Thinking
Why practical AI succeeds where transformation fails
Case studies, thought leadership and practical guidance on using AI to improve everyday work.
Unified Commerce Intelligence: Orchestrating Revenue Across Omnichannel B2B Transactions
A £60M foodservice distributor processes orders through four channels: Shopify Plus, EDI feeds, phone, and WhatsApp. Each channel lands in a different system. Web orders sync to the ERP overnight. EDI orders arrive every 15 minutes but don't update inventory visibility until the next batch run. Staff key phone orders directly into the ERP. Staff transcribe WhatsApp orders into a spreadsheet, then manually enter them. The operations director calculated the cost. Pricing inconsistencies between channels cost £180K annually in margin leakage. A customer quoted £4.20/kg for chicken thighs on the phone receives a different price (£4.45/kg) when they order the same product via the website three days later. They call to query it. The phone team has no visibility of the web order. They create a second order. By the time staff spot the duplicate, the customer has been invoiced twice and the stock has been picked. Resolution takes 90 minutes and involves three people.
Agentic AI Observability: Detecting When Your Autonomous Systems Are Operating on Bad Data
Agentic systems fail quietly. A pricing agent drifts 2% below target margin over eight weeks, processing 4,200 decisions before anyone notices. Margin leakage: £87,000. The agent didn't crash or throw errors—it just made slightly wrong decisions, consistently, for two months. Gartner predicts 40% of agentic AI projects will be cancelled by 2027, primarily due to silent degradation that compounds into operational disasters.
Why Fortune 500 Supply Chain AI Fails at Mid-Market: The Complexity Mismatch Problem
Fifty-seven percent of supply chain leaders cite data quality as the primary barrier to AI adoption. Not model accuracy. Not cost. Data quality. Most mid-market distributors have data spread across five to seven systems with no single source of truth. You cannot train an AI model on conflicting data. The unglamorous work of master data management, integration, and governance must come first.
Inventory Decisions Under Uncertainty: When AI Forecasts Conflict With Safety Stock Rules
Deploying demand sensing AI without retiring legacy safety stock policies creates a hidden cost: dual everyday work. Inventory planners second-guess AI recommendations, override autonomous reorder points, and maintain manual guardrails 'just in case.' This article quantifies the cost of running both systems, explains why most deployments fail at the governance layer, and outlines the workflow clarity required to let one system own inventory decisions.
Enterprise Operations Fragmentation: Why Mid-Market Distributors Can't Integrate Their Way Out of Workflow Delay
Mid-market distributors run on 5-8 operational systems. They've invested in integration platforms. Yet workflow delay persists. Integration solves data movement, not workflow clarity.
System Integration Challenges: When Legacy Meets Modern AI
A legacy ERP connects to a modern AI pricing platform. A field mapping breaks. Prices don't update. Nobody notices for 48 hours. By then, £15,000-£30,000 in margin has leaked. Silent failures are worse than loud ones because they compound. Real-time data observability makes these failures visible and resolvable before they cost money.
Agentic AI Data Architecture: Why Your Supply Chain Intelligence Fails Without the Right Foundation
A pricing agent at a South Yorkshire distributor adjusted 1,200 SKUs based on supplier cost data that was six weeks stale. The pipeline reported healthy. The data quality checks passed. The system lost £34,000 in margin over three months before anyone noticed. The agent hadn't malfunctioned - it had operated exactly as designed on information that was technically valid but operationally worthless. This is the observability gap that makes autonomous systems dangerous. Traditional monitoring tracks schema compliance and pipeline health. Agentic AI needs semantic monitoring - systems that understand whether data makes sense in the context of the decision being made, not just whether it arrives on time in the right format.
The Skill Deficit Crisis: Why 84% of Mid-Market Distributors Lack AI-Ready Talent
Mid-market distributors are hiring data scientists for problems that don't need them. The real gap isn't technical capability—it's workflow clarity and domain expertise. Most AI hiring fails because businesses copy tech company job specs instead of recruiting for operational roles that understand distribution workflows.
LLM Observability: Why Custom AI Models Need Different Monitoring
Custom AI models degrade silently in production. Most mid-market distributors discover this weeks after the damage begins—when pricing errors accumulate, inventory misallocates, or search results decay. Model drift is silent cost leakage that compounds over time.
Predictive Orchestration: The 2026 Supply Chain AI Priority CSCOs Can't Ignore
Most supply chain directors spend Monday mornings reviewing exceptions from last week. A supplier shipment arrived late. A customer order exceeded forecast. A warehouse ran short on a fast-moving line. By the time the team decides what to do, the problem has compounded.
Event-Driven Architecture: Real-Time Commerce Operations at Scale
A West Midlands foodservice distributor runs seven systems: SAP for ERP, Shopify Plus for commerce, Manhattan for warehousing, Salesforce for CRM, Xero for accounting, Akeneo for product data, and a legacy routing tool built in-house. None of them talk to each other in real time. Pricing updates take three days. Inventory visibility lags 24 hours behind actual stock. The operations director spends Monday mornings reconciling conflicts created by systems working from different versions of the truth. This is the middleware bottleneck. Mid-market B2B distributors have invested in best-of-breed systems but lack the integration layer to connect them. The B2B middleware market reached £14.1 billion in 2025, growing at 12.23% annually (MarketsandMarkets, 2024). That growth reflects a painful reality: buying good systems is easy, making them work together is not. Event-driven architecture promises to solve this. Order placed, inventory updated, fulfilment triggered, customer notified—all in real time. But most mid-market stacks lack the middleware to orchestrate that sequence. Without it, teams fall back to manual processes and the operational friction compounds.
Edge Intelligence in Distribution: Real-Time Decisions at the Warehouse Floor
Most mid-market distributors assume warehouse automation requires robotics. Reality: software orchestration delivers faster ROI without the capital spend. A Manchester plumbing merchant spent £480,000 on conveyors but orders still took 90 minutes from pick to dispatch because systems didn't communicate. The problem wasn't hardware, it was coordination. This article covers why warehouse execution systems matter, how AI decision logic works between platforms, real-time resource allocation mechanics, and the implementation reality most vendors skip.
Agentic Supply Chain Orchestration: Controlled Automation Across Multi-Depot Networks
Most mid-market distributors allocate stock across depots manually. A buyer reviews demand signals, checks inventory levels, and decides where to send stock. This takes hours or days. By then, demand has shifted. Autonomous agents change this by making allocation decisions in real-time — but only if governance frameworks define the boundaries within which they operate.
Turning Tacit Expertise into Executable Workflow Support
A pricing manager adjusts margins across 800 SKUs in under an hour. A senior buyer reallocates stock between depots without checking a spreadsheet. A trade counter manager prioritises contractor orders instinctively. All three decisions generate better outcomes than their peers—higher margins, fewer stockouts, faster turnaround. Nobody can explain why.
Task-Specific AI Agents in B2B Commerce: From Procurement to Order Management
A procurement exception flags in your ERP at 4:47pm on Friday. Supplier lead time extends from 14 to 21 days. Inventory on your fastest-moving line drops below reorder point. The alert sits unread until Monday morning. By then, the supplier's order window has closed for the week. You've lost seven days, and your customer delivery commitment is now at risk.
Composable Commerce Architecture: Building Modular Systems That Evolve With Your Business
A Midlands foodservice distributor spent £240,000 recovering from a pricing error in 2024. Their monolithic commerce platform took three days to update prices across 12,000 SKUs. Commodity costs moved faster than their approval cycle. By the time new prices went live, margins had eroded. The problem wasn't the platform. It was the architecture. Pricing, inventory, and order management sat in one system. Upgrading the pricing engine meant ripping out everything else. The platform vendor quoted 18 months and £180,000 for a custom pricing module. The distributor couldn't wait that long. This is the composable commerce question: when one component fails, can you swap it without dismantling the entire system?
Agentic Sales Agents: When AI Handles Customer Negotiations in Distributed Channels
Most B2B distributors take 2-5 days to generate a complex quote. Sales teams spend 30-40% of their time on quote administration rather than relationship-building. Autonomous negotiation systems can issue quotes in under two minutes—but only if decision thresholds are explicit and governance is built in from the start. Without clear boundaries, AI agents commit companies to commercial terms they didn't intend.
Market Sensing Intelligence: Reading Weak Signals Before Competitors
Quarterly forecasts miss demand shifts that happen in days. Demand sensing uses real-time weak signals—POS data, supply disruptions, social trends—to detect changes 24-72 hours before they appear in historical patterns. For mid-market distributors managing volatile categories, this decision speed translates directly into margin capture and stockout reduction.
Budget Allocation Intelligence: How AI Prevents Mid-Market Distributors From Funding Dead Projects
Eighty percent of AI projects fail to deliver intended business value. The failures are predictable and preventable. Data governance gaps, scope creep, stakeholder misalignment, and unrealistic timelines appear weeks before deployment—but most distributors commit capital without spotting them. This article walks through the specific red flags that signal project failure and provides a framework for pre-deployment validation that takes 1-2 days, not months.
Why Big 4 Partnerships Fail at Operational Decision Implementation: The Deloitte-Palantir Model Exposed
Most mid-market distributors engage Big 4 consultancies expecting a one-time project. They end up in perpetual dependency that costs £500K-£2M+ annually. This isn't accidental — it's the business model. Contract structures, methodology licensing, and support dependencies lock you in. Here's how the economics work, and what independence actually requires.
Agentic AI Advisory: When Your Operations Need Real-Time Workflow Support, Not Quarterly Reviews
A foodservice distributor takes three days to change a price. By the time the spreadsheet is approved and uploaded, the margin opportunity has passed. A seven-depot building merchant routes deliveries manually each morning, missing optimisation windows that close by 9am. A fashion retailer delays markdown decisions until stock levels force action, tying up £180,000 in slow-moving inventory for an extra six weeks. Workflow delay costs more than imperfect decisions made fast.
Controlled Automation: The Governance Framework Mid-Market Distributors Need Before Deployment
Autonomous agents can accelerate decisions and reduce costs in B2B commerce. But most mid-market distributors lack the governance frameworks to deploy them safely. Without audit trails, decision boundaries, escalation rules, and performance monitoring, autonomous systems become liabilities. This article maps the four control pillars required before autonomous everyday work goes into production - and the practical roadmap for implementing them without killing velocity.
Agentic AI Governance: Building Control Systems Before Deployment
A pricing agent at a West Midlands distributor adjusted 847 SKUs overnight. By Monday morning, high-margin industrial fasteners were underpriced by 14%, creating £47,000 in margin leakage before anyone noticed. The agent had no anomaly detection. No alert threshold paused execution when margins dropped below cost. No observability layer showed which input triggered the repricing cascade. This happens when teams deploy autonomous agents without monitoring infrastructure. Agents make decisions at scale, without human review, across pricing, fulfilment, and replenishment. A single bad decision compounds across hundreds of transactions before anyone spots it. Observability isn't optional when agents control operational workflows.
Agentic Pricing Intelligence: When Custom Models Set Prices Autonomously
Most B2B distributors take three days to change a price. By the time it's live, the margin opportunity has passed. Autonomous pricing agents compress this cycle from days to minutes - but only if governance is built in from the start. Without it, you hand control to a system that optimises for volume while destroying margin.