.jpg)
Despite the understandable skepticism, generative AI will continue to be weaved into the business fabric and redefine ways of working. Data will get better at painting clearer pictures, playing more instruments, creating new medicines, developing more complex apps, and recommending when, where, and how to launch a product, to name a few. As generative AI permeates the enterprises, business leaders will need to better understand the investment, expertise, and responsibility required to bring its potential benefits and value into fruition.
In this first part of our series on 2024 tech and analytics trends and predictions, we tackle the impact and value of generative AI to organizations, and what business leaders and decision-makers need to rise high above its surge and ride its wave with confidence.
Investing in niche, “brand-specific” AI
The C-suite is increasingly betting on AI to augment their human capabilities, but few are only managing to reap its benefits. More than 40% of surveyed leaders are struggling to demonstrate ROI from each AI model, while almost half of their company's AI implementations never move to production. That’s poised to change with generative AI, which is projected to not just help with IT operations — it will also boost productivity and creative problem-solving by 50% in 2024.
Lingaro Senior Data Scientist and AI Advisor Maciej Michalek remarked that advancements in large language models (LLMs) — combined with the company’s data — will underpin many business processes and workflows. LLMs are projected to be operationalized to automate the customization, sales, and delivery of products and services, which surveyed analytics leaders will prioritize in the next years. In fact, 10% of operational processes will utilize LLM-powered digital assistants in 2024. Nearly 40% of surveyed enterprises revealed, too, that they're already investing in their own LLM.
Using domain- and use case-specific generative AI
What does this mean for businesses? These customized, generative AI-powered enterprise-ready solutions, for example, will enable marketers to scale personalized branding and marketing campaigns. In fact, Forrester predicts that the top agencies will spend US$50 million in external partnerships for these AI models in 2024. ChatGPT Enterprise was just the beginning: Adobe, Google, Meta, and Microsoft are inking deals with these agencies to create “brand-specific” AI models. Salesforce also unveiled plans to incorporate a GPT-based chatbot in its customer relationship management (CRM) platform.
With these in mind, Michalek said that while advancements are underway, current hardware and computing limitations will compel a more pragmatic approach to AI. A single NVIDIA processor for a small-scale generative AI-based application already costs at least US$40,000 in a disruption-prone supply chain. It also consumes up to 700 watts of power, so decision-makers need to be modest. Some are pushing through with smaller, more practical domain-specific models with clearly defined purpose and context. Google, for instance, is working on two models for organizations specifically in healthcare and life sciences. They have another for cybersecurity, while Databricks has its own for retail companies. Bloomberg, too, is developing its own that caters to financial institutions.
These constraints emphasize the importance for enterprises to temper unrealistic expectations and set achievable ambitions, like prioritizing solutions that bring the most ROI. The Lingaro DS&AI practice, for instance, organizes thought leadership and use case hunting sessions that enable decision-makers to better understand how generative AI can enrich how they work with their organization’s data. Lingaro AI advisors like Michalek also provide actionable road maps from which enterprises can base their AI strategy on, whether for automating routine tasks, optimizing daily operations, or improving business decision-making.
Leveling up with multimodal generative AI
Generative AI has become 2023’s technological buzzword. In 2024, further technological strides will be made toward artificial general intelligence (AGI) due to multimodal generative AI — but what does this mean for businesses?
Imagine a system that doesn’t just read data, but also interprets and visualizes texts, sees images, hears sounds, and watches videos like humans do. It then brings them all together and creates its own unique version, in different formats and form factor across multiple touchpoints. Multimodal generative AI brings this to life. Lingaro DS&AI Practice AI Engineering Competency Leader Norbert Fijałek doesn’t see this as an isolated happenstance. This was actually born from the need to streamline and maximize existing investments in AI and other digital technologies due to financial and operational burdens.
Advancements in generative AI capable of consolidating multiple sources and types of data are already available and can be utilized by enterprises. OpenAI released GPT-4V, which has image and vision capabilities and can interpret advanced images and schemas, such as high-level architecture diagrams. Google, too, has been working on multimodal generative AI that can be applied to digital marketplaces by making it easier to upload, search for, and sell products.
Multimodal generative AI can also be used by vehicle makers to automatically inspect car defects by analyzing both videos and images. In consumer industries, it can be used to analyze sensors, cameras, and audio recordings to translate and predict consumer sentiments and behaviors. Meta’s ImageBind — designed to mimic human perception — coalesces six different types of data to generate, for example, virtual worlds that can be used for e-learning or live retail/livestream shopping. By 2026, IDC predicts that multimodal AI will used in at least By 2026, IDC predicts that multimodal AI will used in at least 30% of AI models.
What’s in it for businesses now? As companies drown in a deluge of unstructured data — be it in the form of PDFs, videos, PowerPoint slides, or information from social networks — multimodal generative AI can help streamline it, enhance the efficiency of its use, and eliminate redundancies. Fijałek notes the ripple effect: generating personalized, customizable content and automating the process for creating them. By 2025, for instance, 40% of engagements in the services industry will embed generative AI-based delivery. By 2024, 40% of enterprise applications will have conversational generative AI capabilities.
Realizing business value in multimodal generative AI
The Lingaro DS&AI practice, too, has already been working with CPG and FMCG companies in adapting and applying multimodal generative AI to different parts of their business. These include streamlining their knowledge management processes by consolidating different types of data and fueling their team’s creative juices by generating unique graphics for their brands.
The emergence of multimodal generative AI technologies, such as GPT-4V, LLaVA, and Kosmos-2, promises to transform enterprises by further enhancing efficiency, fostering innovation, and improving productivity. By handling tasks in both text and image forms, these models diversify task management, improve efficiency, and free up human resources. They offer a holistic understanding of visual and textual data, which can be applied to business areas like product development, customer service, and marketing.
Despite the immense potential, Fijałek cautioned that it’s easy to be wrapped around the hype without realizing any value. Fijałek reiterated that implementing a generative AI-based solution, let alone a multimodal one, requires a solid foundation in data and analytics, particularly data quality and relevance. It entails a workforce who can expertly manage and transform this data, given how unstructured data handled by enterprises is projected to double in 2024.
The tall order is partly why enterprises are increasingly keen on collaborating with other experts. With 50% of leaders and teams expected to fail in bridging internal gaps in their AI initiatives in 2024, tech executives will be driven to rethink and expand on how their external partnerships can provide the alliances and capabilities needed to make the most of their investments in AI. This includes creating ethical as well as security- and privacy-aware AI-powered innovations.
Complying with responsible AI mandates
With generative AI becoming a household name, public consciousness is also waking up to pressing questions: What is this technology doing? What consequences will arise from the decisions and predictions it arrives at?
Lingaro Senior Data Scientist and AI Advisor Taylor van Valkenburg noted that AI’s pervasiveness will compel humans to develop more precise guidelines to ensure that its use aligns with best practices and legal mandates. The recently held AI Safety Summit saw 28 countries, alongside technology providers and academic organizations, committing to AI safety and conducting rigorous protocols for using AI.
In Italy and Poland, there are ongoing investigations and lawsuits about ChatGPT’s violation of the GDPR. Indeed, the EU’s efforts in charting out regulatory frameworks reflects a growing recognition of AI’s socio-technical nature. AI professionals will now need to merge their technical prowess with a broader, more holistic perspective. With the EU AI Act, “high-risk” AI initiatives will demand exhaustive documentation, ensuring transparent, accountable, and compliant development processes. With these stringent measures in place, Forrester predicts that a ChatGPT-powered third-party app or system will be fined in 2024 for mishandling personally identifiable information.
Protecting the bottom line with responsible AI
Responsible AI isn’t just about doing the right thing, van Valkenburg furthered. Businesses have a lot to gain — or lose. Apart from costly penalties imposed by the GDPR, the EU AI Act will require organizations that use AI (generative or not) to provide explainability that reflects the risks of their use cases. Failure to do so will translate into penalties of up to 7% of the company’s global turnover. Companies in the US that overstate their use of AI or generate biased or incorrect results from AI systems may incur sanctions from the Federal Trade Commission.
Conversely, companies that apply trust, risk, and security controls in their AI and advanced analytics applications can significantly improve the accuracy of their decision-making by eliminating 80% of faulty data. They’re also more likely to be able to move their AI projects into production faster.
As public uncertainty and regulatory activity heighten, van Valkenburg added that enterprises will have to overcome more complexities in developing AI-powered innovations and increased costs in compliance. Navigating these intricacies will require new business expertise to ensure adherence to industry standards and legal requirements, internally and in interactions with clients and customers. Lingaro’s DS&AI practice, as a Microsoft Gold Partner, adopts the Microsoft Responsible AI Standard when using or integrating AI in Microsoft-powered solutions. The practice works with global CPG companies in conducting impact assessments to understand the effect of their AI solution to their people. They also perform early user testing to identify failures in human interaction with AI systems and understand how they’ll behave.
Indeed, generative AI in the enterprise is inevitable. As organizations forge ahead, those that proactively embrace a holistic, measured, and values-driven approach will not only be more successful at using it, but also in safeguarding their company’s reputation, improving its bottom line, and achieving market leadership. Their efforts will be set apart by the astuteness with which their leaders navigate the challenges of responsible implementation and tangible value realization. In 2024 and beyond, AI’s immense potential will only be matched by visionary leadership, robust governance, and strategic alliances guided by a conscientious, accountable human compass.
Subscribe to our roundup of expert insights into upcoming trends and paradigm shifts in data science, AI, and machine learning that will help enterprises better utilize these innovations.
