![Practical advice for IT to get the most out of Generative AI Practical advice for IT to get the most out of Generative AI](https://inaitoday.com/wp-content/uploads/sites/20/2024/07/Joyce-Gordon.jpeg)
Practical advice for IT to get the most out of Generative AI
In the rapidly evolving technology landscape, Generative AI (GenAI) has swiftly transitioned from a mere buzzword to a powerful tool with practical applications. As IT professionals find themselves at the forefront of this technological revolution, it’s crucial to understand how to implement and manage GenAI solutions effectively.
This article delves into actionable insights and expert advice to help IT teams navigate the complexities of GenAI integration.
Identifying Key Value Areas
Jennifer Fleck, Senior Principal at Slalom Consulting, provides a valuable framework for understanding the potential impact of GenAI. She articulates:
“GenAI drives value in three key areas: productivity (doing what you already do but better and faster), differentiation (doing what you already do in a fundamentally different way), and disruption (changing the essence of your business).”
This approach offers IT professionals a strategic lens through which to view GenAI implementation. By categorising potential use cases into these areas, teams can prioritise initiatives that align with organisational goals and maximise return on investment.
For instance, in the realm of productivity, GenAI could be employed to automate routine tasks, freeing up IT staff for more complex problem-solving. Differentiation might involve using GenAI to create more personalised user experiences in software applications. Disruptive applications could lead to entirely new service offerings or business models powered by GenAI capabilities.
IT leaders should thoroughly assess their organisation’s processes, pain points, and aspirations to identify where GenAI can deliver the most significant impact across these three value areas.
The Power of Interdisciplinary Collaboration
Rio Longacre, Managing Director at Slalom Consulting, emphasises the critical importance of cross-functional teamwork in GenAI implementation:
“My advice would be to refrain from throwing a lot of Gen AI experts together and think they’re going to solve everything. They need to be paired with someone who knows how to do prompt engineering or someone who has domain-level expertise in marketing, personalization, and content or whatever the area is you want to improve or apply generative AI to.”
This insight underscores the need for IT professionals to break out of silos and foster collaboration across various disciplines. While technical expertise in AI is undoubtedly crucial, the true potential of GenAI is unlocked when combined with domain-specific knowledge and skilled prompt engineering.
IT teams should actively seek partnerships with business units, subject matter experts, and prompt engineering specialists. This collaborative approach ensures that GenAI solutions are technically sound and finely tuned to address specific business challenges and opportunities.
Tackling the Challenge of AI Hallucinations
One of the primary concerns in GenAI implementation is the occurrence of AI hallucinations – instances where the AI generates plausible but incorrect or nonsensical information. Joyce Gordon (pictured), Head of Generative AI at Amperity, offers valuable insights on this issue:
“AI hallucinations stem from gaps between training data and real-world queries, ambiguous prompts, and the probabilistic nature of language models. While complete elimination is unlikely, IT can mitigate these issues through high-quality data curation, precise prompt engineering, and implementation of feedback loops. The goal is to reduce inaccuracies without overly constraining the model’s generative capabilities, striking a balance between reliability and the potential for novel outputs.”
For IT professionals, this advice translates into a multi-faceted approach to improving GenAI output accuracy. It begins with a rigorous focus on data quality, ensuring that the AI models are trained on relevant, up-to-date, and comprehensive datasets. This may involve implementing robust data governance practices and regular data audits.
Precise prompt engineering is another critical skill that IT teams need to develop or acquire. This involves crafting clear, specific instructions for the AI that reduce ambiguity and guide the model towards producing accurate and relevant outputs.
Furthermore, the implementation of feedback loops is essential. This could involve creating systems that allow users to flag inaccurate outputs, enabling continuous learning and refinement of the AI models. IT teams should also consider implementing human-in-the-loop processes for critical applications, where human experts review and validate AI outputs before they are used in decision-making or customer-facing scenarios.
Building Trust Through Transparency
As GenAI becomes more prevalent in business operations, the issue of trust becomes paramount. Brandon Purcell, VP and Principal Analyst at Forrester, highlights this challenge in his report The State Of Explainable AI, 2024:
“Artificial intelligence suffers from a trust problem that spans stakeholder groups. Enterprises need to work toward minimising this trust gap to realise the full potential value of their AI investments. Fortunately, enterprises have seven levers available to them to build trust in AI systems. One of these levers is transparency, the perception that an AI system is leading to decisions in an open and traceable way.”
For IT professionals, this emphasis on transparency presents both a challenge and an opportunity. It necessitates the development of systems and processes that make AI decision-making more explainable and traceable. This could involve implementing advanced logging systems that track the inputs, processes, and outputs of GenAI systems. It might also require the creation of user-friendly interfaces or dashboards that allow stakeholders to understand how AI-driven decisions are made.
Moreover, IT teams should consider adopting or developing explainable AI (XAI) techniques. These methods aim to make complex AI models more interpretable, allowing humans to understand the reasoning behind AI-generated outputs. This could involve using simpler, more interpretable models where possible or implementing tools that provide post-hoc explanations for more complex models.
There is no AI without data
Mike Edmonds, Sr. Director of AI Strategy, Global Retail and Consumer Goods at Microsoft, agrees there is no AI without data:
“Many global retailers and consumer goods companies that we work with are actively driving forward with opportunities to unlock value over the immediate term leveraging generative AI. The key to realizing value and moving faster than the competition comes down to setting the right data foundation that provides a constant supply of data from a well-managed and highly integrated intelligent data platform. Data inputs and signals are the building blocks that drive intelligent applications and power data-driven decision-making. To unlock value with generative AI use cases, start by setting the right foundation focused on data readiness.”
The effective implementation of GenAI in IT environments requires a holistic approach beyond mere technical proficiency. It demands strategic thinking to identify value areas, cross-functional collaboration to ensure relevance and effectiveness, rigorous measures to improve accuracy and mitigate hallucinations, and a commitment to transparency to build trust. By embracing these principles, IT professionals can position themselves and their organizations at the forefront of the GenAI revolution, driving innovation, efficiency, and competitive advantage.