How to get started with GenAI: Get ahead or get left behind

The colossal step forward came in November 2022, when OpenAI launched ChatGPT, a free and accessible Large Language Model (LLM) chatbot. The combination of LLMs, which are a massive advancement for AI by combining the ability to gather information with the ability to do something useful with it (such as answer questions), and the global buzz created by ChatGPT has had an unprecedented impact that is transforming the societal and economic landscape.

GenAI in practice

Since the launch of ChatGPT, organizations worldwide have frantically worked to explore the potential use cases of GenAI across the public and private sectors. The long list includes:

• Intelligent search – revolutionizing how we navigate vast repositories of information, these solutions help uncover insights, find answers, and discover patterns across diverse datasets, all with a simple query.
• AI-powered customer service assistants – Natural Language Processing is used to make conversation sound as natural as possible, standard queries are resolved quickly and efficiently and waiting and handling times are reduced. Bank of American’s ‘virtual financial assistant’ Erica, for example, helps users access their balance information, transfer money between accounts, and schedule face-to-face meetings. Starbucks’ customer service chatbot, another good example, guides users through the process of creating and paying for a customized order, notifying them when it is ready for pick-up.
• Virtual assistants – solutions that carry out administrative support tasks and prioritize incoming messages and calls can be used across multiple sectors.
• GenAI solutions that generate summaries of reports, customer calls and meetings (as well as listing actions that result from meetings) free up employees to focus their efforts on more productive tasks.
• Fraud detection in financial services and insurance – LLM models are able to detect unusual patterns, transactions and behaviors.
• Predictive maintenance solutions that automatically schedule essential work before problems arise.
• Supply chain optimization – AI solutions can allocate resources, schedule processes, assign people efficiently, predict future demand and identify pinch points along the supply chain.
• Marketing solutions – including recommendation engines and personalized marketing.

Explore more use cases here.

But if 2023 was the year of Proofs of Concept, 2024 must be the year of action. For many organizations, the evolutionary pace of AI built around LLMs is both exciting and terrifying. AI has become a necessity to gain a competitive edge, by helping organizations work better and more efficiently.

Sitting back and waiting to see how this new world develops is not an option. Being left behind now can give your competitors the upper hand, putting your whole organization at risk. It’s better to start the race, even though no one yet knows what the course looks like or where the finish line will be.

Addressing the risks

The challenge is that LLMs and the infrastructure that supports them (such as ever more powerful microchips) are developing so rapidly that the standard approach to project delivery cannot apply. It is deeply unrealistic to invest in multi-year projects; delivery needs to be fast, and intelligently targeted.

Investing in innovative and rapidly evolving technology is inherently risky and in our conversations with clients, we consistently hear three big concerns:

Accuracy
LLMs are not true artificial intelligence – they are essentially a neural network that predicts the next word in a sentence. These models are not infallible, they can get things wrong.

Bias
Google’s recent difficulties with its Gemini offering were a high-profile reminder of the complex ethical questions that LLMS raise. In particular, Generative AI models can amplify existing biases within the data they are trained on, which can lead to discriminatory outcomes.

Data privacy
LLM models have a voracious appetite for data. The risk is that if private information is not protected, the model could inadvertently disclose confidential data as part of its answers.

These are very real issues and organizations are right to be concerned, but the risks should not be a barrier to exploring and adopting LLMs – they are a reminder that while exploration of potential uses of LLMs and AI should be a priority, it must be carefully done. Accuracy and data quality, for example, go hand in hand. If the GenAI is not given the right information, it will produce inaccurate results. High-quality data, robust data architecture and well-designed interfaces for interacting with the data are all essential.

Implementing LLMs – a roadmap

Organizations are feeling pressure to jump on the LLM bandwagon – this is very promising technology that (mostly) works. It’s very likely that in the next few years, many more organizations will be implementing LLMs into their everyday work.

Waiting to see where LLMs take us is arguably more of a risk than exploring the options today. These models can undoubtedly bring competitive advantages but need to be approached with care.

At Nortal we work closely with clients to explore the potential AI use cases within their organization through a multi-step process:

While the right solutions will be unique to each organization, there are three overarching points to remember when implementing GenAI:

With data, start where you feel comfortable
It can help to visualize the data made available to a LLM in terms of comfort levels. For first use of Gen AI solutions, it makes sense to start with the layer of data that you feel most comfortable with, which usually means your own internal, non-sensitive and non-controversial data. Once you feel more comfortable you can move onto models that use the next level, public data. The final stage is models that use internal, restricted and confidential data.

Remember that AI is not a magic bullet
It’s essential that you focus on the problem that needs to be solved, rather than on the approach that will be used to solve it. AI or GenAI may not be the most appropriate solution for addressing the problem at hand.

Focus on value
The best uses cases for GenAI will be unique for each organization, so it’s important to explore use cases within your own context and focus on where AI can add real value.

A leap into the unknown

We are only scratching the surface of the capabilities of LLMs. DALL-E, developed by OpenAI, is already bridging the gap between visual and written data, and we expect the emergence of NSP, next sentence prediction, to move rapidly this year – this is where models are taught to understand the relationship between sentences and as a result, can better understand the overall meaning of text (and as a result should provide even more accurate responses). More and more use cases that are useful in both the private and public sectors are emerging every day.

At Nortal we have implemented GenAI in our own internal processes, so we already have first-hand experience of the journey that many of our clients are undertaking – we are mindful of the risks and limitations of LLMs, as well as the considerable benefits. Our own LLM solution for clients, Nortal Tark, directly address the risks and concerns most frequently raised by organizations – it is installed within the client’s own controlled environment, for example, so they have full control of their data and privacy.

We are also working closely with clients across the public and private sectors in different industries to transform their organizations with AI. Our work for the German Government’s Procurement Office, for example, has taken significant steps in addressing sustainability and efficiency issues in public sector procurement by making relevant documents easily accessible and searchable. This is an excellent example of how AI can enhance strategic decision-making and promote transparency.

The evolution of GenAI promises to be a rollercoaster for organizations around the world – an exciting leap into the unknown. But organizations who are prepared to take a leap of faith have much to gain.

To learn more about Nortal Tark and the possibilities of GenAI, book a demo below.