Journey as AI Engineers: from Quella to agentic AI
Helani and Meenu share what it means to move from AI experiments to connected assistants that support real business workflows.
Inside the work.
- 01Start with a useful problem
- 02Connect AI to trusted systems
- 03Keep people in control
CHAPTER 01
Learning begins with a real customer need
For Helani and Meenu, the journey as AI engineers has grown through products with a practical job to do. Quella, Fexcon's conversational shopping assistant for WishQue, is one example: customers need help discovering products, understanding delivery options and checking orders without searching across separate pages.
A useful assistant therefore needs more than fluent conversation. It must recognise intent, retrieve the right business information and know when the available evidence is not enough.
CHAPTER 02
The difficult work sits around the model
Applied AI engineering includes the systems that make an answer timely and relevant. APIs, retrieval, access rules, evaluation and fallback behaviour all influence the customer experience. A model can only act on live information when those connections are designed carefully.
Quella's development has involved linking conversational behaviour with commerce data and workflows. That work turns a general AI capability into an assistant with a defined role inside a real product.
CHAPTER 03
Moving toward agentic workflows
Agentic AI introduces the ability to select tools and complete steps toward a goal. It also raises the standard for boundaries and oversight. Reading an order status, recommending a product and changing a customer record carry different levels of responsibility.
The engineering task is to define what the system may do, what needs confirmation and how a person can take over. Evaluation needs to include ambiguous requests, missing data and failed tools alongside successful demonstrations.
CHAPTER 04
A discipline shaped by curiosity and evidence
Working in AI means learning continuously, but progress is easier to assess when it is tied to user outcomes. The team can compare whether people find information faster, whether routine enquiries are resolved and whether escalation works when the assistant reaches its limits.
Helani and Meenu's journey reflects Fexcon's wider approach: explore new technical possibilities, connect them to the systems a business already uses and improve them through evidence from real interactions.
The next stage of AI engineering will be shaped by assistants that are useful, connected and accountable in the workflows they support.