- Explore our industry expertise
- Success stories
- A smart assistant that manages information as a strategic asset
A smart assistant that manages information as a strategic asset
Flexible, secure architecture
The smart assistant leverages a robust Google Cloud architecture to provide secure and fast access to comprehensive company information.
Reduced search time
The AI-powered solution significantly reduces the time spent searching for relevant information, enabling employees to work more efficiently and make well-informed decisions.
Innovative personalization
By offering tailored recommendations and automated responses, the solution delivers a personalized user experience and boosts productivity.
“The smart assistant heralds a new approach to innovation at GFT. With our deep pool of GenAI expertise we have been able to build a solution that addresses a universal market challenge.”
Challenge
As part of its outreach program GFT continually gathers feedback from its diverse global client base. When clients experience universal pain points, GFT strives to deliver innovative solutions using its proven fast-prototyping methodologies that leverage AI large language models (LLMs), retrieval augmented generation systems and multi-agent system to improve data quality and deliver better outcomes.
Many large companies are grappling with information overload on a wide range of internal subjects. This is most obvious when new joiners need to gather information on a company’s processes and procedures. As part of the onboarding process, new recruits may face the daunting challenge of having to consult numerous information sources in diverse silos. The problem is most acute in global corporations that have grown through acquisition where individual companies may have different cultures, business procedures, manuals and technology stacks.
The information challenge is also present in everyday work: for example a lawyer who must search for a specific clause in a 200-page contract can waste up to 60 minutes . Research suggests that poor decisions made due to an incorrect understanding of information may amount to $10 billion per year.
On the positive side, IBM reports** that automation and chatbots can handle up to 30% of help desk queries, saving IT support teams significant time, while Zendesk reports that*** organizations using chatbots experience a 60% reduction in ticket handling time.
With a deep understanding of this universal searching challenge, GFT committed to build an AI-powered smart assistant using its proven echnologies. The proposed LLM solution would interact with documents to gather information and enable a faster, deeper understanding that previously possible.
“Human comprehension of a high-volume documents invariably takes a long time leading to reduced productivity, increased demand for support and heightened business risk. Over time, this leads to knowledge silos with an organization.”
Engagement
GFT has worked with AI and LLMs since their inception and has delivered several chatbots that harness the power of natural language processing to deliver accurate answers quickly. The aim of this project was to empower users to surf through potentially any internal knowledge base using a natural language interaction tool.
The specific objective was to enhance human capabilities by facilitating a real-time conversation direct with any information source – typically long and often complex documents. Although the number of use cases is unlimited, immediate applications could include regulatory compliance in financial services or more general data rights compliance, such as GDPR in Europe. In addition to saving time, the right solution would deliver more accurate, consistent results and improve decision making.
To ensure a scalable and reliable deployment the smart assistant is built on Google Cloud Platform (GCP). As well as delivering a secure infrastructure as a service, GCP provides a set of proven management tools that are continually updated.
Other unique architectural features of the smart assistant include:
- RAG architecture. For an LLM to be effective and offer relevant results it must leverage custom data from multiple sources. The smart assistant uses an adaptive retrieval augmented generation architecture to integrate external sources, deliver an enhanced contextualized understanding and maximize response accuracy.
- Personalized user experience. To reduce time and improve accuracy of results the smart assistant delivers tailored responses based on an individual user profile and interaction history.
- State of the art LLM. AI is evolving at a blistering place. Designed with flexibility in mind, the system can use many of the proprietary LLMs and can switch between them if needed to facilitate fast, accurate responses.
- Authentication and API access. Many organizations are migrating to an open, composable architecture facilitated by application program interfaces (APIs). The assistant maximizes the use of APIs to allow easy integration with any target infrastructure.
The smart assistant is designed to combine scalability with flexibility. As a fully API-enabled solution, it offers a strategic solution to the ongoing challenge of information searching and retrieval. New models or components can be integrated easily with no disruption to business as usual.
“With a broad client base GFT gains unique insight into common business challenges. All organizations must search a burgeoning volume of complex business-critical data from many sources. Our smart assistant uses AI and LLMs to offer a flexible solution to this universal business need.”
Benefit
The new platform offers an optimal combination of an LLM and GFT’s own data retrieval models. As a high-performance model trained with bespoke data it can generate results that are more tailored and granular than is possible with a single LLM.
With its open architecture, the model can be easily trained using new, specialist data on any subject. In practice the model acts as a ‘processing brain’ that can be easily populated, and results can be shared.
This model offers an important advancement for any company without a centraliaed data repository. By sharing search results a company can improve collaboration and develop best practice methods to address common challenges. As well saving time and money, the smart assistant leads to better outcomes and continual improvement.
The solution can also be modified to answer question in a specific style, for example the way it answers legal queries can be varied from answers to customer service queries. The model is fully multi-lingual and will answer questions in the language in which they were asked.
Although it can be deployed in any sector, the smart assistant is most applicable in markets characterized by high volumes of complex data. Trials have been conducted in financial services and outsourcing and the platform is continually enhanced in line with user requirements.
* McKinsey & Company. The Social Economy: Unlocking Value and Productivity through Social Technologies – Executive Summary. Retrieved from https://www.mckinsey.com.
** IBM. Unlocking the Power of Chatbots: Key Benefits for Businesses and Customers. Retrieved from https://www.ibm.com.
*** Zendesk. Official Website. Retrieved from https://www.zendesk.com/it/.
