What successes have you experienced with Generative AI?

We recently implemented Generative AI in our content marketing workflows to auto-generate product descriptions and customer-facing FAQs. Initially, we viewed it as an experiment, but it ended up reducing our content creation time by over 60% while improving engagement metrics.

Data played a huge role — we trained the model using anonymized customer queries and sales data, which helped fine-tune outputs to match real-world language and priorities. The biggest challenge was ensuring output quality and brand alignment, which required a robust human-in-the-loop review process.

The major benefits included faster turnaround, better personalization, and freeing up creative teams to focus on strategy rather than repetitive writing tasks. It’s been an exciting shift!

Yes, absolutely! During the Smart Recruit project, we definitely had some moments where the generative AI provided surprisingly insightful outcomes.

One instance that stands out was when we used the role-based prompts on the collated survey data. We had asked the AI to assume the perspective of a "frustrated hiring manager" describing their pain points with the old spreadsheet system. The AI not only summarized the obvious issues like manual data entry but also highlighted a subtle but critical point about the lack of visibility into the candidate pipeline for the wider team. This wasn't something explicitly stated by many in those exact words, but the AI, in that specific role, synthesized the underlying sentiment from various responses in a way we hadn't immediately seen. It pointed to a real need for shared dashboards and real-time updates in the new system, which became a key feature. The data from the surveys was the foundation, but the AI, through the "hiring manager" lens, brought a new level of interpretation.

A challenge we definitely faced was the iterative process of prompt engineering. Getting the AI to provide truly actionable insights required a lot of trial and error. Our initial prompts were often too general, and the responses lacked the specificity we needed for feature decisions. It wasn't a plug-and-play solution; we had to learn how to "speak" to the AI effectively, and the role-based prompts were a significant breakthrough in that regard.

The benefits we realised were substantial. The AI significantly accelerated the analysis of the large volume of survey data. What would have taken us days of manual collation and interpretation? The AI could have processed it in a fraction of the time. More importantly, as illustrated by the "frustrated hiring manager" example, it provided novel perspectives and highlighted connections within the data that we might have overlooked. This ultimately led to a more robust and user-centric design for the Smart Recruit system.