-filled EX-A customs declaration · EUR.1 / ATR certificate · CMR · L/C pre-check validation report
D365 BC APIShippypro API
Regulatory Knowledge BasePDF Generation EngineRAG / LLM
Priority 2 · Phase 2 · Q1–Q2 2027
📊 Agent 5: AI Demand Forecasting & Replenishment Agent
Replaces Excel-based rolling 3-month sales trend with a machine learning forecasting model trained on 3 years of D365 BC order history across 2,400 SKUs. The model incorporates seasonal patterns, lead time variability per supplier (±12 day average variance), customer-specific ordering patterns, and ABC/XYZ inventory classification. The agent generates weekly purchase order proposals per supplier, dynamically adjusts safety stock levels per SKU based on service level targets, monitors stock positions against reorder points in real time, and sends automated PO draft proposals to buyers for one-click approval. Targets 35–50% reduction in the €620,000 annual stock-out/overstock cost.
SKU Coverage
2,400 active SKUs · ABC/XYZ classification
Forecast Horizon
13-week rolling (daily refresh)
Training Data
3Y D365 BC history (2023–2026)
Complexity
High — 12–16 weeks; requires data prep
INPUTD365 BC sales order history · Supplier lead time actuals · Confirmed POs · Safety stock parameters · ABC classification · Seasonal calendars
OUTPUTWeekly PO proposals per supplier → D365 BC · Safety stock alerts · Power BI forecast accuracy dashboard · Stock-out risk alerts · Overstock disposal flags
50%
Forecast Error Reduction
Azure ML / Slim4D365 BC Data
Power BIAuto-PO WorkflowTime Series ML
Priority 3 · Phase 3 · Q3 2027
🎧 Agent 6: Customer Service & Shipment Tracking Agent
Conversational AI agent handling inbound customer queries about order status, delivery ETAs, stock availability, invoice disputes, and return requests via email and optionally web chat. Integrates D365 BC (order/stock data) + Shippypro (live tracking) + HubSpot (customer context). Proactively pushes shipment status notifications without customer requests. Targets OTIF improvement from 84% → 90%+ by reducing last-mile communication failures.
Copilot StudioHubSpot API
Shippypro APID365 BC
Priority 3 · Phase 3 · Q4 2027
🏭 Agent 7: Supplier Intelligence & Compliance Agent
Automates supplier performance scorecarding (OTD rate, quality defect rate, document completeness score) for all 143 suppliers. Monitors EU/OFAC/UK OFSI sanctions lists in real time, replacing manual spot-checks. Flags HS code discrepancies before customs filing. Auto-requests corrected documents from suppliers. Monitors 76 import suppliers' financial health quarterly. Targets customs error reduction from 10/year to <2/year.
Sanctions Screening APIHS Classification AI
D365 BCSupplier Portal
Architecture Principle: All recommended components must satisfy four non-negotiable criteria: (1) EU data residency, (2) REST API integration with D365 BC, (3) WMS-agnostic architecture (no dependency on the current basic WMS module pending H2 2026 upgrade), and (4) open data portability (no proprietary data lock-in). The recommended architecture is built on Microsoft's existing M365 E3 investment as the primary platform, supplemented by best-of-breed EU-hosted IDP tools.
Recommended Technology Architecture
🧠 AI & Agent Layer
- Microsoft Copilot Studio — Agent builder for order intake and customer service agents; native D365 BC connectors
- Azure OpenAI (EU West) — LLM backbone for reasoning; GDPR-compliant; EU data residency
- Azure ML Studio — Demand forecasting model training and deployment
- Microsoft AI Builder — Document extraction, form processing within Power Platform
📄 Document Intelligence
- Rossum (EU-hosted) — Primary IDP for customs documents, invoices, shipping docs; pre-trained trade document models; D365 BC connector
- Azure Document Intelligence — Fallback OCR and prebuilt invoice model for standard formats
- Power Automate Premium — Orchestration layer connecting IDP outputs to D365 BC and HubSpot
🔗 Integration & Analytics
- Power Automate Premium — Primary integration middleware; 1,000+ connectors; existing M365 E3 licence
- D365 BC REST API — Primary system of record integration point
- HubSpot API (v3) — CRM synchronisation with ERP order data
- Power BI (extended) — AI ROI monitoring, forecasting dashboards, operational KPIs
- Shippypro API — Carrier tracking data into agent contexts
IDP Vendor Comparison — Document AI Platforms
| Criteria |
Rossum (Recommended) |
Azure Document Intelligence |
Docsumo |
UiPath Document Understanding |
Nanonets |
| EU Data Residency | ✓ Yes | ✓ Yes (EU West) | ✓ Yes | △ Configurable | ✗ US-only |
| Pre-trained Trade/Customs Models | ✓ Extensive | △ Invoice/receipt only | ✓ Good | △ Custom training needed | △ Limited |
| D365 BC Connector | ✓ Native | △ Power Automate | △ API | △ API | △ API |
| Multi-language (ZH, DE, EN) | ✓ Yes | ✓ Yes | ✓ Yes | ✓ Yes | △ Limited Asian lang. |
| Agentic / Workflow Orchestration | ✓ Built-in | ✗ Extraction only | △ Basic | ✓ Full RPA platform | △ Basic |
| Pricing Model | Per page / volume | Per page (Azure) | Per document | Per robot + licence | Per document |
| SMB/Mid-market Fit | ✓ Excellent | ✓ Good | ✓ Good | ✗ Enterprise-heavy | ✓ Good |
| IT Effort to Deploy | Low–Medium | Low | Low | High | Medium |
| AEO-C Audit Trail | ✓ Full | △ Partial | △ Partial | ✓ Full | ✗ Limited |
| Overall Recommendation | PRIMARY CHOICE | SECONDARY/FALLBACK | ALTERNATIVE | NOT RECOMMENDED | NOT COMPLIANT |
Demand Forecasting Platform Comparison
| Platform | D365 BC Integration | SMB Suitability | EU Hosting | Complexity | Est. Annual Cost | Recommendation |
|---|
| Slim4 by Slimstock | ✓ Native connector | ✓ Built for distributors | ✓ NL-based | Low | €15K–€30K/yr | RECOMMENDED |
| Azure ML (custom model) | △ Via API | △ Requires ML expertise | ✓ EU West | High | €8K–€15K/yr | IF IT SKILLS AVAILABLE |
| Inventory Planner | ✓ Native | ✓ Excellent | △ Configurable | Low | €6K–€12K/yr | ALTERNATIVE |
| SAP IBP | ✗ SAP-native | ✗ Enterprise only | ✓ | Very High | €80K+/yr | OVER-ENGINEERED |
Regulatory Framework — Meridian's Obligations
| Regulation | Applies To | Current Status | AI Impact |
|---|
| GDPR | All customer & employee data processing | ✓ Compliant | AI tools must use EU-hosted data centres; no personal data to US-only clouds |
| AEO-C Status | All customs & trade operations | ✓ Active | AI customs agents must maintain full, auditable decision log; human override must be logged |
| EU AI Act | AI systems used in regulated workflows | △ Planning | Document AI and trade compliance agents may qualify as "limited risk" — transparency obligations apply |
| REACH (Chemicals) | Safety product data sheets | ✓ Compliant | Agent 2 must flag REACH-relevant product documents for human compliance review |
| EU Customs Code | Import/export declarations | ✓ Active | AI HS classification must be reviewed by licensed broker; agent provides recommendation, not final ruling |
| OFAC / EU Sanctions | All export transactions | ✗ Manual/Weak | Agent 7 must automate screening; manual spot-checks create material compliance risk |
EU AI Act — Risk Classification for Meridian's Agents
| AI Agent | EU AI Act Risk Level | Obligations |
|---|
| Order Intake Agent | Limited Risk | Transparency disclosure to customers that AI processes their POs |
| Document Extraction Agent | Limited Risk | Audit trail; human review for customs declarations |
| AP Invoice Agent | Limited Risk | Full payment audit trail; exception escalation log |
| Export Doc Agent | Limited-High | Customs doc AI must have licensed broker oversight; AEO-C audit compliance |
| Demand Forecasting Agent | Minimal Risk | Internal decision support; buyer retains final PO approval authority |
| Customer Service Agent | Limited Risk | Must disclose AI nature to customers; human escalation path required |
| Sanctions Screening Agent | High Risk (potential) | Must not autonomously block transactions; flag for human compliance officer decision |
EU AI Act Enforcement Note: The EU AI Act's enforcement provisions began applying to high-risk AI systems in August 2025. For Meridian's use cases, the primary obligations are transparency (customers must know when AI processes their orders), human oversight (no fully autonomous customs or sanctions decisions), and audit trail maintenance. All recommended implementations include these controls by design.
Cybersecurity Posture for AI Deployment
Current Posture (Adequate as Foundation)
- Microsoft 365 E3: MFA enforced, DLP policies, Defender endpoint protection ✓
- Annual penetration test ✓
- ISO 27001 in planning — accelerate for AI programme credibility ⚡
- GDPR compliance policy and IT security policy documented ✓
- Basic data backup procedure in place ✓
Additional Controls Required for AI
- AI model access governance policy (who can configure/override agents)
- API key management and rotation policy for all agent integrations
- AI audit log retention policy (minimum 7 years for customs decisions)
- Data classification policy extended to cover AI training data
- Vendor security assessment checklist for all AI SaaS providers
- Agent "kill switch" protocol — documented escalation and suspension procedure
CFO Summary: At the stated budget midpoint of €150,000 for Year 1, the base-case model projects a cumulative net saving of €66,931 by end of Year 1, growing to €577,937 by end of Year 2 and €1,438,943 by end of Year 3. This represents a 3-year net ROI of ~540% on cumulative investment of €290,000. Payback occurs at approximately month 13–14 under the base case.
3-Year Cumulative Savings vs. Investment — Base Case (€)
Annual Savings Stack by Category (€)
Detailed 3-Year Financial Model — Base Case (65% Capture Rate)
| Cost / Saving Category | Annual Addressable Cost |
AI Capture % | Full Annual Saving |
Y1 (6mo ramp) | Y2 (full) | Y3 (optimised) |
| Email Order Entry — Labour | €131,546 | 80% | €105,237 | €52,618 | €105,237 | €115,760 |
| Import Document Processing — Labour | €43,848 | 80% | €35,078 | €35,078 | €35,078 | €38,586 |
| AP Invoice Processing — Cost/Invoice Reduction | €52,896 | 81% | €42,846 | €42,846 | €42,846 | €47,130 |
| Export Document Prep — Labour | €70,158 | 75% | €52,619 | €13,155 | €52,619 | €57,880 |
| Order Entry Errors — Credit Note Reduction | €100,320 | 70% | €70,224 | €35,112 | €70,224 | €77,246 |
| Customs Errors & Penalties | €28,000 | 80% | €22,400 | €11,200 | €22,400 | €24,640 |
| Inventory — Stock-out Cost Reduction | €390,000 | 40% | €156,000 | €39,000 | €156,000 | €187,200 |
| Inventory — Overstock Carrying Cost Reduction | €230,000 | 40% | €92,000 | €23,000 | €92,000 | €110,400 |
| Customer Churn Reduction (9% → 7%) | €165,600 | 30% | €49,680 | — | €49,680 | €74,520 |
| Supplier Performance / Sanctions Reduction | €38,000 | 50% | €19,000 | — | €19,000 | €28,500 |
| TOTAL GROSS ANNUAL SAVINGS | €1,250,368 | — | €645,084 | €252,009 | €645,084 | €761,862 |
| Investment costs (implementation, licences, advisory, training) |
| Implementation & Advisory (YTS + Vendors) | | | | −€120,000 | −€45,000 | −€35,000 |
| Software Licences (annual SaaS) | | | | −€30,000 | −€32,000 | −€35,000 |
| Internal Staff Training & Change Management | | | | −€15,000 | −€8,000 | −€5,000 |
| TOTAL INVESTMENT | | | | −€165,000 | −€85,000 | −€75,000 |
| NET ANNUAL POSITION | | | | +€87,009 | +€560,084 | +€686,862 |
| NET CUMULATIVE POSITION | | | | +€87,009 | +€647,093 | +€1,333,955 |
Scenario Analysis
3-Year Net Cumulative — Conservative / Base / Optimistic Scenarios (€)
Payback Curve — Monthly Cumulative Net Position (€)
| Scenario | Assumptions | Y1 Net | Y2 Cumulative | Y3 Cumulative | Payback Month | 3Y ROI |
|---|
| Conservative | 50% capture rate; 6mo delays; higher implementation overrun (+20%) | +€28,000 | +€390,000 | +€850,000 | Month 17–18 | ~360% |
| Base Case | 65% capture; standard timeline; budget at midpoint | +€87,009 | +€647,093 | +€1,333,955 | Month 13–14 | ~540% |
| Optimistic | 80% capture; fast adoption; demand forecasting exceeds targets | +€145,000 | +€920,000 | +€1,850,000 | Month 10–11 | ~720% |
Sensitivity Note: The single highest-impact variable is the Demand Forecasting Agent (Agent 5), which accounts for €248,000 of the base-case Year 2 annual saving. If this agent underperforms by 50% (€124,000 actual saving), the base-case 3-year cumulative drops to ~€1.09M — still a 3-year ROI of ~340%. Even the worst-case scenario (conservative with forecasting underperformance) delivers payback within 24 months — within the CFO's stated requirement.
Critical Dependency Note: Meridian is in early discussions with a WMS upgrade vendor for H2 2026. All Phase 0 and Phase 1 AI implementations are architected at the ERP/document layer and are 100% WMS-agnostic. No Phase 1 agent depends on the current D365 BC basic WMS module or any future WMS. Warehouse automation agents (GRN digitalisation, slotting optimisation) are deliberately deferred to Phase 2, post-WMS selection, to ensure compatibility.
Phase 0 · Foundation Sprint
Data Readiness, Architecture Design & Governance Setup
📋 Full data quality audit (customer, product, supplier, HS codes)
🔗 D365 BC API documentation & test environment setup
🏗 ERP–CRM integration architecture design (HubSpot ↔ D365 BC)
📁 Legacy document archive migration roadmap
⚖ EU AI Act compliance review & risk classification
📊 AI ROI baseline measurement (establish current KPIs)
👥 Change management kickoff — departmental AI champions appointed
🔒 AI governance policy drafted (access, audit, override procedures)
🤝 WMS vendor AI-compatibility requirements documented
Deliverables: AI Programme Charter · Data Quality Report · Integration Architecture Document · AI Governance Policy · Phase 1 Vendor Selection
Budget: ~€15,000 · Owner: YTS Consulting + Pieter Loos (IT Manager)
Phase 1 · Quick Wins — Highest Immediate ROI
Order Intake Agent + AP Invoice Automation + Document Extraction Agent
🤖 Agent 1: AI Order Intake Agent — pilot (10% of email orders) → production rollout
💼 Agent 3: AP Invoice Automation — deploy on Rossum IDP + D365 BC 3-way match
📄 Agent 2: Import Document Extraction Agent — deploy on top 15 suppliers first
🔗 HubSpot ↔ D365 BC integration — stock/order visibility for sales team
📊 Power BI AI ROI Dashboard — live KPI monitoring goes live Week 12
🎓 Staff training: logistics team (Agent 2), finance (Agent 3), CSR team (Agent 1)
⚡ Power Automate enhancement — build on existing PO acknowledgement flows
✅ 90-Day CFO Review Gate — KPI vs. baseline report presented to Sophie Janssen
Expected Outcome: 127 FTE hours/week freed from manual data entry · €130,000 annual savings active · Month-end close reduced from 8 → 5 days
Budget: ~€90,000 · Owner: YTS Consulting + IT Manager + Finance Director
Phase 2 · Core Operations — Maximum Cost Impact
Export Document Agent + Demand Forecasting AI + Warehouse GRN Digitalisation
🚢 Agent 4: Export Document Generation Agent — all 6 export markets
📊 Agent 5: Demand Forecasting ML Model — training, validation, production deploy
🔄 Auto-PO proposal workflow live — buyers review, one-click approve
🏭 Warehouse GRN digitalisation (contingent on WMS selection)
💳 Credit limit automated alerts to sales reps (Power Automate)
📋 Backorder management module activation in D365 BC
📈 Quarterly price list automated refresh workflow
🎓 Procurement team training on AI demand forecasting tool
Expected Outcome: Full €591,000 annual saving trajectory achieved · OTIF improvement trajectory starts · 48h/week export doc prep eliminated
Budget: ~€45,000 · Owner: YTS Consulting + COO + Procurement Manager
Phase 3 · Intelligence Layer — Customer & Supplier Excellence
Customer Service AI + Supplier Intelligence + Compliance Automation
🎧 Agent 6: Customer Service AI Agent (order status, tracking, query resolution)
🏭 Agent 7: Supplier Performance Scorecard Automation
⚖ Real-time EU/OFAC sanctions screening automation
🔢 HS code classification AI — systematic review of 2,400 SKUs
🌐 Customer self-service portal upgrade (tracking, invoices, returns)
💱 FX risk monitoring dashboard (CNY exposure flagging)
📊 Full Power BI AI operations dashboard — all 7 agent KPIs live
Expected Outcome: OTIF target of 92%+ achieved · Customer churn 9% → 7% · Sanctions compliance automated · 9% churn → 7% saving ~€33K/year
Budget: ~€35,000 · Owner: YTS Consulting + COO + Customer Service Manager
Phase 4 · Full Agentic Operations (Strategic Horizon)
Multi-Agent Orchestration, Autonomous Operations & New Market Expansion
🌐 Multi-agent orchestration platform — agents collaborate autonomously
🤝 Autonomous supplier negotiation support agent
🔮 Predictive customer churn & upsell AI
🤖 WMS robotics integration (AMR readiness post-WMS upgrade)
🌍 AI-assisted expansion into 2 new export markets (stated strategic goal)
📡 Digital twin — supply chain simulation & scenario planning
Risk Heat Map — Probability vs. Impact
Risk Category by Priority Score (Probability × Impact)
Probability: High · Impact: High
Customs docs at 2/5; customer master duplicates; HS codes inconsistent. AI models trained on poor data produce inaccurate outputs, potentially creating compliance exposure (AEO-C) and customer-facing errors.
✅ Mitigation: Phase 0 data sprint before any agent deployment. Dedicated 4-week data quality remediation on top 500 SKUs (HS codes), customer master deduplication, and supplier master refresh. Use Agent 2 (IDP) to bootstrap structured customs data from existing PDFs retroactively. Assign named data stewards per domain.
Probability: High · Impact: High
Pieter Loos (IT Manager) + junior admin cannot manage AI programme delivery, BAU support, WMS upgrade evaluation, and integration development in parallel. Risk of project delays and technical debt.
✅ Mitigation: YTS Consulting embedded as technical AI programme lead, reducing internal IT burden. Prioritise Microsoft-native, low-code tools (Copilot Studio, AI Builder, Power Automate) that minimise custom development. Defer WMS-dependent work to Phase 2 to prevent timeline collision.
Probability: Medium–High · Impact: High
COO flagged change management as a concern. 210 h/week of manual work will be automated. Risk of non-adoption, passive resistance, or key staff exits — particularly in the CSR team (order entry) and AP team (invoice processing).
✅ Mitigation: Frame AI as augmentation, not replacement. Commit publicly to no redundancies driven by AI in Year 1. Redeploy saved time to customer success, data quality improvement, and export market development. Appoint 1 AI champion per department. Involve affected staff in agent design workshops. Celebrate quick wins loudly.
Probability: Medium · Impact: High
Sophie Janssen has stated a clear ROI requirement. If Phase 1 results are not measured and communicated clearly within 90 days of go-live, budget withdrawal for Phase 2 is a realistic risk.
✅ Mitigation: Implement live AI ROI dashboard (Power BI) from Day 1 — tracking: orders processed automatically, invoices auto-matched, hours saved, error rates. Schedule formal 90-day CFO review with quantified results vs. baseline. Phase 1 agents are specifically chosen to show measurable financial results within 60–90 days of deployment.
Probability: Medium · Impact: Medium
Planned WMS upgrade (H2 2026) could conflict with warehouse AI if not designed compatibly. Risk of duplicate implementation effort or architectural incompatibility.
✅ Mitigation: All Phase 1 AI agents operate at ERP/document layer — completely WMS-agnostic. Include AI compatibility as mandatory evaluation criterion in WMS vendor RFP. YTS Consulting to review WMS vendor shortlist for AI integration readiness. Warehouse automation deferred to Phase 2 post-selection.
Probability: Low–Medium · Impact: High
Company policy prohibits US-only cloud storage per GDPR. Many AI vendors default to US data centres or have ambiguous residency commitments, creating contractual and compliance risk.
✅ Mitigation: Mandatory EU data residency clause in all AI vendor contracts. Preferred Microsoft Azure EU West/North Europe regions — natively GDPR-compliant. Rossum (EU-headquartered) selected as primary IDP vendor. All vendor contracts reviewed by YTS Consulting for data residency compliance before signature.
Probability: Low · Impact: Very High
Authorised Economic Operator status is a commercial asset. AI-assisted customs decisions that cannot be fully audited or that bypass licensed broker review could jeopardise AEO-C status.
✅ Mitigation: All AI customs agents provide recommendations, never final rulings. Full decision audit log with timestamps. Licensed customs broker (Van der Berg Logistics BV) retained in the approval workflow for all customs declarations. AI dossiers explicitly flagged as "AI-assisted, human-approved" in customs system.
Probability: Low · Impact: Medium
COO flagged vendor lock-in concern. UiPath (proprietary robot licences) and Microsoft Copilot (platform dependency) were cited concerns during previous evaluation phase.
✅ Mitigation: All agent integrations use open REST APIs (D365 BC, HubSpot, Shippypro). Data stored in D365 BC (client-owned). Avoid proprietary AI data formats. Negotiate data export portability in all SaaS contracts. Rossum and Azure Document Intelligence both support standard JSON/XML data export. Modular architecture allows component-by-component replacement.
Critical Finding: 80–90% of Meridian's enterprise data is currently trapped in unstructured formats (PDF, email, scanned documents, spreadsheets). The single most important pre-condition for AI success is not the AI tools themselves — it is the systematic extraction and structuring of this data. No AI agent, however sophisticated, can deliver accurate outputs from unstructured, incomplete, or inconsistent input data.
Data Remediation Sprint Plan (Phase 0 — 8 Weeks)
| Domain | Current Score | Actions Required | Weeks | Owner | Blocks Agent |
|---|
| Customs & Trade Docs | 2/5 | Deploy IDP on last 12 months of import PDFs to create structured archive; establish doc intake SOP; supplier document quality checklist | Weeks 1–6 | Trade Compliance Officer | Agent 2, Agent 4, Agent 7 |
| Customer Master Data | 3/5 | HubSpot–D365 BC deduplication run; standardise address format (CASS/postal validation); mandatory contact refresh by sales reps | Weeks 2–4 | Sales Manager + IT | Agent 1, Agent 6 |
| Product & HS Codes | 3/5 | Manual review sprint on top 500 highest-volume SKUs; AI HS classification cross-check; standardise product description format | Weeks 3–8 | Trade Compliance Officer | Agent 2, Agent 4, Agent 7 |
| Supplier / Vendor Master | 3/5 | Email campaign to all 143 suppliers for data refresh (banking, lead times, contacts); update D365 BC vendor master; mandatory annual refresh SOP | Weeks 1–4 | Procurement Manager | Agent 3, Agent 5, Agent 7 |
| Logistics & Shipping | 3/5 | Shippypro API → D365 BC structured tracking data capture; historical carrier performance retroactive import | Weeks 4–6 | IT Manager | Agent 4, Agent 6 |
| Pre-2023 Sales History | 4/5 | Assess pre-migration data for demand forecasting model; cleanse or exclude from training if quality too low | Weeks 2–3 | IT Manager + Finance | Agent 5 |
Ongoing Data Governance Framework
Data Stewardship Model
- Customer Master: Sales Manager (steward)
- Product/HS Codes: Trade Compliance Officer
- Supplier Master: Procurement Manager
- Inventory/WMS: Warehouse Manager
- Financial Data: CFO / Finance Manager
- AI Training Data: IT Manager (custodian)
Data Quality KPIs (Post-AI)
- Customer master duplicate rate: <1%
- Supplier master completeness: ≥95%
- HS code accuracy rate: ≥99%
- Inventory accuracy: ≥98% (from 91%)
- Document extraction accuracy: ≥98%
- AI training data refresh: quarterly
No-Code Data Integration
- HubSpot ↔ D365 BC bi-directional sync via Power Automate
- Shippypro API → D365 BC automated tracking updates
- Rossum IDP → D365 BC structured document data pipeline
- Power BI refresh: daily from D365 BC, hourly for operational KPIs
- ExactOnline bridge: monthly reconciliation automation
The COO has explicitly flagged change management as a concern. Industry data consistently shows that AI programme failures are caused more often by people factors than by technology failures. Gartner projects that 40% of agentic AI projects will be cancelled by 2027 — not because the technology does not work, but because organisations skip the human foundations. This section addresses those foundations systematically.
Stakeholder Impact & Engagement Plan
| Department | FTE Affected | Primary Impact | Concern Level | Engagement Strategy | Redeployment Opportunity |
|---|
| Customer Service / Order Entry | 3 FTE (partial) | 90h/wk email order entry automated → freed for value-add tasks | High | Early involvement in agent design; demonstrate speed benefit to them; appoint CSR as AI champion | Customer success management, proactive account development, portal onboarding support |
| Finance / AP Team | 1–1.5 FTE (partial) | 37h/wk invoice keying automated → freed for financial analysis | Medium | CFO to communicate redeployment plan; AP clerk becomes "exception reviewer" — higher value role | Working capital optimisation, early payment discount capture, spend analytics |
| Logistics / Trade Compliance | 2 FTE (partial) | Document collection and export prep automated → more complex trade work | Medium | Trade compliance officer positioned as AI governance owner — elevated role | HS code governance, new market regulatory research, AEO-C programme management |
| Procurement / Buying Team | 1 FTE (partial) | Demand forecasting replaces Excel; buyer reviews AI proposals vs. manual calculation | Medium | Frame as decision-support, not decision-replacement; buyer retains final PO authority | Supplier relationship development, negotiation, dual-source strategy, new market sourcing |
| Warehouse Operations | 22 FTE | Phase 1 does not affect warehouse directly; GRN digitalisation in Phase 2 | Low | Brief update in Phase 0; involve warehouse supervisor in Phase 2 GRN design | Process improvement roles; quality control; customer returns management |
| IT Team | 2 FTE | AI programme adds workload; YTS Consulting absorbs programme management | Medium | Clear scope separation: YTS leads AI delivery; IT team owns BAU and WMS evaluation | API governance, AI security administration, WMS integration architect role |
Change Management Framework — 5-Stage Model
Stage 1
Awareness
CEO all-hands briefing on AI programme vision. No job loss commitment in Year 1 communicated clearly and in writing.
Stage 2
Understanding
Department-level AI workshops (1hr each). Demo of Agent 1 prototype with CSR team. Questions answered transparently.
Stage 3
Buy-In
AI champions appointed. Staff involved in agent testing and feedback. Quick wins celebrated and communicated company-wide.
Stage 4
Adoption
Phased rollout with support buddies. Exception handling owned by affected teams — they become the "quality controllers" of AI output.
Stage 5
Advocacy
Top performers become internal AI trainers. Savings reinvested visibly. Staff redeployment to strategic roles shows personal career benefit.
KPI
Success Measurement
AI adoption rate per agent (target: ≥85% by Month 6); staff satisfaction survey; exception escalation rate (declining = trust building).
Meridian vs. Best-in-Class: Key Operational KPIs
2026 AI Adoption in B2B Wholesale & Logistics
Industry Benchmark Comparison
| KPI | Meridian Current | Industry Average | Best-in-Class (AI-enabled) | Meridian AI Target | Gap to Close |
|---|
| OTIF Delivery Rate | 84% | 88% | 97% | 92% | +8pp (target) / +13pp (BIC) |
| Order Error Rate | 4.1% | 2.5% | 0.8% | 1.2% | −2.9pp (target) |
| Days Sales Outstanding | 47 days | 42 days | 32 days | 40 days | −7 days |
| AP Invoice Cost | €14.50 | €10.80 | €2.78 | €3.50 | −€11.00 (target) |
| Inventory Accuracy | 91% | 93% | 99.5% | 98% | +7pp (target) |
| Forecast Error (MAPE) | ~35% est. | 25% | 12% | 18% | −17pp (target) |
| Customer Query Response | 4.2 hours | 2.5 hours | 15 minutes (AI) | 30 minutes | −3.7 hours |
| Customer Churn Rate | 9% | 7% | 4% | 7% | −2pp (target) |
| Order Automation Rate | 28% (EDI only) | 45% | 85% | 75% | +47pp (target) |
| AP Straight-Through Processing | ~0% | 35% | 85% | 80% | +80pp (target) |
2026 Market Intelligence Headlines
Agentic AI ROI Benchmark
Companies report average 171% ROI from agentic AI deployments. US enterprises average 192%. This exceeds traditional automation ROI by 3×. 74% of executives achieved ROI within Year 1 of deployment. (Source: AIMonk/Landbase 2026)
Logistics & Supply Chain
McKinsey estimates AI in supply chain reduces logistics costs by 5–20% and forecasting errors by up to 50%. AI-powered innovations could reduce logistics costs by 15%, optimise inventory by 35%, and boost service levels by 65%. (Source: Microsoft/McKinsey 2025)
AP & Document Automation
Best-in-class AP teams using AI spend €2.78/invoice vs. Meridian's €14.50 — a 5.2× disadvantage. Automating AP reduces processing costs by up to 80% and cycle time by similar margins. AI achieves 99–99.9% extraction accuracy. (Source: APQC/Parseur 2025–2026)
Overall Recommendation: Proceed with the phased Agentic AI programme as described. The business case is sound, the technology foundation is adequate, and the financial returns significantly exceed the investment. The primary success factor is not the technology — it is disciplined execution of the Phase 0 data readiness sprint and robust change management. Both are entirely within Meridian's control.
Immediate Actions — Next 30 Days
| # | Action | Owner | Deadline | Priority |
|---|
| 1 | Approve AI Programme Charter and confirm Year 1 budget allocation (€150,000) | CEO + CFO | Week 2 | Critical |
| 2 | Appoint internal AI Programme Owner (recommend: Operations Manager reporting to COO) | COO | Week 1 | Critical |
| 3 | Engage YTS Consulting on advisory retainer for programme delivery support | COO + CFO | Week 2 | Critical |
| 4 | Document WMS vendor AI-compatibility requirements before issuing RFP | IT Manager + COO | Week 3 | Critical |
| 5 | Issue CEO all-hands communication on AI programme — vision, no-redundancy commitment, and benefits | CEO | Week 2 | High |
| 6 | Initiate Rossum EU-instance proof of concept — process 100 sample customs invoices and packing lists | IT Manager + Trade Compliance | Week 4 | High |
| 7 | Begin customer master deduplication in HubSpot — assign to Sales Operations | Sales Manager | Week 3 | High |
| 8 | Brief Sophie Janssen (CFO) on this audit financial model — confirm 90-day KPI review gate process | COO + YTS | Week 1 | High |
Recommended Technology Stack — Final Summary
| Layer | Recommended Tool | Purpose | Est. Annual Cost | Rationale |
|---|
| Agent Platform | Microsoft Copilot Studio | Order intake & customer service agents | Included in M365 | Native D365 BC connectors; EU-hosted; no extra licence |
| AI Reasoning | Azure OpenAI (EU West) | LLM backbone for all agents | €8K–€15K/yr (usage) | GDPR-compliant; EU data residency; enterprise SLA |
| Document Intelligence | Rossum (EU-hosted) | IDP for customs, invoices, shipping docs | €12K–€20K/yr | Pre-trained trade models; D365 BC connector; AEO-C audit trail |
| Demand Forecasting | Slim4 by Slimstock (NL) | ML forecasting for 2,400 SKUs | €18K–€28K/yr | NL-headquartered; D365 BC native connector; SMB-optimised |
| Integration Middleware | Power Automate Premium | All agent orchestration & ERP–CRM sync | €8K–€12K/yr | Existing M365 investment; 1,000+ connectors; low-code |
| Analytics | Power BI (extended) | AI ROI monitoring & operational KPIs | €3K–€6K/yr | Already deployed; familiar to finance team; D365 BC native |
| TOTAL ESTIMATED ANNUAL LICENCE COST | €49K–€81K/yr | Within Year 1 budget envelope of €150K (including implementation) |
Individual Stakeholder Briefing Notes
For Erik van den Berg (COO)
The WMS-agnostic architecture protects Phase 1 investment regardless of WMS outcome. 210 h/week of manual processing can be reduced by ~60% by end of Phase 1. Change management is the primary execution risk — allocating a dedicated internal AI champion role is the single most important governance decision you can make.
For Sophie Janssen (CFO)
Even the conservative scenario (50% capture, +20% cost overrun) delivers payback within 18 months and 3Y ROI of ~360%. Phase 1 agents (AP + Order Entry) target €147,696 in annual savings against €90K in Phase 1 costs — a first-year return exceeding the investment. The 90-day KPI gate provides a low-risk decision point before Phase 2 budget is committed.
For Pieter Loos (IT Manager)
The recommended stack is Microsoft-native and cloud-managed. Your primary role in Phase 1 is API governance, D365 BC test environment management, and security policy. YTS Consulting leads vendor management and agent configuration. The WMS evaluation should include an AI integration readiness checklist — we will provide this. No custom development is required in Phase 1.
Summary: Projected KPI Improvement — Current vs. AI-Enabled Target
Ready to Begin?
Schedule Your Discovery Call with YTS Consulting
Yasar Tezeren · AI Advisor & Ex-CEO · 40+ years executive experience · PMI-certified Project Manager
🌐 yts-consulting.com
📧 yasar@tezeren.net
📞 +33 6 42 26 74 12
💼 yts-agents.com