CHALLENGE

Peak season pressure when demand outruns customer support

As the leading online flowers and gifting company in the Middle East and the UK, Floward’s business is built around delivering gifts such as flowers, cakes, chocolates, and perfumes quickly and reliably, especially during high‑demand periods. Floward also partners with local and international brands to offer a wide range of curated gifts. Same‑day delivery is a core part of its value proposition, ensuring that gifts arrive on time for important occasions like Mother’s Day, Valentine’s Day, birthdays, anniversaries, and seasonal celebrations.

This model creates intense seasonality. Key occasions such as Valentine’s Day and Mother’s Day concentrate on a large share of annual demand in just a few days. On those peaks, customer conversations can surge up to 13 times the volume of an average day, as senders and recipients ask questions, update delivery details, and track orders in real time.

To keep up, Floward had been scaling its customer service team to around 500 agents during peak periods. While this ensured coverage, it was increasingly unsustainable and costly to sustain such short windows of time. At the same time, customer journeys were fragmented. Customers might provide delivery addresses through a chatbot, seek answers to FAQs in a separate chat, and still need a human agent for anything outside a narrow script. 

During peak days, these friction points made it harder to protect same-day delivery. If a customer couldn’t complete their journey smoothly or if address details weren’t collected quickly, deliveries were at risk of delay. Floward needed a way to manage peak demand more efficiently, reducing reliance on temporary teams while guiding customers through multi-step journeys in a single, consistent experience. 

SOLUTION

A unified platform for AI-powered customer service

Building on an existing partnership with Infobip, Floward expanded the relationship to tackle a new challenge: scaling customer service for extreme peak periods. Together, we moved beyond traditional rule-based chatbots toward agentic AI by implementing AgentOS, Infobip’s AI-native unified platform, as the backbone of the new approach. Floward now uses intelligent agents that understand context, make decisions, and adapt to customer needs in real time. AgentOS orchestrates the end-to-end journey, from the first utility message on WhatsApp and guided address collection to AI-powered assistance and seamless handoff to live agents who receive the full context of the customer’s case, when needed. The goal: handle extreme peak-day volumes without relying on hundreds of additional agents, while keeping customers informed and orders on track for same-day delivery.

Infobip’s AI consultants worked closely with Floward’s team to map existing customer journeys, identify friction points such as address collection and fragmented FAQs, and design new workflows that bring together AI agents, customer data, and WhatsApp Flows. This collaboration enabled Floward to design, test, and deploy the new solution in under two months, fast enough to be ready before a major peak period.

Three customer journeys in one intelligent system

Floward’s customer service now runs three core journeys, all orchestrated through AgentOS and connected to Floward’s CRM.

• Gift senders: Customers who place an order through the app, website, or WhatsApp receive automated confirmations and status updates over WhatsApp. AgentOS uses customer data to recognize them as senders, surface the right order, and answer questions about delivery times, changes, or payment, reducing the volume of routine inquiries reaching human agents.
• Gift recipients: When a sender places an order, Floward uses AgentOS’s customer engagement and automation capabilities to automatically send a utility message on WhatsApp to the recipient, notifying them that a gift is on the way and asking for delivery details. This step is critical for same-day delivery. With AgentOS orchestrating the journey, Floward can ensure address collection happens quickly and without confusion, even when volumes spike.
• New customers and general inquiries: People messaging Floward’s WhatsApp number without an existing order are handled by an AI knowledge agent. Using a structured knowledge base, the agent answers common questions about products, delivery areas, pricing, and payment methods, resolving many inquiries without manual routing.

Across all three journeys, AgentOS integrates with Floward’s CRM, identifying whether a customer is a sender, recipient, or new contact, and ensuring seamless routing. This context-aware routing reduces back-and-forth, cuts down on duplicate conversations, and helps keep agents focused on complex cases instead of basic triage.

Orchestrating the utility message, address collection, and AI assistance

For each order, a single orchestrated flow now replaces the fragmented experience Floward had before.

When an order is placed, AgentOS sends a utility message via WhatsApp to the recipient, informing them about the upcoming delivery and prompting them to share their address details. WhatsApp Flows then guide the recipient through structured address collection within WhatsApp, including location, building details, and delivery instructions. This reduces the risk of incomplete or unclear information, which previously caused delays and threatened same-day delivery.

If, instead of following the prompts, the recipient starts asking questions such as “Who sent this?” or “When will it arrive?”, an AI agent steps in within the same conversation. It answers questions using Floward’s knowledge base while keeping the primary goal of collecting the address at the center of the flow. The customer does not have to switch channels or start over. 

If the AI agent detects that the issue is outside the predefined scope (for example, a complex change to an order or a sensitive complaint), the conversation is handed to a human agent in AgentOS’s cloud contact center. The agent receives the full context of the conversation, including any data collected so far, avoiding repetition and speeding up resolution.

Behind the scenes, Floward runs a multi‑agent setup on AgentOS, where one coordination agent routes each conversation to the right specialized AI agent for tasks such as address collection, FAQs, or order changes. This multi‑agent orchestration helps keep journeys efficient and consistent, even when volumes surge.

Using AI to resolve routine queries and reduce agent load

Behind these journeys, Floward uses AI agents powered by retrieval‑augmented generation (RAG), which draw on extensive Floward documentation to answer FAQs accurately and in the right context. For dynamic information such as delivery timelines, address details, and order status, the agents pull real‑time data from Floward’s CRM, ensuring customers always receive up‑to‑date answers. As a result, most routine questions are resolved automatically, reducing the number of conversations that need human intervention.

Real‑time analytics and tailored workflows ensure a smooth customer experience and give Floward the insights needed to adjust staffing and improve containment. For example, by identifying the times of day when address‑collection drop‑offs were highest, Floward was able to fine‑tune WhatsApp Flows and staffing plans ahead of peak periods.

By using AI agents, customer data, WhatsApp Flows, and live agent escalation together in one platform, Floward has turned previously fragmented journeys into a single, coherent system designed to handle peak-day spikes without sacrificing service quality. 

Extending AI-powered journeys to e-invitations

Beyond customer service, Floward has started using the same AgentOS foundations to support new, revenue-generating experiences. One example is its e‑invitations feature inside the Floward app, which allows customers to create personalized invitations for occasions, including event details and RSVP options.

The approval workflow behind these invitations, ensuring message content and images are reviewed before being sent over WhatsApp, is orchestrated through AgentOS. Once recipients respond, automated journeys connect engagement with commerce: guests who accept an invitation are prompted to choose a gift for the host, while those who decline are offered alternative gifting options.

This extension shows how Floward is taking the capabilities initially implemented to solve peak-day support and address collection challenges and applying them to new journeys that drive both engagement and incremental revenue.

RESULTS

45% fewer peak agents while handling 13x more conversations and 15% cost reduction 

The shift to agentic AI on AgentOS transformed how Floward manages peak demand. By combining AI automation with its existing team. Floward decreased agent dependency by 45% while handling 13x more conversations and reducing customer service costs by 15%. On Valentine’s Day alone, Floward handled 54K customer conversations. Despite having fewer agents, Floward maintained service quality, with agents now focusing on complex, high‑value cases instead of routine inquiries.

Customer experience metrics also improved significantly. Following the implementation of Agentic AI, Floward increased CSAT by 12 percentage points year over year, alongside a 4% uplift in customer response rates.

Containment rate improvement

Floward’s key metric was containment rate, the percentage of customer issues resolved by AI without human escalation. The previous AI assistant achieved limited containment; too many customers were handed to agents. With AgentOS, containment rates improved by 17 percentage points year over year on average. This means more customers receive instant answers, fewer wait for human support, and agents can focus on higher-value interactions.

Consistent response times at peak

Even under peak pressure, Floward has maintained reliable service levels. The contact center now sustains an average response time of one minute with 95% SLA achievement, including during major events such as Valentine’s Day. This level of consistency helps protect Floward’s same‑day delivery promise, even when volumes spike well beyond normal levels.

What Floward is saying