Credit: Dreamstime
Generative AI, or GenAI, is the hot technology of the moment, gaining traction with both enterprises and individual users.
Bloomberg Intelligence research in a June 2023 report says the generative AI market is poised to expand dramatically over the next 10 years, rising to $1.3 trillion from a market size of $40 billion last year. The report says market growth is being driven by training infrastructure in the short term, and this will gradually shift to inference devices for large language models (LLMs), digital ads, and specialised software and services in the longer term.
A June 2023 report by consulting firm KPMG notes that generative AI technology is advancing at a rapid pace. The firm surveyed 200 US executives in June 2023, and 75 percent said generative AI will be a top-three emerging technology over the next 12 to 18 months.
Generative AI use cases include chatbots and virtual assistants; content development; data analytics; design and development; and predictive maintenance. We spoke to three CTOs about how they are leveraging the technology in their organisations and what they expect looking ahead.
Copilots for innovation and efficiency
Avanade, a joint venture created by Microsoft and Accenture in 2000, has collaborated on generative AI since Microsoft first launched a series of GPT services in preview with OpenAI in early 2021, says company CTO Florin Rotar.
For example, Avanade leveraged the Copilot technology Microsoft introduced to its Microsoft Graph and Microsoft 365 applications to integrate LLMs with its own data to increase productivity.
Avanade is also one of the first users of Microsoft Viva Sales, which it’s using to streamline its sales processes, Rotar says, and is embedding the new releases of Viva Sales that include generative AI as Microsoft brings these new capabilities to market.
“Bigger picture, we are seeing GenAI being the inflection point for AI becoming democratised, providing new opportunities to unleash innovation and efficiency benefits,” Rotar says. The company is seeing benefits such as increased productivity, and is using generative AI across its sales and marketing operations to discover insights and improve sales account planning, he says.
“As we’re actively working with clients across the world, we’re seeing organisations realise additional benefits,” Rotar says. For example, a manufacturer has identified opportunities to use generative AI to inspire designers, engineers, and marketing teams to create and deliver new product designs. A nonprofit organisation is exploring the use of generative AI to help generate grant reporting and redirect time spent on related administrative tasks.
Rapid prototyping and product delivery
Anaqua, a provider of intellectual property (IP) management software, is using generative AI for internal operations to support development and in its products to support a growing list of use cases, says Erik Reeves, CTO.
The company has been using and following AI, machine learning, and advanced search for years, “and some of the new capabilities emerging now are simply different in their level of quality, without having to invest a huge amount of effort,” Reeves says. “These new services offer extremely valuable benefits for rapid prototyping, testing, and market validation in a much shorter timeframe, giving us the ability to try a number of things [and] validate, kill, or accelerate projects.”
At the same time, the company recognises that it needs to be practical and apply these tools intelligently to solve customers' real challenges, Reeves says.
“Some of our early initiatives are quite simple, but impactful nonetheless,” he says, such as looking for ways to reduce manual and redundant work. “We are also pursuing more advanced use cases that apply in more sophisticated and nuanced ways to business and data challenges in IP, but we're maintaining a very flexible and practical approach to delivery,” he says.
Generative AI “is a formative area for us as we look at options to leverage this new [type] of technology,” Reeves says.
“While we are a ways from having AI produce an application for us, especially in a complex area like ours, we already have leveraged it for simple tasks that can help consistency, efficiency, and quality in development. I think what’s critical for us here is to establish a learning community and spread that knowledge, just like we do with basic coding practices, security, and operational mechanics of development.”
Where Anaqua has seen immediate benefits is more ad hoc, for areas such as code review, code comparison, or asking specific questions for validation or feedback. “What we get back isn’t always going to be something you take to the bank, but there is a shockingly high amount of value to be gained already—and it will only get better,” Reeves says.
Code generation is the tip of the iceberg
Quick advances in generative AI technologies have led to rising interest and progress in code generation tools, according to Marktechpost Media, an AI news platform. These tools use machine learning algorithms and natural language processing (NLP) to help developers automate some aspects of coding.
“AI-generated coding enables developers to work on more creative and fulfilling tasks,” Rotar says. “By spending less time and effort on the more mundane aspects of coding, developers are freed up to focus on discovery and innovation, coming up with new ways to use programs, apps, or coding. The AI-generated concepts can also offer opportunities for net-new code solutions or even coding languages that don’t exist today, which may solve for current or future challenges.”
Software company SAS is exploring code generation and flow generation capabilities with co-pilot integrations, and currently deploying marketing content generation using ChatGPT, says SAS Executive Vice President and CTO Bryan Harris.
“Looking to the future, we are testing the integration of a chatbot into our AI platform to allow customers to interact with our tools and their data using natural language,” Harris says. “We are infusing our [products] with this functionality, so our customers can realise true value” from generative AI faster.
For example, financial planners might seek ways to better maintain their employers’ margins, Harris says. “Our approach to [generative] AI helps explain the data, point out anomalies detected by another AI model, and automate their normalisation,” he says.
“Similarly, a logistics planner might ask our agent to review options to optimise supply chain costs by comparing supplier bids against margin targets. This is a precise interaction that we don't yet see in the public domain.”
What's ahead for generative AI
While there is no official or widely agreed-upon definition of what technologies fall under the generative AI umbrella, SAS considers digital twins, synthetic data generation, and LLMs, all to be generative in nature.
“Our software already uses many fundamental AI capabilities, such as reinforcement learning, to generate synthetic data,” Harris says. “In this system, we have a ground truth source, a model that randomly generates tabular data, and a discriminator. The discriminator attempts to determine whether the generated data is plausible or false and returns feedback to the model.”
This is the “breeding ground” for building synthetic digital twins, Harris says. “For example, we can generate several types of data similar to vehicle telemetry data, and then run ‘what-if’ scenarios to predict the behavior of this complex system,” he says.
“I think what is equally exciting in code generation—and other topics, frankly—is the ability to provide a ‘smart assistant’ to a developer; something that can [in] real-time provide quality control, make suggestions, and help enforce consistency,” Reeves says. “Some mundane things can be relegated to automation or simple review, while higher-order design thinking and user experience can become more central to the creative exercise.”
The CTO's role in GenAI adoption
Given the hype in the market, CTOs need to be vigilant about setting expectations about generative AI within the C-suite and with the managers and teams working under them.
Part of this is making it clear that generative AI is still relatively new in terms of business use cases, and needs to be deployed with some level of caution.
“It’s tricky to keep up with the pace of the market,” Rotar says. “One year from now it will look very different. It’s tough to manage the hype and keep pace. The key is ensuring that leaders are experimenting, learning, and also working on what their AI strategy will be.”
While many leaders have ideas for how to leverage generative AI, Rotar says, they need to consider risk management, compliance, regulation, security, and ethics. “That means leaders need to consider more than the technology implications of AI,” he says. “To harness the benefits of AI, leaders need to assess and monitor multiple business and IT domains to maintain the AI readiness of their organisations and people.”
While businesses are eager to experiment with generative tools, “we must be mindful that current applications are experimental at best,” Harris says. “The content generated by [generative] AI-based solutions is the result of sourcing data and artifacts created by humans—and humans are prone to inserting bias, making mistakes, and contradicting themselves.”
SAS customers work with sensitive data, Harris says, which is a primary reason for the company to take a cautious approach to generative AI. “Simply asking a question through ChatGPT may result in the disclosure of confidential information that could be used to retrain the underlying model in a way the customer didn’t intend,” he says.
The novelty of generative AI can mask potential pitfalls and result in collateral damage, Harris says. “Users tend to be overly trusting of automated programs, and individuals may not question generative outputs, then make ill-conceived decisions based on misinformation, fake content, or ambiguous statements promoted by the algorithms,” he says.
“These oversights could have serious ramifications if they spread to a live production environment where the results could affect the real world.”
While generative AI shows immense promise, “without checks and balances securely in place there is also great potential for misuse,” Harris says. “These lapses could result in harmful deepfakes, copyright infringements, misappropriation of intellectual property, and other improper outcomes.”
Asking the right questions
Advances in generative AI are accelerating faster than either governing bodies or society have had time to reasonably address, and issues of business value, risk, and ethics have yet to be reconciled, Harris says.
“Our business model does not promote pushing the latest technology just for the sake of its novelty,” Harris says.
“Our many years of experience have instilled the importance of taking a closer look at when, where and how best to apply new methodologies by examining both their strengths and weaknesses. Our commitment to data ethics forces us to answer the question, ‘if we can, does it mean we should?’ If yes, to what degree?”
