This report has ten chapters, structured from methodology to concrete advice, plus an appendix with all sources.
This report is the deliverable of the Redesaign Scan: an individual twenty- to thirty-minute conversation with each member of the leadership team, combined with public research from the major consultancies, industry analysis, and Redesaign's own expertise. The outcome is not a technical inventory, but a sharp picture of which work processes gain the most from redesign, and why.
Five individual interviews, each structured around the same twenty questions, supplemented with organizational data and external research.
CEO, COO, CFO, commercial director, and HR director, each twenty to thirty minutes, twenty questions per conversation.
Every work process mentioned is scored on redesign value, ease of implementation, and strategic impact.
Checked against research from McKinsey, BCG, Celonis, and current industry data for logistics and supply chain.
Redesaign doesn't look at isolated tasks, but at the assumption a work process was originally built on. Every organizational structure that exists today was rationally designed decades ago around three fundamental constraints. Once it's clear which constraint drives a process, it becomes visible whether AI can play a structural role there, and what kind.
Decision-makers rarely have all the relevant information at the right time and in the right place. Work is therefore designed around collecting, passing on, and checking information: reports, forms, handoff moments.
Friedrich Hayek described in 1945 that knowledge in an organization or economy is never concentrated in one place, but dispersed across many individuals. Classic structures, hierarchy, reporting lines, approval rounds, are essentially mechanisms for bringing that dispersed information together.
AI systems can bring together large volumes of scattered, unstructured information, documents, transactions, sensor data, in a short time and make it searchable. What used to take a reporting cycle of weeks can now be continuously available and up to date.
Human judgment is a limited resource. Every organization must choose where people spend their attention: on high-value judgment, or on repetitive processing that crowds that judgment out.
Herbert Simon introduced the concept of bounded rationality: people don't make decisions with unlimited computing power, but seek a satisfactory solution within the limits of time, information, and cognitive capacity. Many processes are therefore designed with simple rules and fixed steps, precisely to keep the cognitive load manageable.
Generative and predictive AI can take over a large share of routine judgment work, freeing human cognition to concentrate on exceptions and the decisions that truly matter.
Coordinating work between people, teams, and departments costs time and resources: meeting, handing off, waiting for a reply before the work can continue.
Ronald Coase showed in 1937 that firms exist because coordinating work through the market, buying in every step separately, carries transaction costs. Internal hierarchy exists to limit those coordination costs, often at the expense of flexibility and speed.
Agentic AI can initiate, follow up on, and complete tasks without a human having to coordinate every step, and can work across system boundaries in a way that is cumbersome for people. Coordination costs fall as a result, especially at handoff points between departments.
These three assumptions were never wrong. They were the rational basis for organizational design in a world where information, cognition, and coordination were scarce and costly. AI doesn't change the assumptions themselves, it changes the costs underneath them. That's why process redesign must come before tool implementation: only when a process is redesigned around the new cost structure does it create structural value, rather than a faster version of the old process.
Source: McKinsey & Company (March 2025), The state of AI: How organizations are rewiring to capture value.
A 3PL provider that coordinates warehousing, transport planning, order fulfillment, and returns logistics for around sixty regular clients in food, non-food retail, and light industry.
In 2023, a regional investment firm took a minority stake in Meridian to finance growth. The family retains control, but the organization must demonstrably scale within four to five years. Staff shortages in the warehouses are squeezing margins, and clients increasingly ask for real-time transparency instead of periodic reporting.
A warehouse management system (WMS) and a transport management system (TMS, since 2021) run alongside each other without a full integration. Forecasting and staff planning happen largely through Excel. Customer service works with a standalone CRM, phone, and email.
For each leadership-team member: the biggest energy drains, which of the three assumptions weighs heaviest, the balance between judgment and routine work, and the strongest opportunities. Click a profile for the full interview.
Redesaign distinguishes three assumptions that classic organizations are built on: information is scarce, cognition is scarce, coordination is costly. AI fundamentally changes all three. Below is the weighted average across the five interviews.
Score per assumption: the weighted percentage of mentioned work processes that rely on that assumption, across all five interviews.
Information scarcity dominates: the WMS, TMS, CRM, and accounting package don't talk to each other, and figures are manually pulled together before a decision can be made. Coordination costliness follows closely: between DCs, between operations and customer service, and between HR and operations. Cognition scarcity shows up mainly in customer service and finance, where volume crowds out room for judgment. This pattern points directly to the three opportunity areas later in this report.
Every work process mentioned is scored on redesign value, ease of implementation, and strategic impact. The combination of those three scores determines the top 3.
| Process | Redesign value | Ease of implementation | Strategic impact |
|---|---|---|---|
| 1. Demand forecasting & inventory management | |||
| 2. Customer service & order handling | |||
| 3. Invoicing, documents & POD processing | |||
| Management reporting | |||
| Transport planning & routing | |||
| Supplier & contract management | |||
| Warehouse staff scheduling | |||
| Returns logistics & claims handling | |||
| Recruitment & onboarding | |||
| Quality control & incident reporting |
The logistics sector generates more operational data per employee than almost any other industry, yet lags behind in AI adoption. That gap is exactly Meridian's opportunity.
AI adoption rate by sector. Source: TechWize (2026), based on the Gartner Supply Chain Technology Report 2025 and McKinsey State of AI 2025.
Meridian already has the data: the WMS records every scan, every delivery, every deviation. The problem isn't data availability, it's that the data isn't used to look ahead. That's exactly the difference between automating an existing process and redesigning around what the data already makes possible.
Not every opportunity calls for the same technology. Redesaign distinguishes three levels of AI deployment, increasing in autonomy. Each opportunity area in this report calls for a different level, and often a combination.
AI helps draft, summarize, and search. The human does the work and decides, AI speeds up the input.
AI carries out a defined task independently, from signal to recommendation, within preset boundaries. The human steers and checks the exceptions.
AI carries out an entire process or sub-process, from start to finish, and only surfaces on exceptions. The human sets the boundaries.
BCG calls this four waves of integration depth: chatbots and copilots, AI-driven workflows, autonomous agents, and multi-agent systems. Value increases as AI decides and acts more, rather than just assisting, and most organizations are still in the first two waves. Microsoft's own transformation methodology uses a comparable three-level distinction, and stresses that the level can differ by step within a process, not just by process as a whole.
Sources: BCG (July 2026), Driving Sustained Structural Cost Advantage with Applied AI. Microsoft (2026), Becoming a Frontier Firm: Our Frontier Playbook.
Each area below combines the score on redesign value, ease of implementation, and strategic impact with the four AI actions: eliminate, automate, strengthen, preserve. Each area also shows which technology level applies, and how the situation differs before and after redesign.
Inventory decisions at Meridian rely on manual Excel forecasts and planners' experience. Sales data from the WMS is barely used to look ahead. This process scores highest on strategic impact: it touches margin, customer satisfaction, and warehouse recruitment pressure all at once.
Forecast in Excel, one to two weeks behind reality. Planners spend most of their time on number-crunching. Excess inventory and missed service levels exist side by side.
Forecasts update automatically every day based on current WMS and sales data. Planners only see what deviates from the pattern. The conversation with clients shifts from "what went wrong" to "what's coming".
Manual Excel forecasts and duplicate data entry between the WMS and purchasing system.
Baseline forecasting per client and item based on historical sales, seasonal, and client data.
Planners get forecasts with confidence margins and exception alerts, and focus on deviations and supplier conversations.
Inventory choices for new clients without history and for unusual peaks remain human work.
Potential ranges based on McKinsey research into AI in distribution operations, via Open Sky Group (2026). Actual results for Meridian are established during the Redesaign process.
About eighty percent of client contact at Meridian consists of repetitive track-and-trace questions that, with the right integration between the WMS, TMS, and CRM, can be answered independently. The remaining twenty percent, genuine disruptions and escalations, is where customer service adds the most value and currently has the least time for.
Customer service calls operations, operations calls back, and only then does an answer go out. Our best people give answers a system could give just as well.
Standard questions are answered instantly and independently from live order data. Customer service has room for disruptions, escalations, and genuine account conversations.
Repeatedly answering the same status question by phone or email.
A first line that connects directly to WMS and TMS data for status, delivery time, and documentation.
Customer service staff focus on disruptions, escalations, and account relationships.
Complex complaints and the largest accounts always stay personal contact.
Reference values based on comparable agentic customer service implementations, as described in Redesaign's research base (Sierra, Decagon). Actual results for Meridian are established during the Redesaign process.
PODs, transport invoices, and credit notes arrive at Meridian by email, scan, and portal, and are largely matched by hand. This process scores highest on ease of implementation: the pattern is predictable, the volume is high, and the tools to recognize and match documents are readily available.
PODs and invoices are retyped from emails and scans. Month-end close takes two weeks of manual checking, with a team that mostly enters data instead of analyzing it.
Documents are recognized and matched automatically, deviations stand out immediately. The finance team focuses on exceptions, supplier conversations, and analysis, and the close shifts from weeks to days.
Manually retyping PODs and transport invoices from emails and scans.
Document recognition and matching of PODs, invoices, and contract terms, with automatic flagging of discrepancies.
The finance team focuses on discrepancies, supplier conversations, and analysis instead of data entry.
Contract negotiation and judgment on major disputes remain human work.
Reference values based on comparable document-processing cases in professional services, from Redesaign's research base. Actual results for Meridian are established during the Redesaign process.
Alongside the top 3, the interviews surfaced seven other processes with AI potential that currently score lower on redesign value, ease of implementation, or strategic impact. None of these are hopeless, they follow after the first round of redesign, or once the prerequisites are further in place.
Monthly figures arrive late and are pulled together by hand. Automatic, continuously updated reporting would enable faster steering.
Already partly tooled through the TMS since 2021. Further optimization adds value, but less redesign value than the top 3 since the foundation is already in place.
Carrier contract terms and rates are tracked manually. Overlaps with the invoicing opportunity, but smaller strategic impact on its own.
Returns and transport damage claims are assessed manually case by case. A recognizable volume pattern, but smaller in scale than the top 3 at Meridian.
Weekly scheduling is done manually and depends on forecast quality. Only gains real value once demand forecasting has been redesigned.
Warehouse incidents and quality deviations are logged separately. Potential for pattern recognition, but currently limited volume to automate against.
Screening and paperwork are labor-intensive, but the strategic impact is smaller than the top 3. Still relevant given the structural recruitment pressure.
Process redesign is the core, but it only sticks if the organization has the prerequisites in place. This isn't a reason to delay starting, but it is context for a realistic action plan.
The WMS, TMS, CRM, and accounting package all run separately. For all three opportunity areas, a basic integration between these systems is a prerequisite, not an afterthought. A shared data foundation, not a full system replacement, is the starting point: agents can only reason reliably if they draw from the same source.
Meridian isn't alone in this. Only nineteen percent of organizations use multi-agent systems today, the vast majority are still exploring. The data foundation, not the ambition, is usually the bottleneck.
Needs attentionFour of the five leadership-team members report curiosity within their teams, especially among the employees who do the most repetitive work. That's a good foundation, provided it's deliberately guided: honest communication about what's changing, redesigning roles rather than cutting them, and managers who lead by example turn out to be the strongest predictors of adoption.
A specific point of attention for Meridian: new warehouse employees learn the trade partly through the routine work that is now being automated. Entry-level work needs to be redesigned, not simply cut, or the training path into more experienced roles disappears.
Solid foundationThere is no ownership yet per AI initiative, and no framework for responsible use. That doesn't have to block the first step: four questions are enough to get started. Who owns each AI initiative? What actions may an agent take independently, and what actions may it not? How does an action stay traceable and reversible? And how is client and business data protected?
This is a widely recognizable gap: among comparable organizations, 41% report having no policy for responsible AI use, and only 21% have actually formalized one. Setting up governance can happen alongside the first Redesaign process, it doesn't need to come first.
Needs attentionOrganizations that extract structural value from AI don't anchor it as a standalone pilot, but as a business transformation with a clear owner. Five elements recur: clear business goals rather than productivity alone, an operating model that embeds the new way of working, sustained investment in people, documenting the organization's own methods and quality standards, and security that grows alongside the autonomy of agents.
None of these prerequisites need to be fully in place before Meridian can start. They are context for a realistic action plan, and become part of the Redesaign process itself: data integration, ownership, and the training approach are addressed per process rather than as a separate project upfront.
A process redesign doesn't stick because you ran a training, it sticks because you led the change itself. Redesaign uses John Kotter's eight-step model for this, summarized into five practical blocks worked through per process. For Meridian, that is made concrete below for the first two redesigns: demand forecasting and customer service.
Sanne and Rutger share concrete figures on recruitment pressure, missed service levels, and manual work with the full leadership team and the teams themselves. Rutger and the senior planner form the driving coalition for the first redesign.
For each process, it's made concrete what changes for the planner, the customer service employee, and the finance employee: what work disappears, what work stays, and what new work appears. Not an abstract AI vision, but a vision per role.
Training happens in small, role-focused groups tied to the day-to-day work, not a one-off classroom session. Obstacles are actively removed: freeing up time for training, and making room for questions and concerns in regular team meetings.
The first visible result, for example less manual POD matching within a few weeks, is shared widely across the organization. Visible quick wins build trust for what follows.
The new way of working becomes part of job profiles, onboarding, and performance reviews, so the organization doesn't slip back into the old routine once attention fades. This aligns with Redesaign's principle of self-sufficiency: the organization can carry the new way of working forward itself.
Source: John Kotter (1996), Leading Change, summarized as the 8-step model for change.
The Scan stands on its own as valuable advice. Below is the follow-up track as Redesaign would propose it, purely for illustration.
Five leadership interviews, an organization-wide picture, and this report with the top 3 opportunity areas.
A Flow Canvas session per chosen process, starting with demand forecasting & inventory management, together with the team involved.
Route A for implementation, route B for buy-or-build advice, or route C for the Academy, matching what the Redesaign process delivers.
None of these follow-up steps are mandatory. The Scan is a standalone product and stands on its own value. The Redesaign process and the follow-up routes are only recommended if the Scan gives reason to, as is the case for Meridian.
Every data point in this report is cited. The interviews themselves were written to be illustrative for this sample report.