Delivering Smarter Customer Search with AI Integration

Our client has amassed a wealth of valuable behavioural data based on user demographics and usage patterns. Although this data holds enormous potential for discovering new business opportunities, it remained buried inside a set of large data repositories, which were challenging to interrogate. This meant that the marketing and member engagement teams were unable to efficiently access this valuable business intelligence.

Customer

Our client is a leading provider of intelligence services to the construction industry. They offer a comprehensive digital platform that supports the needs of construction firms, professionals, and interns. The platform enables users to access information on over one million construction projects across the UK.

Problem Statement

To help users find projects relevant to their interests, the platform includes a powerful search engine equipped with numerous filters. These filters can be combined in various ways, allowing for highly tailored search results. However, this complexity often creates a barrier. Users must understand how the filters work and remember how to apply specific combinations, a process that typically requires prior experience or training.

As digital expectations evolve and AI becomes more widespread, such a steep learning curve is increasingly viewed as a potential risk to user satisfaction and business growth. Our client recognised this challenge and sought to explore how AI could simplify the search process, making it more intuitive and accessible.

Data Potential Realization

User-Friendly Search Experience

LLM and ChatGPT

Natural Language Query

Solution Development

Arrk partnered with the client to explore the feasibility of using natural language processing (NLP) to transform the search experience. The goal was to allow users to input natural language queries—such as “show me commercial developments starting in Manchester next quarter”, and receive relevant results without needing to configure multiple filters.

We assessed several NLP models, including GPT-4o Mini, Anthropic’s Claude, Meta’s LLaMA, and Mistral. Our approach combined both technical and commercial evaluation, considering not just performance and integration but also the operational costs of deploying AI at scale for a high-traffic application.

Over the course of six months, we moved beyond traditional chatbot or AI assistant solutions. We designed a tailored approach that accounted for the specific user behaviours, business context, and platform architecture. Our team brought together deep domain expertise, platform familiarity, and real-world experience of embedding AI into live environments.

Requirement Collection

Structured Query Conversion

Data Repository Integration

Outcomes

Developed an AI-enhanced search capability that enables natural language queries, reducing reliance on manual filters.
Significantly improved user experience for both seasoned professionals and new users unfamiliar with advanced filtering techniques.
Reduced onboarding time and support queries related to search functionality.
Helped de-risk digital complexity and positioned the platform for future AI-led enhancements.
Demonstrated a scalable approach to AI integration, with consideration for platform load, cost of inference, and long-term maintainability.
Provided the client with a forward-looking, differentiated digital offering that reflects current user expectations while being grounded in practical application.