Returns, order status and delivery changes, finished in one conversation.
AI agents that identify the customer and act across your commerce, order and service tools, on chat, voice and WhatsApp. They complete the request instead of logging it.
Book a demo- Live for one retailer in four weeks, acting across 40+ tools
- 82% resolved without a person in the loop, on beta
- Your team keeps the exceptions, with the full history
Book a proof of concept
Two hours with your leadership to start. A working agent on your data in four weeks.
14% of customer issues are fully resolved in self-service today.
Gartner, 202415.8% of retail sales are returned: $850bn a year in the US.
NRF and Happy Returns, 20258% of first delivery attempts fail, at $17.20 each.
Loqate, 202130% of service cases are handled by AI today. Teams expect 50% by 2027.
Salesforce, 2025What changes in customer service.
- Requests are completed, not logged
- Returns, cancellations and delivery changes are carried out in the conversation, across every system they touch.
- One conversation on every channel
- Chat, voice and messaging share the same skills and the same customer history.
- People handle the exceptions
- Your team takes the cases that need judgement, with the order, the history and the conversation in front of them.
Five workflows we deploy for retailers.
Each is a named workflow with an owner and a target, not a general assistant.
Customer service
Order status, returns and cancellations
Identifies the customer, finds the order and completes the request: a return label, a refund, a cancellation or a changed address. It works on chat, voice and WhatsApp.
of customer issues are fully resolved in self-service
Gartner, 2024
of incoming e-commerce support requests ask where an order is
Gorgias, 2023
Delivery and services
Delivery, assembly and service scheduling
Books and moves delivery, assembly and repair appointments against real capacity, and tells customers about a delay with a new time they can accept or change in the same message.
of first delivery attempts fail in the US, at $17.20 each
Loqate, 2021
of online shoppers had a delivery problem within three months
Descartes, 2025
Sales
Product advice before the purchase
Answers product, stock and compatibility questions from your catalogue, and hands ready buyers to checkout or to a store.
conversion among online shoppers who use Lowe’s AI assistant
Lowe’s, 2025
of consumers want generative AI in their shopping
Capgemini Research Institute, 2025
Stores
Store staff assistant
Gives floor staff stock, policy and product answers in seconds, from the same sources head office uses, so they do not leave the customer to find out.
more issues resolved per hour by newer support staff with an AI assistant
NBER, 2023
to plan a shift with Walmart’s AI task tool, down from 90 minutes
Walmart, 2025
Back office
Supplier and invoice queries
Answers suppliers’ payment and delivery questions from your ERP, matches invoices to orders, and sends disputes to the right person in finance.
per invoice for the best finance teams, against $9.90 on average
Ardent Partners, 2026
of invoices need manual handling, against 11.8% for the best
Ardent Partners, 2026
The figures are published research and industry benchmarks, not results from our own deployments. The Lowe’s and Walmart figures are those companies’ own reports, and shoppers who choose to use an assistant may already be more likely to buy. Sources:
- Gartner, Customer service survey, via destinationCRM (2024)
- NRF and Happy Returns, 2025 Retail Returns Landscape (2025)
- Loqate, Fixing Failed Deliveries (2021)
- Salesforce, State of Service, 7th edition (2025)
- Gorgias, What’s the Secret to Reducing WISMO Requests? (2023)
- Descartes, Home Delivery Consumer Sentiment Study (2025)
- Lowe’s, Q3 2025 earnings call, via CX Dive (2025)
- Capgemini Research Institute, What Matters to Today’s Consumer 2025 (2025)
- NBER, Generative AI at Work (2023)
- Walmart, New AI-powered tools for 1.5 million associates (2025)
- Ardent Partners, Accounts Payable Metrics that Matter in 2026 (2026)
Proof: IKEA.
A retailer’s customer service agent, taken from an FAQ chatbot to acting across 40+ tools, then to voice.
- 1Explore
- A chatbot answering frequently asked questions from a knowledge base.
- 1 week
- 2Deploy
- Customer identification and integration with 40+ tools: the agent acts on the customer’s behalf, from returns to cancellations.
- 2 weeks
- 3Expand into voice
- The same skills extended to voice, proving the platform is not bound to one channel.
- 1 week
live from kickoff
resolved without a person in the loop, on beta
The platform underneath.
- Inside your systems
- Agents read and act in your commerce platform, order management, delivery, payments and contact-centre tools. We connect them during deployment; nothing is replaced.
- One company context
- Customers, orders, products and policies in one governed layer, so every channel and every agent answers from the same record.
- Governed by you
- Access follows your SSO, sensitive actions wait for a person, personal data is masked before any model call, and every step is logged.
- Any model, hosted where you choose
- Frontier, open-weight or on-premise models, in our cloud, your cloud or on your own hardware.
Live in four weeks.
Our engineers and a product lead work inside your environment until the first workflow pays, then hand it to your teams.
- Two hours with your leadership
- We map where agents pay in your business and agree the first workflow.
- A business case in ten days
- An opportunity map, the numbers for the first workflow, and a deployment proposal.
- A working agent in four weeks
- A returns and order-status agent, built on your own systems and data by our engineers, with your team alongside.
- Your team takes over
- Your people run what is live and build the next workflow; we review, then step back.
Questions from retail teams.
It acts. It identifies the customer and completes returns, cancellations and delivery changes in your commerce and order systems. For one retailer it works across more than 40 tools.
Next step
Start with returns.
Two hours with your leadership. Ten days to a business case. Four weeks to a working agent on your data. No commitment.