In the competitive world of Quick Service Restaurants (QSR), the ‘secret sauce’ isn’t always in the kitchen. For El Jannah, the iconic Australian charcoal chicken chain, it’s increasingly found in the cloud.
Since launching its AI pilot in September 2025, El Jannah has transformed how it interacts with its” community. By integrating Salesforce Agentforce into WhatsApp, the brand has moved beyond simple automated replies to a sophisticated, autonomous service model that handles the heavy lifting of digital enquiries.
The digital overhaul:
Working with Xenai Digital, El Jannah integrated WhatsApp directly into its Salesforce Service Cloud environment. This wasn’t just about adding a new chat window; it was about meeting customers where they already live.
The impact was immediate:
- 7,300+ total digital conversations managed since launch.
- 75% of WhatsApp enquiries (over 1,100 cases) resolved entirely by AI.
- 20% of all chat-based cases migrated to WhatsApp, significantly reducing the strain on traditional phone support.
“Customers expect fast answers on messaging platforms they use every day,” says Tyler Mason, El Jannah’s Chief Technology Officer. “AI handles the routine, ensuring our human team is ready when a conversation truly matters.”
Guardrails and grounding
A common fear with AI is the hallucination — the tendency for models to confidently state incorrect information. El Jannah mitigated this through a rigorous grounding process. Mason comments, ‘We had a lot of learnings when we first launched Agentforce. What helped us stay within the guardrails was a thorough QA process before publishing. We also loaded a large set of pre-approved FAQs from the outset to ensure responses were accurate and on-brand”.
Before the AI went live, the team loaded a massive set of pre-approved FAQs and structured knowledge articles. This ensures the AI stays within the guardrails” of the brand’s specific menu nuances and store locations.
During the pilot, the team noticed inconsistent responses when the AI tried to pull from multiple knowledge articles at once. The solution? Re-structuring the data into formats the agent could more clearly interpret. It’s a process Mason likens to ‘onboarding a new team member—you don’t just hire them and walk away; you train, monitor, and refine.
The team continuously monitors hand-offs to human agents, using those insights to refine future AI responses. “We continuously monitor hand-offs to human agents and use these cases to build out more information for future interactions. The reality is that not all questions can be answered by AI, which is why we maintain a dedicated support team seven days a week ready to step in when needed”.
A test-and-learn approach
During the pilot phase, the system initially struggled with some queries that referenced multiple knowledge articles, often providing oversimplified or inconsistent answers. “LLM-powered chat agents are not purely deterministic”, advises Mason. “They can pass a test one day and fail the next, as they are dynamic and learning in nature. Early in the pilot we saw inconsistent responses, particularly when the agent referenced multiple knowledge articles on the same topic and produced different answers”.
“Consistency improved significantly once the knowledge sources were re-structured into formats the agent could clearly interpret. This highlights an important lesson: deploying AI agents may be relatively quick, but achieving reliable performance requires ongoing training and optimisation after go-live. It’s similar to training a new team member, you don’t stop after onboarding and expect them to be perfect”.
A unified guest view
One of the most powerful features of the deployment is its integration with Salesforce Data Cloud. When a customer messages El Jannah on WhatsApp, the system doesn’t see a stranger; it sees a unified customer profile.
- Real-time sync: Every interaction is immediately attached to the customer’s existing profile.
- Contextual support: If a customer previously ordered via the app or had an issue in-store, the AI has that history at its fingertips to provide a tailored response.
- Human hand-off: While escalations are becoming rare, the system uses sentiment-based triggers to alert human agents the moment a conversation requires a personal touch.
“All interactions with Agentforce (via WhatsApp or our website) are immediately identified and attached to the customer’s existing profile inside Salesforce. This allows us to provide tailored responses based on the guest’s profile and interaction history”, says Mason.
From support to strategy
The 75% resolution rate is just the beginning. El Jannah and Xenai Digital are already scoping transactional AI capabilities.
Future updates may include:
- Automated Vouchers: If the system detects a delivery has exceeded a 40-minute threshold, it could automatically trigger a “we’re sorry” voucher via WhatsApp.
- Direct Ordering: Moving beyond support to allow guests to place their favorite charcoal chicken orders directly within the chat interface.
As El Jannah continues its national expansion, this AI-first approach ensures that whether they have 50 stores or 200, the guest experience remains as consistent and authentic as their Granville roots.
“AI will continue to play a central role in our guest experience strategy. As we open new locations, it’s important that we can scale customer interactions while maintaining responsiveness and consistency”.
“Insights from these conversations help us understand common customer needs and friction points, which can inform service improvements and targeted communications. Our focus is on making interactions easier and more relevant, not more intrusive, so any personalisation will prioritise usefulness and simplicity for the guest”, concludes Mason.
