Back to Blog

How to set up an AI agent in JTL: step by step

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder at Superkind

A weathered dark wooden shipping crate with metal corner caps and the orange Superkind mascot sitting on top holding a clipboard

You want an AI agent in JTL that actually earns its keep: Monday, 09:12. Sven, who runs operations at a mid-sized online retailer, opens JTL-Wawi. 61 new orders overnight across JTL-Shop, Amazon and Kaufland. Three of them are B2B with a credit limit question, nine have returns from last week waiting to be booked, a supplier email about a delayed container is still unanswered, and two colleagues in service have questions about tracking numbers. Every day looks like this before anyone touches a product. This guide shows how to set up an AI agent in JTL step by step, so most of that work is done before Sven’s first coffee. You will also see how Superkind’s ready-made AI team for JTL dispatches orders, keeps product data clean and books returns on its own.

Two paths lead to an AI agent in JTL today. JTL’s own agent, Juno, was announced at JTL Connect in Cologne on 2 October 2026 as a closed beta with the open beta in November 2026, and it will cover reorders and customer-service tickets at launch. In parallel, external AI employees like Superkind connect to JTL-Wawi, JTL-Shop and your marketplaces today and cover the full back office. This guide uses Superkind for the step-by-step, so you can start before Juno is generally available and take advantage of both when it is.

Key takeaways

Start in minutes, not months - sign up for free at app.superkind.ai, connect JTL-Wawi and your inbox, and give the first task like “sort incoming orders”. You see work done the same day, without a JTL upgrade or a plug-in on every workstation.

You decide what needs approval - routine orders, standard returns and supplier chases run on their own. Price exceptions, B2B credit over your limit and refunds over EUR 150 wait for your OK in Teams, Slack or Outlook.

Explain it once, it keeps running - a rule like “Whenever a new article is created in JTL-Wawi: write the description, map the category, push to all channels” turns one task into a routine you never touch again.

Get a ready-made AI team for JTL - Superkind comes with an operations team (Order Dispatcher, Product Keeper, Return Handler, Supplier Scout) that run automatically on new orders, articles, returns and purchase orders. Start for free with unlimited users, or build your own.

Keep control, keep the data in Europe - Superkind is hosted and processed in Europe, GDPR compliant, EU AI Act aligned, no training on your data, DPA through the Trust Center.

Your AI team for JTL

orders-jtlOperations
Order DispatcherAI employee
Routine · every 15 minutes
47 new orders this morning: 44 booked and ready for picking, 3 wait for your OK (B2B credit limit).
JTL-Shop orders22 booked
Amazon + Kaufland22 booked
Over credit limit3 pending
B2B order over credit limitSchulz GmbH, EUR 4,820 (limit EUR 3,500). Waits for your OK.ApproveReject
Example: Order Dispatcher at work in Microsoft Teams

What is an AI agent in JTL?

An AI agent in JTL is a software employee that reads and writes in JTL-Wawi, JTL-Shop and JTL-POS on your behalf: it books incoming orders across your sales channels, writes and pushes product data, processes returns and credit notes, chases supplier confirmations, updates stock and answers routine tickets with real order data, while a person approves the steps you choose. JTL stays your source of truth; the agent handles the work on top of it. One example is Superkind: a ready-made AI team that connects to JTL-Wawi via the 1,500+ integrations and does these tasks across Outlook, Teams, Slack and the Superkind web app, hosted in Europe.

The difference to a generic chatbot or a Zapier flow: an AI agent does not stop at an answer or a single trigger. It knows the context of your JTL shop (which article belongs to which category, which customer is a repeat buyer, which SKU is low on stock) and it keeps the knowledge across tasks, so the next return or the next B2B order uses the same rules.

How it works with Superkind: your AI team for JTL

Superkind is the AI employee for your team: you write to Superkind by email (Outlook or Gmail), in Microsoft Teams, in Slack or in the Superkind web app, like you would write to a colleague, and Superkind does the task in your programs. Hosted in Europe, GDPR compliant, with 1,500+ integrations into the systems where JTL merchants already work: JTL-Wawi, JTL-Shop, Amazon, Kaufland, eBay, Shopify, DATEV, lexoffice, DHL, DPD, Outlook and Teams.

You do not start from scratch: Superkind comes with ready-made AI employees for JTL merchants. They run on their own in the background, triggered by a new order, a new product or a new ticket, and they wait for your OK only on the steps you want checked. Use them right away, adapt them, or build your own.

Order Dispatcher

AI employee for JTL order management

Orders arrive from your JTL-Shop, Amazon, Kaufland and eBay. The Order Dispatcher reads every new order in JTL-Wawi, checks stock, assigns the right warehouse, writes the delivery note and books the order. For B2B customers it checks the credit limit and holds anything above it for a quick OK in Teams.

Works inJTL-WawiMicrosoft TeamsOutlookDHL+ 1,700 more
orders-jtlOperations
Order DispatcherAI employee
Routine · every 15 minutes
47 new orders this morning: 44 booked and ready for picking, 3 wait for your OK (B2B credit limit).
JTL-Shop orders22 booked
Amazon + Kaufland22 booked
Over credit limit3 pending
B2B order over credit limitSchulz GmbH, EUR 4,820 (limit EUR 3,500). Waits for your OK.ApproveReject
Example: Order Dispatcher at work in Microsoft Teams

Product Keeper

AI employee for JTL product data

New products come from suppliers in Excel, with missing attributes, no SEO text and the wrong category tree. The Product Keeper takes each new article in JTL-Wawi, writes the description for JTL-Shop and Amazon from the supplier data, maps it to your category tree and pushes it to every channel. New SKUs are live the same day.

Works inJTL-WawiExcelSharePoint+ 1,700 more
Superkind
New chat
My tasks
AI employees
Product Keeper
Order Dispatcher
Return Handler
New article · Steel bottle 500 ml black

Whenever a new article is created in JTL-Wawi: write the shop description, map the category, push to all channels.

Reasoning ⌄
Product Keeper 11:02
New article in JTL-Wawi

Description written, category Drinkware > Bottles > Insulated, pushed to JTL-Shop, Amazon and Kaufland. Attributes complete.

Attributes filled14 of 14
ChannelsJTL-Shop, Amazon, Kaufland
Live in shop11:04

Runs automatically for every new article

Send a message ...
Example: Product Keeper at work in Superkind

Return Handler

AI employee for JTL returns

A DHL return parcel is announced, the customer wrote a short note in the box. The Return Handler reads the DHL notification, matches it to the order in JTL-Wawi, books the return, writes the credit note, puts the item back into stock and emails the customer. Refunds over EUR 150 wait for your OK.

Works inJTL-WawiOutlookDHL+ 1,700 more
Return 4198827 · Order SB-20314
service@yourshop.de
Customer sent back one bottle, wrong colour. Please process.
Return HandlerAI employee
New return notification
Return booked in JTL, item back in stock, credit note 2026-R-0411 written, customer notified. Refund EUR 29.90.
Items1 bottle
Reasonwrong colour
RefundEUR 29.90
Runs on its own under EUR 150
Example: Return Handler at work in Outlook

Supplier Scout

AI employee for purchasing and delivery dates

Half your delivery dates in JTL-Wawi are guesses because the supplier never confirmed. The Supplier Scout reads every open purchase order, emails the supplier for confirmation, parses the reply and updates the expected date in JTL-Wawi. Delays over three days get flagged to purchasing in Slack.

Works inJTL-WawiOutlookSlack+ 1,700 more
#purchasing
Supplier ScoutAI employee
Routine · every morning at 7:30
18 purchase orders chased today. 14 confirmed, 2 delayed by less than 3 days, 2 delayed more than a week.
Confirmed14 of 18
Delayed 7+ days2 (Hansen, Wolff)
Updated in JTL18
Runs automatically every morning
Example: Supplier Scout at work in Slack

Every employee is ready to use right away. Or build your own on the platform.

Safe by default

Order data, customer addresses and prices from JTL stay in Europe. Superkind never uses them to train a model.

From the first task to your own JTL app

The same four-stage path you see on the Superkind site, told once on an JTL example. Each stage builds on the last.

  1. 1

    Start

    Ready in minutes. Sign up for free at app.superkind.ai, connect JTL-Wawi and Outlook, and write the first task: “Sort the incoming orders and tell me what needs approval.” First run in minutes.

  2. 2

    Relief

    Tasks become routines. The one-off order sort turns into a rule: every 15 minutes new orders are booked, B2B credit above the limit waits for your OK. The same for product data pushes and standard returns.

  3. 3

    Knowledge

    Learns from day one. Your Company Brain grows with every task: your category tree, your customer price lists, your return reasons, your supplier contacts. When a colleague moves on, the knowledge stays.

  4. 4

    Growth

    When it gets more complex, you do not start over. Build a custom JTL cockpit on the Superkind platform, with its own interface for your margins per channel, on top of what Superkind already knows. Built yourself or with the Superkind team, and it belongs to you - your own Super App.

Send Superkind your first JTL task

Start for free in Outlook, Teams or Slack. Connect JTL-Wawi in minutes.

A dark matte industrial handheld barcode scanner with a coral-orange trigger button on a soft grey studio backdrop

What you need

Check these points before you start. Most of them are already in place in any JTL-run shop today.

Checklist before the first task

  • A free Superkind account at app.superkind.ai
  • A running JTL-Wawi (any current version) with API access enabled, or SQL read access for the parts the agent needs
  • JTL-Shop and your marketplace channels (Amazon, Kaufland, eBay) connected to JTL-Wawi the way you already run them
  • Access to the shared mailbox the shop uses (service@..., bestellung@..., retouren@...) in Outlook or Gmail
  • Microsoft Teams or Slack for the approvals and the status pings
  • Your DATEV or lexoffice connection, if you want the agent to prepare bookings
  • Your DHL, DPD or GLS account for return labels and tracking
  • An agreed approval policy per topic (e.g. B2B orders over EUR 3,500: approval required; refunds over EUR 150: approval required; everything else: runs on its own)
  • An admin who can confirm which JTL areas and shared mailboxes Superkind may read and write
  • Thanks to the 1,500+ integrations in Superkind, JTL-Wawi, your marketplaces, DATEV, Outlook and Teams can be connected without new tools or custom code

Step by step: your first AI agent in JTL

Six concrete steps, from a blank Superkind account to a JTL agent that works a full day on its own. Each step is something you can do today, and each ends with a short “how you know it works” check.

Step 1: Create a free Superkind account (Start stage)

Open app.superkind.ai, sign up with your work email and skip the credit card. The free start comes with a one-off up to EUR 150 in credits, unlimited users and every feature. Add your two or three colleagues from operations and service to the workspace right away, so approvals can land in the channels they already use.

  • How you know it works - you see the empty workspace at app.superkind.ai, your colleagues are listed as members and the sidebar shows the AI employees from the template library.

Step 2: Connect JTL-Wawi via the Superkind integration

In the Superkind integrations panel, pick JTL-Wawi (one of the 1,500+ integrations). An admin confirms which areas Superkind may read and write: orders, products, stock, returns, purchase orders. For JTL-Shop you connect the shop as a separate channel, same for Amazon, Kaufland, DHL and DATEV. No API keys to copy by hand, no plug-in on every workstation.

Superkind web appIntegrations

Connect JTL-Wawi. Scope: read orders, products, stock, returns. Write: order status, credit notes, product descriptions.

  • How you know it works - the Integrations page shows JTL-Wawi as “Connected”, with the scopes you approved, and the next test task reads live order numbers from your shop.

Step 3: Give the first task in Outlook or Teams

This is the moment that proves the agent works for you. Pick one real task you do every morning. For most JTL merchants that is sorting incoming orders.

TeamsSven to Superkind

Look at all new orders in JTL-Wawi since 06:00. Book the ones under EUR 1,000 for B2C. For B2B, check the credit limit and list everything above it for me.

What Superkind does: it reads the new orders in JTL-Wawi, checks stock, writes delivery notes for the straightforward B2C orders, and posts a short list of the B2B orders over the limit in the Teams channel.

  • How you know it works - you see the booked orders in JTL-Wawi with the right status, and the Teams message lists the exceptions with Approve / Reject buttons.

Step 4: Set the approval policy (what waits, what runs)

This is where you keep control. Tell Superkind, once, which steps need a human OK before they go out. The rest runs on its own.

TeamsSven to Superkind

Approval rules for JTL: B2B orders over EUR 3,500 wait for me. Price changes over 5 % against the price list wait for me. Refunds over EUR 150 wait for me. Everything else runs.

Superkind confirms with “Noted.” Every exception comes with a short card: the order, the price list, the reason, Approve or Reject. Everyone on the team sees the same cards in the channel they work in, and the audit log records who approved what.

  • How you know it works - the first B2B order over EUR 3,500 shows up as a card in Teams, and nothing above the thresholds leaves the system without a click.

Step 5: Turn one-off tasks into routines (Relief and Knowledge stages)

After three or four days of one-off tasks, stop forwarding orders by hand. Tell Superkind what should happen every day, every hour or on every trigger. This is where tasks become routines and the Company Brain starts to grow.

SlackSven to Superkind

Routines: every 15 minutes, sort and book new JTL orders. Whenever a new article is created in JTL-Wawi, write the shop description and push it to Amazon and Kaufland. Every morning at 7:30, chase open purchase orders. Returns under EUR 150 run on their own.

Each new routine starts to feed the Company Brain: your category tree, your customer price list, your return reasons, your supplier contacts. The knowledge stays even when the person who taught it leaves the team.

  • How you know it works - the Routines page in Superkind lists all four, each with its trigger and last run time, and the morning JTL-Wawi dashboard shows no backlog of untouched orders.

Step 6: Build your own Super App on top (Growth stage)

When the four JTL employees cover your daily work, the next step is a view your team actually opens every morning. On the Superkind platform you build a custom app with its own interface, on top of everything Superkind already knows about your JTL data. For a JTL merchant that often means a “Shop Cockpit”: open orders per channel, returns of the week, slow-moving SKUs, supplier delays, in one page.

  • Built by your team or with Superkind - no new middleware, no new login, the app lives in the same Superkind workspace as the AI employees.
  • Keeps using the 1,500+ integrations - the cockpit reads live from JTL-Wawi, Amazon, Kaufland and DATEV, so there is no second data layer to maintain.
  • Belongs to you - the app, the knowledge and the data stay yours.
  • How you know it works - the whole operations team opens the cockpit instead of five tabs, and the agent reports its own work there every day.

Set it up as a routine

Once the first week of one-off tasks has run cleanly, keep adding small routines. A simple pattern for JTL merchants:

  • Every 15 minutes - sort and book new orders across JTL-Shop, Amazon, Kaufland and eBay. Exceptions to Teams.
  • Every hour - check stock for low-running SKUs and prepare reorder suggestions for purchasing.
  • Every morning at 7:00 - chase open supplier confirmations in Outlook, update delivery dates in JTL-Wawi, post delays in Slack.
  • Every morning at 7:30 - book standard returns under EUR 150, write the credit note, send the customer email, put stock back.
  • On every new article - write the JTL-Shop description, map the category, push to Amazon and Kaufland.
  • Every Friday at 16:00 - build the weekly report from JTL data: orders per channel, returns per reason, margin per SKU group.

When this grows, for example a dedicated routine for B2B customers or a cross-channel price rule, you do not start over. On the Superkind platform you build a custom app with its own interface, on top of everything Superkind already knows about your JTL data. The 1,500+ integrations keep the app connected to JTL-Wawi, Amazon, Kaufland and DATEV without new middleware. See how to build an AI agent without code for the method.

Data protection and EU hosting

JTL data is customer data: names, addresses, order history, prices and sometimes bank details. These records must stay in Europe. German consumer protection and competition authorities also look closely at refund and warranty handling, so the audit trail matters as much as the hosting location.

What this means for a JTL merchant: the order data Superkind reads, the credit notes it writes and the customer emails it sends never leave the EU and are not used to train a model. Every change goes into an audit log, so you can show in an audit who booked a refund or changed a product description. For most shops this is the first time refund and return handling is traceable step by step.

Other ways to automate work in JTL

Superkind is not the only path. Here is an honest comparison of the common options JTL merchants evaluate today.

WayEffortError riskBest for
SuperkindLow: connect JTL-Wawi, write rules in plain languageLow: approval where you want it, audit log, EU hosting, 1,500+ integrationsJTL merchants who want orders, returns, product data and supplier chases handled across Outlook, Teams, Slack and the web app without new middleware
JTL-Workflows and JTL-AmeiseMedium: built-in, but limited to rule-based automation and imports/exportsLow inside JTL, no cross-channel reasoningFixed, deterministic steps that never change (status transitions, mass updates)
Juno by JTL (beta)Low once generally available: ships with JTL, requires ERP version 2.2+Low in its first scope (reorders, service), narrower than a full back-office agentJTL shops that want reorder suggestions and standard customer-service answers close to the ERP
Zapier or Make flowsMedium: build and maintain triggers per task, per channelMedium: brittle on edge cases, no shared knowledge, data often leaves the EUSimple two-system links (e.g. new order to Google Sheet) for a very small shop
Hire a part-time back-office clerkHigh: recruiting, onboarding, holiday coverLow on known tasks, high on peak days (sales, Christmas)Teams who prefer a human for every judgement call, where volume is modest and stable

For other sides of the same stack, see our guides on AI order management, verifying invoices with AI and AI agents for your ERP.

Common mistakes

These five mistakes cost the most time when a JTL shop starts with an AI agent.

  • Starting with the hardest task - do not point the agent at Christmas returns on day one. Start with sorting incoming orders. The flow is clear, the volume is high and the proof of value is in the inbox within an hour.
  • No approval policy - a shop with no rule set tends to either approve everything by hand (which defeats the point) or let everything run (which breaks on the first exception). Write the approval policy on day one.
  • Treating JTL-Wawi as a black box - give the agent the scopes it needs (orders, products, returns, purchase orders). Half-connected agents produce half-done work and raise more questions than they answer.
  • Letting product data in a free-text field - send the agent the supplier Excel, the category tree and the SKU logic. Without these, you get generic text and the wrong category.
  • Ignoring the audit log - the log is your evidence for consumer-protection audits and for the EU AI Act transparency rules. Open it once a week in the first month.

6 tips for your AI agent in JTL

1. Pick one channel first, then widen

Start with JTL-Shop orders only, add Amazon in week two, Kaufland in week three. Each new channel brings its own quirks (Amazon FBA, Kaufland handling fees), and adding them one by one keeps the agent accurate.

2. Put approvals in Teams or Slack, not in a new tool

Approvals should land where your team already works. Superkind sends the exception as a card in the channel the responsible person uses, no new login, no new inbox.

3. Write rules the way you would brief a new colleague

“B2B orders over EUR 3,500: wait for me. Returns under EUR 150: run. Price exceptions over 5 % against the price list: wait for me.” Short, concrete sentences work better than scoring models nobody maintains.

4. Keep JTL as the source of truth

The agent should read and write in JTL-Wawi, not run a parallel database. If you ever disconnect the agent, your JTL shop keeps running exactly as before.

5. Make data protection a conscious choice

The DSGVO and the EU AI Act (Regulation (EU) 2024/1689) require that you know where your AI runs and that your team is trained to use it. An EU-hosted agent like Superkind, with a signed DPA from the Trust Center, settles both in one conversation with your DPO.

6. Watch the Juno beta but do not wait for it

JTL’s own agent, Juno, launches in open beta in November 2026 and will cover reorders and customer service at launch. Enrol in the beta, and keep running Superkind on the full back office in parallel. The two cover different jobs and work side by side.

Conclusion

Setting up an AI agent in JTL does not mean handing your shop to a bot. It means the reading, booking and chasing that eats your morning is done for you, your rules are applied every time, and only the handful of calls that need a person land as a card in Teams, Slack or Outlook. The fastest way to see whether it fits your shop: open app.superkind.ai, connect JTL-Wawi, and give the first task “sort my orders” before lunch. If you want more context first, read our guide on AI for online shops or on how to qualify leads with AI for the B2B side.

Frequently Asked Questions

Yes. JTL announced Juno, its own AI agent for the ERP, as a closed beta at JTL Connect on 2 October 2026 with an open beta in November 2026. External AI agents like Superkind connect to JTL-Wawi today via the JTL API, through the SQL database or via file exchange, and work with every version. You do not have to wait for Juno to free up time on orders, returns and product data.

The tasks that cost real hours every week: book incoming orders across channels, write product descriptions and push them to all shops, process returns and credit notes, chase supplier confirmations, update stock, answer standard customer tickets with order data, and build weekly reports from JTL data. Signing contracts, deciding on price exceptions and talking to key accounts stay with your team.

Via its 1,500+ integrations, which cover JTL-Wawi, JTL-Shop, Amazon, Kaufland, eBay, Shopify, DATEV, lexoffice, DHL, DPD, Outlook and Teams. The admin confirms which areas Superkind may read and write (orders, products, returns, purchase orders), and Superkind uses the JTL API where it exists and SQL views or file exchange where it does not. No new middleware, no custom code.

The way you set it up. The default pattern that works for most JTL merchants: order confirmations under a certain value and returns under EUR 150 run on their own; everything above waits for a human OK in Teams, Slack or Outlook. You can tighten or loosen that rule any time, and the audit log records who approved what.

JTL data contains customer names, addresses, order history and prices and must stay inside Europe. Superkind is hosted and processed in Europe, is GDPR compliant and aligned with the EU AI Act, does not train on your data and offers a DPA (AVV) through its Trust Center. Access rules let you decide who sees which part of the shop and the ERP.

With Superkind you start for free at app.superkind.ai with a one-off up to EUR 150 in credits, no credit card and no sales call. The Business plan starts at EUR 50 per month for the whole company with unlimited users; a typical JTL merchant works through orders, returns and product data for a few hundred credits per day. EUR 100 covers up to 1,000 tasks or 5,000 questions. Stand: October 2026.

No. JTL stays your source of truth for products, stock, prices and orders. The AI agent reads and writes in JTL on your behalf, the same way a back-office clerk would, so you keep your JTL setup and skills. The agent is on top of JTL, not instead of it.

Related Articles

Henri Jung, Co-founder at Superkind
Henri Jung

Co-founder of Superkind, the AI employee teams write to in Outlook, Teams and Slack. Henri helps companies, from small teams to the Mittelstand, hand their routine work to AI - without new tools, consultants or an in-house AI team.

Send Superkind your first JTL task.

Start for free in Outlook, Teams or Slack. Or let us show you in 30 minutes how it runs on your JTL-Wawi.