How a global consultancy introduced AI data insights

The Client

The client is the custom data analytics arm of a global, US-based consultancy operating in the consumer space. Leveraging the most granular consumer transaction datasets available today, they provide deep insights that help customers gain a competitive edge. 

• Industry: Alternative Data Analytics 

• Headcount: 10,000+

• Region: US-based HQ/Locations globally

The Problem 

Alternative data sources are the client’s superpower, and for good reason. They provide customers valuable insights into this untapped data, particularly within the financial industry. The client had begun building a text-to-SQL tool in-house, an AI-powered solution that allows users to ask data questions in plain English and self-serve answers. The project held huge potential internally and as a value-add to their customers, but building it was challenging in practice. They had a clear vision, but an in-house build wasn’t the right way to realise it. 

Scoping A Solution

The client’s Co-Founder decided to set up pilots with several Data Analysis providers, including those harnessing AI, like Fluent. The client’s team knew what their ideal solution needed to look like. It needed to offer:

• High rates of accuracy, with fast answer times. 

• Easy integration of their data warehouse.

• Wouldn’t limit user conversations to a few exchanges. 

It was this core success criteria that they initially judged each provider on. With that in mind, they kicked off the pilots with a selection of providers. 

Fruitful Feedback Loops

Fluent let the client stress-test the LLM, often asking the same questions to measure consistency. They marked the LLM’s ability to clarify vague queries and the quality of data visualisations. Fluent proved to be a highly suitable solution, and during the weekly feedback sessions held with the Fluent team, the client could request potential additions and tweaks to the product. Additional features scoped out in the pilot included QA reporting and custom visualisations like Marimekko charts. 

The client aimed to roll out the selected AI solution to internal business users. Members of their product, client relations, and data science teams were chosen to test, and each was onboarded in dedicated sessions. 

The client provides untapped insight, seeing things others don’t within their data. Fluent’s question clarification flow and flexibility when handling complex or vaguely worded questions were ideal for this. The Fluent LLM didn’t make too many assumptions with incoming queries, and the user experience was highly contextual so users felt guided by clear guardrails when interacting with the product. 

Ultimately, the combination of product quality, consistent customer support, and a transparent pilot process convinced the client to choose Fluent as their Text-to-SQL solution provider.

Scaling Usage 

Six months after signing, things were running smoothly. The client identified several use cases they could roll out internally, including for an internal consulting team to use in live retail cases. Fluent also proved particularly useful in helping users get up to speed with new industries quickly, allowing for more effective, quicker pitching to prospective clients, too. 

The client scaled up the number of datasets connected to Fluent from 1 to 10 in less than 8 months. The more datasets connected to Fluent, the larger the potential applications for Fluent within the client’s business. As the LLM continued to learn from the client’s datasets, accuracy rates quickly reached target, jumping from ​​73% to 94%, and finally, a 99% accuracy rate was recorded in January 2024.

The Future

For Q1 2024, the goal was to introduce 10 internal users to Fluent. By Q2 2024, 40 employees were using Fluent internally, including a number who had become internal champions of Fluent. A further 10 external users have been onboarded from two of the client’s live clients; both are globally recognised brands in the restaurant and retail sectors. 

Fluent continues to roll out new features for the client. QA reporting, custom data visualisations, and Insights and Breakdown features were introduced. These allow users to understand exactly how answers have been generated and what assumptions have been made and to deepen their analysis with useful recommendations from Fluent's AI. Now able to converse with their data as if it were an on-call human analyst, the client has made Fluent a trusted solution within their arsenal of internal data analysis and value-adds to their clients.  

If some of the pain points mentioned here rang true for your business, let's talk. Request a demo with our team, here.