New Way Now: Smyths Toys delivers agentic customer support to meet peak Christmas demand
Read the Portuguese translation →
Author: Google Cloud · Watch on YouTube
Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.
Original language: English
Original video: New Way Now: Smyths Toys delivers agentic customer support to meet peak Christmas demand · Channel: Google Cloud · 03:11 · Watch on YouTube
Topics: AI agents, Enterprise AI applications, Data, databases and embeddings, AI infrastructure and hardware
Summary
Section titled “Summary”Smyths Toys describes its goal of building a digital workforce to help manage customer support, especially during the Christmas peak. Its Codie system includes tools for email, social media, and website support, with non-technical business users able to work with the solution through a bot manager. The company says Codie Web handles about 60% of customer queries, while Codie Email processes 90% of incoming emails, reducing the human team’s workload. Smyths uses Google Cloud technologies, BigQuery, and real-time order-tracking data to provide customer responses, and reports a zero email backlog on January 1, 2026.
Key points
Section titled “Key points”- Smyths Toys describes itself as a multinational Irish retailer with a significant web business and a larger retail business.
- Christmas is the company’s peak season; it says 70% of its revenue is done in three months of the year.
- The company’s goal was to build a digital workforce.
- A bot manager allows non-technical business users to work with Smyths’ AI solution, Codie.
- Codie has customer-service tools for email, social media, and the company website.
- Codie Web uses Kubernetes and runs on Google ADK.
- Smyths says Codie Web takes about 60% of customer queries, while Codie Email processes 90% of the emails it receives.
- For 2026, the company reported having zero emails backlogged on January 1.
- BigQuery and data agents provide relevant data, including real-time order-tracking information, to ground customer responses.
Chapters
Section titled “Chapters”| Start | Chapter | Summary |
|---|---|---|
| 00:00 | Smyths Toys and its retail business | The video introduces Smyths Toys and describes its retail focus, multinational presence, and seasonal revenue pattern. |
| 01:05 | Introducing Codie | Smyths explains how its bot manager enables business users to work with Codie across email, social media, and web support. |
| 01:35 | Handling customer queries at scale | The company outlines Codie Web and Codie Email’s role in handling queries and reducing the human team’s workload. |
| 02:05 | Data, agents, and real-time answers | Smyths describes its SAP and MCP setup, BigQuery data, Gemini, and the use of live order-tracking information. |
| 02:39 | Grounding responses and Google support | The video explains how tracking data informs customer responses and closes with Smyths’ praise for Google’s support. |
Quotes
Section titled “Quotes”The ultimate goal for us was to build a digital workforce. — 00:33
Codie for customer service has three flavors. — 01:05
Codie Web takes about 60% of the load of customer queries, which is huge. — 01:35
For 2026, we had zero. — 02:05
The real value add from Google is the people we interact with and the support that we get. — 03:11
Concepts and entities
Section titled “Concepts and entities”Smyths Toys, Codie, Codie Web, bot manager, Google ADK, Kubernetes, ABAP SDK, SAP, MCP catalog, BigQuery, Gemini, Google Cloud
Full transcript (original language)
Section titled “Full transcript (original language)”00:00 Hey, let’s play. Having an AI strategy is mostly the realization that when you’re in retail, you are a heavily technology-focused business. Smyths Toys are the largest specialty toy retailer in the world.
00:33 Smyths Toys is an Irish company. As a brand it’s quite iconic, especially in Ireland and in the UK. Now we are multinational. The web side of the business is quite significant, but the retail is still bigger because the children want that toy experience. They want to go into the store. For us, peak is Christmas season. Roughly 30% of our revenue is done in nine months the year, with the other 70% done in the other three. The ultimate goal for us was to build a digital workforce.
01:05 With that in mind, we actually built something which we call the bot manager. And we then allow the end users, and these are non-technical business users, to work with our AI solution, which we have now called Codie. Codie for customer service has three flavors. We have it for emails. We have a bot for social media. And the other one is Codie Web, which sits in the website and that is directly customer-facing. And Codie Web utilizes Kubernetes and it runs on Google ADK.
01:35 Codie Web takes about 60% of the load of customer queries, which is huge. You realize that 600 of them are now being intercepted on the website. Then what we have is Codie Email. That picks up those 400 emails that do come through. That’s processing 90%. And then the human only really has to handle about 50 a day. So 1st of January, typically you would be facing a backlog of thousands of emails.
02:05 For 2026, we had zero. We’re using the ABAP SDK and all the agents themselves we build in our app. We’ve got our own sort of SAP, MCP catalog, and that we expose data agents and action agents in the MCP catalog, and business users, in fact, have control of what data points they require. And it’s at a click of a button. And on the ADK side, we have federated some of the relevant data
02:39 we require into BigQuery. And it uses data from BigQuery to answer, to respond to customers in real time. We’ve just moved to Gemini, and the differences in the quality of answers is huge when you do this. And a lot of the time, actually, you can’t tell it was done by an AI model. We have integrated with a lot of tracking companies. It grounds the responses coming out of Gemini by using the data agents we have created that’ll access real-time tracking information for the orders of interest to the customers,
03:11 and it just passes the data back Gemini and then back out to the customer. The real value add from Google is the people we interact with and the support that we get. It feels as if Google kind of are responsible for Codie in many ways. They’ve really helped us get it there, but it’s the people who did that.