How Generative AI will Move into the Enterprise

By Greg Bucko, Data Consultant

How Generative AI will Move into the Enterprise

Artificial Intelligence: Understanding the risks and rewards

In the ever-evolving landscape of technology, artificial intelligence (AI) stands as a beacon of potential and promise. With its rapid advancements and integrations, AI is reshaping the way businesses operate, strategize, and innovate. As we delve deeper into the world of Generative AI and its implications for the corporate realm, it’s evident that we’re on the cusp of a transformative era.

Generative AI may be reaching maximum hype right now, and business leaders are frantically scrambling to understand how this technology can be applied at scale. On the surface, Generative AI tools such as ChatGPT seem almost magical, creating a sense that with enough time and data they can do almost anything. But business leaders need more than just magic chatbots, they need to understand how this technology can generate that magic in ways that either drive revenue, lower costs, or preferably, both. In today’s blog post we focus on moving past the hyperbole around generative AI and discuss what enterprise-wide adoption at enterprise scale could look like and the very real limitations of generative AI that could pose challenges to this adoption.

Understanding Generative AI

Generative AI, a subset of artificial intelligence, has quickly become part of the everyday lives of millions. Whether it’s content creation, data analysis, or customer service, Generative AI is making an impact. The mainstream awareness of AI’s capabilities has surged, putting organizations in a unique and somewhat difficult situation. On one hand, organizations are smitten with the promise of automation and greater efficiencies for knowledge workers, and on the other hand they are (and should be) concerned about the risks.

Watch our Presentation on Demystifying Generative AI

Risks of Generative AI for businesses

Although the benefits are great, these risks cannot be ignored and can be summarized into three risk groups (note that I’m not touching on either existential or economic risk, which are both subjects far too speculative for consideration here).

  1. Accuracy Risk: The problem of Generative AI tools such as ChatGPT making mistakes is already well documented. This is one of the main reasons I encourage people to use these tools in their personal lives, to get to know both the promise and the peril of what these tools can produce before using these tools in the workplace where the stakes are often much higher. These tools make mistakes because their job is to predict the next word or phrase in a sentence based on a distribution of possible answers, with some amount of that distribution containing contextual inaccuracies. Since ChatGPT doesn’t fact check in real time, it will always give you some sort of answer. It’s the user’s job to make sure that this is accurate.
  2. Security Risk: The size and shear compute power required to run these Generative AI models means that, as of right now, they are hosted virtually by incredibly well-funded companies such as OpenAI, Google, and Anthropic. On top of that, the chat itself is used to further fine-tune the model, which means anything that you put into ChatGPT, for example, will end-up on the servers of a third party. Although you can turn this feature off in ChatGPT, according to OpenAI1 your data will still remain on their servers for 30 days to monitor for abuse. This creates a clear vulnerability for sensitive data such as personally identifiable customer, client, and employee data, not to mention exposing trade secrets.
  3. Legal and Compliance Risks: Courts will be working overtime in the next few years to set the precedents that will dictate how we engage with Generative AI in ways that respect copywrite holders, protect privacy, and comply with global regulations. Governing bodies in the US, EU, and elsewhere are drafting legislation as fast as they can to square this incredible technology with consumer protections such as the GDPR in Europe and the CCPA in California. The lack of a proper regulatory framework for Generative AI makes it a real headache for business leaders (and their attorneys) concerned about not running afoul of the law.

How Generative AI will work its way into organizations

There are four ways in which Generative AI will begin to integrate itself into enterprises at scale: through basic office software, enterprise-level tools, embeddings, and increased AI awareness. This is, of course, in addition to how ChatGPT has already made its way into businesses in a grassroots way. However, as organizations clamp down on ChatGPT use specifically, these four will become a more scalable way to bring Generative AI into the business. Let’s break each one down, and discuss some of the benefits and issues organizations will face:

  1. Basic office software: The ubiquity of Microsoft Office makes it one of the most obvious places where Generative AI will make its presence felt. Whether you are the CEO or an intern, you are using Word, Excel, Outlook, and PowerPoint to some degree every day (or for some, their Google equivalents). So, when both Microsoft and Google announced major integrations of Generative AI into their tools, it was clear that aside from ChatGPT, this would be the world’s real introduction to the technology. The promise of this software according to Microsoft and Google is incredible, with marketing materials that show these Generative AI assistants writing emails, producing marketing briefs, analyzing data, and ultimately creating a presentation complete with charts, graphs, and imagery. But these tools, at least at first, will come with a high price-tag and limited access. For example, in July, Microsoft announced it was expanding its beta testing to 600 companies, while at the same time announcing the cost of copilot at $30 per Office 365 seat, essentially doubling the price of an Office 365 enterprise license today. In addition to cost considerations, the same risks mentioned earlier will still apply…and the costs of hallucination errors will now be much higher. With even the best large language models in the world generating these errors on a regular basis, it remains perhaps Generative AI’s biggest challenge to adoption. If we embrace these tools as our assistants, then we will set the same level of expectations we would for a human assistant we were paying. And as of today, a human assistant who consistently produced mistakes and outright falsehoods as frequently as ChatGPT wouldn’t be long for the job. We will need to hold our AI chatbots and tools to the same level of accountability.
  2. Enterprise level tools: Tools that are designed to serve a particular business function such as accounting software, CRMs, ERPs, and HR tools, will also be on the frontier of AI integration. Major software-as-a-service (SaaS) providers are bending over backwards to roll out AI-related features or at the very least convince people it’s on their roadmap. While some of the larger players such as Salesforce (Einstein) and IBM (Watson) have been working on their own AI solutions for several years, most niche SaaS players will undoubtably rely on APIs with 3rd party language models such as GPT-4 to provide, like thousands of mobile apps that have emerged in 2023, the ability to say that your application is “powered by AI” or something to that effect. Regardless of how these SaaS providers introduce AI into their tools, one thing is for sure…this is where a large proportion of global marketing dollars will be spent in promoting AI. In the young history of the SaaS industry, we’ve already seen countless trends and buzzwords that make great ad copy but are somewhat limited in their application (Hadoop anyone?). It remains to be seen whether adding new AI features will drive greater and greater revenue for SaaS providers, but it won’t be for a lack of trying.
  3. Enterprise embeddings: This refers to the customization of large language models (LLMs) using specific enterprise information. By fine-tuning foundational LLMs such as GPT-4 with internal corporate data such as emails, press releases, earnings reports, employee handbooks, presentations, and more, businesses can create a tailored ChatGPT-like experience that not only has the communication skills of a large language model, but is also deeply familiar with the nuances, terminology, and intricacies of their specific enterprise. The use cases for this are obvious. A language model that could bring together disparate sources of information across a business could be used to improve strategy, find cost savings, identify new markets, and much more. Assuming, of course, that the corporate data being used to fine-tune the model is itself accurate, non-biased, and comprehensive, which could be a major challenge for some organizations. For example, will fine-tuning a language model with twenty years of annual strategy decks produce a better, more creative strategy or just an average strategy that fits the distribution of the historic data? Will hallucinations of corporate history occur, and if so, at what frequency and what checks and balances should be put forth? These are the kinds of questions organizations will need to be ask themselves and understand prior to creating their own corporate chatbot.
  4. Project-oriented AI: One benefit of the current AI hype cycle is that it once again raises awareness about the importance of data innovation in organizations. Although Generative AI currently owns the headlines, much more specific AI applications that are good at one thing (sometimes referred to as “Narrow AI”) should also experience increased demand. Examples of these types of very specific applications of AI are everywhere today, with tools such as spellcheck, adaptive cruise control, and smart bulbs being just a few of the many examples.

Harnessing AI’s business potential

What the hype around Generative AI has already begun to do is to raise awareness among business leaders that AI can solve problems, optimize processes, and do things that humans don’t want to or simply cannot do. This increased awareness will result in more and more AI applications, regardless of whether they are of the generative language variety or not. This increased demand is at its infancy, with McKinsey reporting that 40% of companies will be increasing their investment in AI across the board due to the advances in Generative AI specifically. However, that number will likely grow as the same report highlights that only one-third of companies said they were using AI in more than one department. No doubt that functional leaders from across thousands of organizations are thinking about how AI could be used within their specific function, something that would not have been the case in a pre-ChatGPT world.

The integration of Generative AI, specifically large language models, into business operations presents real risks including accuracy, privacy, and legal risks. However, the potential upside of this technology cannot be overstated. The ability to streamline processes, provide instant access to vast amounts of information, and enhance decision-making capabilities can revolutionize the way businesses operate. Although the allure of increased efficiency and innovation will be irresistible to many companies, it’s critical that they also be prepared to address the challenges of Generative AI head-on. This includes investing in training and awareness programs for employees, establishing robust data governance frameworks, and continuously monitoring and refining AI implementations to ensure they align with organizational goals and values. At the end of the day, the organizations that will truly benefit from Generative AI will be those that recognize its power, understand its limitations, and are strategic about its deployment.

At Keiter Technologies, we are highly experienced in helping our clients harness the power of AI and advanced analytics. If you’re interested in learning more about how we can help your business maximize its impact through Generative AI, contact our Keiter Technologies team today. We are here to provide innovative data solutions to help your business improve processes and drive growth.


Sources:

Share this Insight:

About the Author


Greg Bucko

Greg Bucko, Data Consultant

Greg is a highly experienced analytics professional who has helped lead data strategy in a variety of industries including hospitality, insurance, retail, agriculture, and financial services. He is a data consultant at Keiter Innovative Data Solutions and provides thought leadership and advisement on matters of enterprise data and analytics strategy.

He is a passionate leader focused on helping organizations understand their customers, markets, and products better through the use of advanced data analytics and consumer insights. Greg specializes in building and leading data analytics teams in organizations that are transitioning to a more data-driven, customer-centric business strategy. He is also skilled in delivering complex analysis in a way that is clear, concise, and easy to apply directly to corporate strategy.

More Insights from Greg Bucko

The information contained within this article is provided for informational purposes only and is current as of the date published. Online readers are advised not to act upon this information without seeking the service of a professional accountant, as this article is not a substitute for obtaining accounting, tax, or financial advice from a professional accountant.

Categories

Contact Us