Resolved more than 300,000 customer emails
Maintained 90%+ customer satisfaction
Built working prototypes in hours rather than weeks
Delivered responses to customers in minutes instead of 48 hours
Smyths Toys uses Google Cloud AI agents to automate customer service and cut responses times from days to minutes.
Smyths Toys uses Google Cloud AI agents to automate customer service and cut responses times from days to minutes.
Seeking to improve speed, reliability, and customer satisfaction
As Europe's largest specialty toy retailer, Smyths Toys serves millions of families across Ireland, the UK and continental Europe through hundreds of stores and a rapidly growing ecommerce business. Roughly 70% of the Irish-founded company's annual revenue is generated during the final three months of the year, making speed, reliability, and customer satisfaction mission-critical during its busiest period.
That seasonal success also created one of the company's biggest operational challenges. Every holiday season, Smyths' customer service organization was inundated with thousands of customer inquiries, the vast majority revolving around a single question: "Where's my order?" Although approximately 70% of inquiries were simple order-tracking requests, each still required a member of the customer service team to manually retrieve information from multiple systems before responding. During peak periods, customers often waited as long as 48 hours for an email response.
I am loyal to solutions, not platforms. What mattered was the partnership. Google Cloud brought engineers who wanted to solve problems alongside us, and that made the difference.
Adam Sexton
Head of AI, Smyths Toys
"We wanted to prove what AI could actually do for the business," says Adam Sexton, Head of AI at Smyths Toys. "Customer service looked like a simple proof of concept because everyone immediately understands the problem. In reality, it's one of the hardest places to start, but it gave us a measurable business case."
Building an AI-powered customer service platform on Google Cloud
For Smyths, selecting Google Cloud involved more than technology alone. “I am loyal to solutions, not platforms,” says Sexton. “What mattered was the partnership. Google Cloud brought engineers who wanted to solve problems alongside us, and that made the difference. I chose Google Cloud because I believed they could help us learn quickly and build something real, not because their technology stack is untouchable. I made a people and technology choice.”
The function-calling capabilities of Gemini allow Codie to collect exactly the data it needs, when it needs it. Instead of relying on static information, the agents pull real-time business data directly into the response.
Deepika Ramachandra
SAP Business AI Developer, Smyths Toys
The release of the Google Cloud ABAP SDK dramatically accelerated development by allowing the team to build working prototypes in hours rather than weeks using live SAP data. That speed transformed AI from an interesting concept into a practical business initiative.
The company launched its proof of concept—Codie—in January 2025 with a narrowly focused objective: automatically answer "Where's my order?" inquiries by connecting to shipping providers and payment platforms.
Codie consists of two complementary environments. Business-critical SAP processes run through an ABAP implementation powered by the Google Cloud ABAP SDK, while more computationally intensive AI workloads execute within a Python environment running on Kubernetes using Google's Agent Development Kit (ADK). This architecture allows Smyths to route complex AI tasks to infrastructure better suited for large-scale inference while preserving the performance of its SAP transactional systems.
Each customer interaction begins with an intelligent classification stage. Leveraging models accessible through Gemini Enterprise Agent Platform, incoming requests are analyzed before being assigned to a supervisor agent responsible for planning how the enquiry should be resolved. The supervisor then delegates work to specialized sub-agents, each responsible for completing a specific task. Those agents can invoke additional retrieval agents to access enterprise knowledge or execute function calls against operational systems.
Through Gemini's native function-calling capabilities, Codie securely retrieves live information from SAP S/4HANA, carrier APIs such as Royal Mail, DPD, and DHL, and payment platforms including Stripe and Worldpay. If an action requires issuing a refund or retrieving shipment status, the framework can complete those activities autonomously before generating a customer response.
Smyths also uses different Gemini models depending on the stage of orchestration, selecting Pro or Flash models according to the complexity of the task. “The function-calling capabilities of Gemini allow Codie to collect exactly the data it needs, when it needs it,” explains Deepika Ramachandra, SAP Business AI Developer at Smyths Toys. “Instead of relying on static information, the agents pull real-time business data directly into the response.”
Transforming service during peak season and giving customers answers in minutes
By the end of 2025, Codie had expanded far beyond order-tracking inquiries to support customer interactions across email, web, and social media. During peak shopping periods, the platform handled approximately 60% of the entire customer service workload, while maintaining customer satisfaction above 90%.
Codie Web intercepts approximately 60% of customer enquiries before they ever become emails, while Codie Email automatically resolves around 90% of the remaining email volume. Instead of beginning January with thousands of unanswered inquiries—as had historically been the case—Smyths entered 2026 with zero email backlog, while another market that had not yet deployed Codie still faced approximately 5,000 outstanding messages.
The operational benefits extended well beyond customer experience. Over six months, Smyths:
- Resolved more than 300,000 customer emails
- Processed 10,000 emails for approximately $1
- Saved more than $1.3 million in temporary staffing costs
- Eliminated the need for the traditional seasonal hiring surge while maintaining 90%+ customer satisfaction
Perhaps most importantly, customers who previously waited up to 48 hours for a response now frequently receive answers within minutes, while website visitors often receive responses almost instantly through Codie Web.
Additionally, Smyths deliberately designed Codie to be managed by business users rather than IT. Through an SAP Fiori application called Bot Manager, customer service teams can update prompts, manage AI behavior, and continuously improve individual agents without developer intervention. As a result, Smyths positioned AI as a way to eliminate repetitive work while enabling employees to become prompt engineers and AI supervisors. The company's permanent customer service team remained intact while AI absorbed work that previously required seasonal hiring.
Our ambition was never simply to automate customer service. Our ambition was to capture the knowledge of our people and build a framework that allows AI to work alongside them across the entire organization.
Adam Sexton
Head of AI, Smyths Toys
Smyths is now extending the same agentic framework into its commercial buying organization, where AI agents will proactively analyze purchasing decisions, supplier performance, inventory levels, and merchandising opportunities. Rather than waiting for buyers to ask questions, Codie will proactively recommend better purchasing decisions—for example, suggesting an alternative supplier capable of delivering inventory earlier or identifying opportunities to improve stock availability and inventory turnover.
To support this, Smyths is combining enterprise data from SAP with additional datasets in BigQuery while continuing to leverage Google Cloud's scalable AI infrastructure. As the complexity of workloads grows, the distributed architecture spanning SAP and Kubernetes will enable the company to analyze significantly larger datasets without impacting mission-critical transactional systems.
For Smyths Toys, Codie has evolved from a single customer service proof of concept into an enterprise AI platform capable of orchestrating a digital workforce across the business. "Our ambition was never simply to automate customer service," says Sexton. "Our ambition was to capture the knowledge of our people and build a framework that allows AI to work alongside them across the entire organization."
Smyths Toys is an Irish toy retailer that specializes in children’s toys, games, and entertainment products with more than 300 shops across western and central Europe.
Industry: Retail
Location: Ireland
Products: BigQuery, Gemini Enterprise, Agent Development Kit
