Pickware has extended its merchandise management system with artificial intelligence. Pickware AI is firmly integrated into the ERP and WMS for Shopware and Shopify and works with the data that already resides there: orders, stock levels, storage locations, suppliers, customers and revenues. For context: An ERP (Enterprise Resource Planning) is the software for merchandise management, i.e. for orders, stock levels and purchasing. A WMS (Warehouse Management System) controls the processes in the warehouse.
Pickware names three ways to use this AI: the AI Chat in the administration, an MCP server through which, for example, Claude or ChatGPT can connect to the Pickware data, and the AI Agents that run in the background. This article is about the Agents. They are the most recent of the three building blocks.
The AI Chat: Questions in Your Own Words
In the AI Chat you ask Pickware directly in the ERP, just as you would ask a colleague. How many orders are open, how does the margin change if a supplier raises purchase prices, which customers have overdue invoices. The answer comes from the shop data, without anything having to be exported beforehand or totaled up in a spreadsheet. Tasks such as a return, an exchange or a dunning run can also be triggered in the chat. Writing actions of this kind are included from the Advanced plan onward.
What Are Pickware AI Agents?
The chat responds when you address it. An agent, on the other hand, works on its own. It starts according to a schedule or on an event in the shop, checks the data and files the result as a message in an inbox, a shared message inbox in the administration. This makes agents suitable for audits and reports that recur regularly, such as a daily warehouse briefing or a check of the master data.
A run can take 20 minutes or longer depending on the task and amount of data. The schedule should therefore be set so that the result is in the inbox before work begins.
How to Set Up an Agent
You create an agent in the chat. You describe in your own words what it should take over, for example: check every working day at 6:30 a.m. which items are oversold, and only send a message if there are such items. If an important detail is missing, the chat asks for it. It shows the draft as an "Agent definition" card with task, trigger and settings, which you confirm or change in the chat. Anyone who does not want to start from scratch begins with a template such as the AI warehouse manager. You will find the agents in the administration in the menu under "Agents".
Very important: An agent cannot ask questions during a run. Everything it needs belongs in the task, i.e. which data it should check and from what value something counts as noteworthy.
When an Agent Starts
Schedule: for example daily at 7 a.m. or every Monday.
Event in the shop: such as when a goods receipt is completed or an order is shipped. Which events are available depends on the installed extensions.
Manually: via "Run" on the detail page you start a run immediately.
There is no trigger of the kind "a threshold has been fallen below". You solve such cases with a schedule and a check in the task, for example check every morning whether a product is below 20% margin.
Trigger: schedule, event in the shop or manual start
Check: The agent evaluates the data according to the thresholds from its task
Inbox: The result is placed as a message in the shared inbox
Decision: You review the proposal and implement it. The agent itself does not change any data.
What You Should Keep in Mind
An agent does not change any data during its runs. It evaluates and makes suggestions; the implementation remains up to you, in the administration or in the chat on the respective message. It does not send emails to customers or suppliers, but it can prepare a text suggestion for this.
Because an agent is based on an AI language model, two runs of the same task can turn out differently and figures can be inaccurate. You should therefore check critical values in the administration before acting on them. State thresholds concretely: "below 25% margin" is usable, "margin too low" is not.
Availability in the Plans
Until 31 December 2026, agents can be used in all Pickware plans and during the trial phase. After that they are included from the Advanced plan onward; the full range of functions is available from the Professional plan.
Our Recommendation:
Until the end of the year, trying it out costs nothing extra. We would start with a single agent that delivers a report that someone currently compiles by hand, and compare the first results against the figures in the administration for a week. Only once the messages are reliable is the next agent worthwhile.
With the "Pickware ERP, WMS and POS" plugin, Pickware combines merchandise management, warehouse logistics and point-of-sale system in a single Shopware installation. For the implementation we support you with our Pickware consulting and planning, from process analysis to training your team.
Related products
Comments
No comments yet.
Write a comment