Taking generative AI from experiments to high-impact production

6 min read


Presented by Capgemini

Generative AI has shown proven benefits for organizations — but where do you start? In this VB Spotlight, experts from Google, Capgemini and VentureBeat share the real-world ROI companies across industries are realizing with gen AI, and actionable insights for implementing it at scale and more.

Register to watch free on-demand now.

Generative AI has been making headlines all year, driving radical business transformation across processes and products. In this VB Spotlight, industry experts share how generative AI can make the best of your organization’s knowledge and data, and why it’s crucial to start moving from experiments to real-world results.

“Right now, boards and C-suites in all companies out there are asking themselves how generative AI will transform the business they live in,” says Rodrigo Rocha, apps and AI global ISV partnerships leader at Google Cloud. “The companies that can address and respond to that question first, and roll out and implement high value-add use cases absolutely have a competitive advantage.”

And companies don’t have the luxury of waiting until the technology is more mature, adds Mark Oost, global offer leader AI, analytics and data science at Capgemini.

“The number one thing executives should know is that if you’re not moving, your competitors will,” Oost says. “Solutions like the ones from Google are very mature already. It’s time to move. But make sure that you tackle the right use cases that bring your company forward. Don’t just do it for the sake of innovation, following the same use cases that everyone is using. Do this at an enterprise scale. Your competitors are already moving, but you can still catch up.”

From experimentation to scaling

The whole space began with a lot of experimentation, Rocha says. What we’re seeing now is that transition between experimentation into focusing on use cases that deliver end customer value.

“It’s less about experimentation and more about discussion of use cases, understanding the impact of those use cases in your customer value chain and the pieces your customer expects of your company,” Rocha says. “Trying to pull that innovation into the business processes to help those customers, transform those conversations from pure experimentation into value-add, which is ultimately what’s going to propel generative AI in the enterprise segment.”

Enterprises need to move from using off-the-shelf AI models for ordinary consumer applications, to building their own business processes, apps and product design, infusing their models with their own data — a move from self-servicing to self-generating processes, and building the business case to show leadership what’s possible.

“What generative AI taught us in the last couple of months is you can very clearly get to successes,” Oost says. “However, now that it adds a lot of value for our clients, we now get questions about data privacy, but also how you’re going to scale up. We’re now moving from an era of big data to an era of big models. You need to start scaling up across your company in a way that preserves privacy, and in a trusted way.”

At Google Cloud, the customer conversation starts on two fronts. First there’s the technical discussion, and crucial questions about the technology itself, including the posture around data sovereignty, data protection, governance provided as a platform and data control.

Alongside that is the conversation about use cases, separating out the pure experiments with no enterprise value, from the front-and-center use cases that unlock business value.

“In these workshops around use cases, we really go down [to] the business processes,” Rocha says. “What are the steps that today are automated and could be made intelligent, or interactive even? That unlocks the incremental benefit to the end customer. It’s a parallel track, an engineering and tech-savvy one, and then one that’s very much related to business processes.”

The hottest use cases in the market

The pharma and financial services industries have dived head-first into the knowledge mining possibilities of generative AI, and have a head start as these sectors are already very conscious about regulations and data privacy. There’s also a lot of movement in retail, particularly around product description generation.

“It’s a way to get marketers in those companies to quickly go from ideation on the product, understanding what the product is all about, to writing full product descriptions that they can later use on their websites, all infused with generative AI,” Rocha says. “That space is also using a lot of image generation for product marketing catalogs.”

Partners like Typeface have developed a solution to support marketers around the world at scale to better portray their products online and ensuring that customers are better informed about the products they’re looking for.

In the human capital management space (HCM) companies like Workday are infusing generative AI into job description creation. Building a robust job description is a managerial task that can take many hours; with the support of generative AI, they can create those far faster and more ethically, with models trained to be sensitive to gender bias, and even point out potential inequalities in previous job descriptions.

Launching a secure and private gen AI solution

Privacy is crucial to build into a generative AI solution right from the start, Oost says. That means infusing models with your own data in a secure way, and ensuring you add guardrails that keep responses on-topic, ethical and responsible.

At Google Cloud, they encourage customers to ask their providers about their data policies, especially around the data used to train the model — data should be responsibly sourced, and the model should include IP protection and IP rights that ensure that there’s no concern around IP being used to train a model. And customers should ask how their own data is used to train models.

In Google’s case, they use a stateless approach, and don’t use customer data to train models; all the questions that customers ask their models are stateless by nature, encrypted in transit, and in the end the whole session is dismantled.

“Ultimately we believe that the customer should be in control of their destiny,” Rocha adds. “We believe in optionality. We work with the customer to ensure that they’re picking the solution or solutions that best fit their needs.”

This is where considerations about data privacy, protection and controls (both in training the model and then serving the inferences and requests) come in when developing an organizational solution. The next decision is commercial versus open source solutions. With commercial offerings, you get data governance tools and protection of your data as part of the service. With open source alternatives, you need to look at data governance and these safeguards yourself.

“Don’t try to do this alone,” Rocha adds. “Bring the rest of the ecosystem. Bring cloud providers like ourselves. GSIs like Capgemini. Have that holistic conversation about your use case, the tradeoffs you can make to get your solution to market faster, and address customers at scale.”

To learn more about the ways generative AI is transforming enterprises, actionable steps toward building a solution that can scale and more, don’t miss this VB Spotlight!

Register now to watch on-demand.


  • How to change the nature of processes from self-servicing to self-generating
  • How to leverage pre-trained models for your own purpose and business needs
  • How to address concerns regarding data and privacy
  • How to scale use cases and make them available across the enterprise


  • Rodrigo Rocha, Apps and AI Global ISV Partnerships Leader, Google Cloud
  • Mark Oost, Global Offer Leader AI, Analytics & Data Science, Capgemini
  • Sharon Goldman, Senior Writer, VentureBeat (Moderator)


Source link