Bottleneck 1: order entry that is still manual
At most B2B wholesalers, orders arrive through a mix of email, PDF, portal, and EDI. A staff member reads the PDF, re-enters the data into the ERP, and checks the article codes. That easily takes five to ten minutes per order, and at a volume of a hundred orders per day that is half an FTE doing nothing but retyping. The AI solution here is document AI: a system that reads incoming order files, matches articles based on customer-specific codes, and presents the completed order line to a staff member for approval. The person decides; the AI does the groundwork. Off-the-shelf packages sometimes offer this as a module, but they only work well when customers send neatly standardised files. In practice, that is a small minority. Once you are dealing with dozens of customers who each use their own layout, custom development wins: an extraction layer trained on your customer base, connected to your ERP logic.
Bottleneck 2: inventory management by gut feeling rather than data
Phocas published a report in March 2026 based on a global study of more than a hundred distributors. The finding: 70% manage more than 5,000 SKUs, yet only 11% rate themselves as 'highly accurate' in demand forecasting. That is a significant gap. Most wholesalers work with historical averages and manual corrections by buyers. AI-driven demand forecasting takes a different approach: it combines seasonal patterns, customer behaviour, and lead times into a single model and gives buyers a concrete replenishment recommendation. McKinsey links this type of approach to inventory savings of 20 to 30% and lower storage costs. Whether you need custom development depends on how distinctive your product range and customer mix are. For a relatively stable range with few exceptions, a standard inventory optimisation tool works fine. As soon as you are dealing with items that are heavily dependent on external factors, such as weather-sensitive products or commodity prices, a purpose-built forecasting layer is preferable.
Bottleneck 3: a customer portal that customers do not use
Many wholesalers already have a customer portal, but adoption falls short. Customers still call and email because the portal does not fit the way they work. They cannot see real-time stock, cannot track order status using their own reference numbers, or the ordering process is too generic for their procurement procedure. An AI solution does not fix this in one go, but can intervene in three places. A chat interface on top of the portal: customers ask questions in plain language and get answers from live data. A smart search function: finding articles by description rather than exact code. Proactive notifications: the portal automatically flags that a reorder makes sense based on the customer's order pattern. Off-the-shelf packages sometimes deliver this as configurable modules. But if your customers use their own article references, have specific contract prices, and are restricted from ordering certain items, a standard portal quickly runs into exceptions. In that case, custom development is cheaper in the long run than endless configuration.
Bottleneck 4: returns logistics as a blind spot
Returns are an underestimated problem in B2B wholesale. They arrive without a clear reason, are assessed manually, and disappear into a separate stream that is disconnected from the regular inventory administration. The result: items sit unprocessed in a corner for weeks, credit notes fall behind, and buyers have no visibility into what is available as used stock. An AI solution for returns logistics starts with structure: every return is assigned a reason, an expected condition, and a routing at the point of registration. A model looks at historical patterns and immediately advises whether an item can go back into regular stock, to an outlet, or to disposal. The connection to the ERP ensures the credit note is created straight away. Whether this requires custom development depends on the volume and complexity of your return flows. With more than a few dozen returns per day, or with items requiring specific inspection, a purpose-built workflow pays off.
Which bottleneck do you tackle first?
The decision rule is straightforward: tackle the bottleneck where manual work disappears every day that can be directly traced to a concrete error or delay. In most wholesale businesses, that is order entry. The volume is high, the errors are visible, and the impact on customer satisfaction is immediate. An AI solution for order entry also has a short payback period: the hours freed up are measurable, and the implementation does not touch the entire ERP at once. Inventory management follows as the second priority if you manage more than 3,000 SKUs and regularly face shortages or overstock. The customer portal becomes a priority when adoption is low and customers structurally place orders outside the system. Returns logistics deserves attention when processing takes longer than two working days and credit notes are running behind. The mistake many directors make: they start on all four tracks simultaneously and fail to build sufficient momentum on any of them. Choose one bottleneck, build it properly, learn from it, and then take the next step.