As the new year arrived, many businesses that have listened to the gospel of GenAI and rushed in to experiment with it might be forgiven for rethinking their haste, given the somewhat subdued impact early pilots have shown in real-world situations.
Data management and trust in GenAI results have been the main obstacles for many businesses, points out Chris Walker, vice-president for solutions architecture for Asia-Pacific and Japan at Elastic, which provides enterprise search and observability tools.
Many organisations still struggle with vast amounts of data to be used for AI, while “hallucinations” from GenAI often undermine trust, he says.
Yet, businesses do recognise the impact of GenAI, he stresses, adding that cybersecurity is one area where it can help teams to protect, investigate, and respond to threats before damage is done.
“In the coming 12 to 18 months, GenAI also has the potential to provide organisations with greater observability of their IT infrastructure, thus improving resilience and uptime,” he tells Techgoondu, in this year’s first Q&A.
Q: Many organisations have been pushing out GenAI pilots in the past 18 months but many have not moved to production or scaled up. What are the main hurdles?
A: Put simply, data management and trust in GenAI results. When surveyed by Elastic, 99 per cent of 3,200 IT analytics respondents from around the globe recognise the positive impact generative AI (GenAI) can have on their organisation. However, 89 per cent of these respondents also reported that their use of GenAI is being slowed due to a variety of factors.
For starters, organisations struggle with managing vast lakes of data and complex data sets. As GenAI models require high-volume, high-quality data, and diverse data for optimal performance, they need to bring the right data together at the right time to generate context-specific material. For GenAI pilots to move into production, organisations must ensure they have all their data available to draw insights from.
Secondly, businesses are wary of instances where GenAI models produce inaccurate information or are not grounded in enterprise truth. These “hallucinations” can undermine trust in the AI’s outputs and delay projects.
For organisations to confidently move GenAI pilots into production, they must be able to bridge enterprise proprietary data – both structured and unstructured – from multiple sources with public large language models (LLMs).
This is essential for delivering context-driven outputs that improve data confidence. GenAI projects will remain stuck in the pilot stage until business and IT decision-makers can be confident that those projects will produce reliable results.
Q: From speaking to your customers in the region, can you say what have been the easiest wins for GenAI so far? Chatbots?
A: GenAI has certainly proven to be a good fit for chatbots and other types of virtual assistants. Our conversations with customers reflect this trend.
For example, two of our customers in Singapore, who are leading local banks, have built new GenAI-powered chatbots or virtual assistants to boost their individual bank’s productivity across departments and locations. These chatbots enable their employees to query an internal knowledge base, with access to their own information, product pages, and more, to get personalised answers, fast.
Beyond chatbots, GenAI can also help organisations bolster their cybersecurity posture, helping to address cybersecurity labour shortages. That’s because GenAI can detect anomalous behaviour, predict and identify vulnerabilities, and support fast resolution including automated reporting.
Furthermore, conversational search capabilities can improve visibility, analytics, and response speed in security, as well as make learning and training more accessible to junior analysts via natural language.
Overall, our research found that in Singapore, 63 per cent of organisations surveyed are using GenAI. The research also noted that more than 60 per cent of global respondents believe GenAI could drive change in improving customer experiences and engagement.
Q: How do Elastic’s search capabilities help in GenAI?
A: For GenAI to work, organisations need a search engine to bridge the gap between enterprise proprietary data and public LLMs to support grounding.
Such integration provides more context-driven outputs and ultimately improves confidence in GenAI results. Search AI-powered tools and capabilities such as Elastic Vector Search and our Search AI Lake support organisations by integrating their proprietary data with LLMs, enhancing GenAI’s ability to provide contextually accurate and relevant insights.
In addition, Elastic’s retrieval augmented generation (RAG) enables enterprises to search proprietary data sources and provide context that grounds large language models for more accurate, real-time responses.
RAG helps to ensure that an LLM’s responses take into account contextual data – making GenAI responses more reliable and relevant – as well as characteristics that are non-negotiable in industries including healthcare, finance, and education.
Elastic’s semantic search capabilities also enable a deeper understanding of query intent, made possible by using only keywords, resulting in more relevant and accurate search results. The platform’s support for vector search facilitates similarity search and recommendation systems, on top of core functionalities for many AI-driven applications.
Finally, Elastic’s robust data visualisation tools aid in data exploration and pattern identification, which are essential steps in developing effective GenAI models. By empowering users to find answers that drive results in a secure and scalable manner, Elastic helps organisations unlock the full potential of their data through GenAI.
Q: We’ve seen the potential of GenAI and are reminded often of the profound changes approaching with each new GenAI iteration. In the next 12-18 months, realistically, what are a few of the biggest real-world use cases for organisations?
A: Cybersecurity presents one of the greatest opportunities for GenAI to help organisations in the near term. GenAI can bolster an organisation’s security, enabling teams to protect, investigate, and respond to threats before damage is done.
The technology offers security professionals the opportunity to make better, faster decisions with less manual effort by pulling relevant information, best practices, and recommended actions from the security community’s collective intelligence. By aggregating all this data, it also democratises it, further enabling security professionals to analyse massive amounts of information in near real-time.
GenAI could also help threat detection systems identify potential threats based on landscape changes, which security teams might have previously only stumbled upon by accident. Comprehensive contextual information will strengthen security teams by allowing them to quickly understand an attack’s nature and what actions should be taken.
For example, Elastic’s Attack Discovery enables security operations (SOC) teams to analyse security activity to find the most relevant alert details. This enables an LLM to identify and prioritise attacks, taking into consideration severity, risk scores, asset criticality and other factors.
With insightful guidance from the LLM, SOC analysts can concentrate their resources on the most significant threats, saving them the time and effort typically required to identify potential threats among vast quantities of data.
In the coming 12 to 18 months, GenAI also has the potential to provide organisations with greater observability of their IT infrastructure, thus improving resilience and uptime.
Context-aware GenAI and advanced machine learning methods can reduce labour-intensive troubleshooting and streamline triage activities so that teams can focus on innovation and transformation.
With machine learning, an organisation can easily identify trends and detect anomalies in business and operational performance logs and data.
In tandem with fortifying security, GenAI can also be used to mask data for specific internal clearance to improve data privacy and ensure sensitive information remains secure. This helps minimise disruptions and ensures systems remain reliable and operational, ultimately improving overall business continuity.
