Why the cloud is a key ingredient of GenAI success

Generative AI – now commonly called GenAI – is thrilling consumers and organizations alike with its ability to generate new and original content such as images, text, audio and even entire virtual environments.  

In the corporate world, this emerging technology has several applications. First, it can streamline and automate content creation. For example, marketing teams can generate larger volumes of personalized content, social media posts or product descriptions in less time than before – even if some of the GenAI content still needs manual fine-tuning.

Additionally, by generating diverse ideas and concepts, GenAI can serve as a valuable tool for brainstorming sessions, product design and problem-solving. It can also assist in prototyping and generating realistic simulations, so that organizations can visualize and test ideas before they invest in physical prototypes.

Furthermore, GenAI can improve customer experience through personalized recommendations, smart chatbots and virtual assistants like Copilot for Microsoft 365.

Organizations should approach these new opportunities strategically, considering potential challenges such as unintended bias in AI-generated content and privacy and security concerns relating to the sharing of corporate data with AI tools.

• ALSO READ → Get ready for GenAI to transform your business operations

Cloud is the backbone of GenAI

However, a corporate GenAI strategy shouldn’t start and end with selecting and deploying AI tools. The underlying IT infrastructure that enables GenAI is just as important, and cloud computing is at its core.

• GenAI algorithms require significant computational power and resources to generate and process large amounts of data. With cloud computing, organizations can scale up or down their computing resources cost-effectively to handle various types of workloads.
• Implementing GenAI in legacy monolithic applications presents significant challenges – primarily due to the applications’ rigid architecture – whereas the application programming interface (API) architecture and modularity inherent in cloud-native applications facilitate the seamless integration of GenAI components.
• GenAI tools often need access to vast amounts of data for training and generating content. Cloud computing enables organizations to store and manage these large datasets, with easy sharing of data across teams and locations to facilitate collaboration. Employees can access GenAI applications and resources from anywhere, at any time, using various devices.
• Organizations that have switched to cloud computing don’t need to invest in their own expensive hardware upfront. Instead, they leverage cloud service providers’ infrastructure on a pay-as-you-go basis.
• Cloud service providers invest in robust security measures and infrastructure to protect data and ensure high availability. This is particularly important when dealing with sensitive data used in GenAI applications. Cloud platforms also have built-in redundancy and failover mechanisms, minimizing the risk of data loss or business disruptions.

Planning is the key to success

So, if you want to implement GenAI in your business, where do you start? Try to answer the following questions as you strategize:

Why do you need GenAI?

Improving efficiency is a common goal for organizations deploying GenAI. By automating tasks and processes, GenAI can streamline operations, reduce manual effort and increase productivity. Improved efficiency can lead to cost savings, faster turnaround times and operational effectiveness.

However, deploying GenAI can also be driven by the goal of generating revenue directly. GenAI can be used to create new products or services that generate additional revenue. For instance, a GenAI-powered recommendation system that suggests personalized products to customers may lead to increased sales.

These are not mutually exclusive goals, though, as improving efficiency through GenAI can also indirectly contribute to revenue generation. Ultimately, though, the deployment of GenAI should align with the overall strategic goals of your organization.

What are your use cases?

What exactly will you use GenAI for? Map out all the use cases that you have in mind. These will help you decide whether you want to develop and deploy GenAI within your own infrastructure or use public GenAI services provided by cloud service providers. Private use gives you more control over data privacy and security but requires a significant investment in infrastructure and skills. Public use, on the other hand, offers more convenience and scalability but may raise concerns about data privacy and vendor lock-in.

Also keep in mind that GenAI can involve various types of content generation, such as text or image generation. Understanding your specific use-case requirements is therefore crucial. For example, if a manufacturer wants to automate the generation of responses to alerts issued by a factory-floor monitoring system, text generation may be sufficient. But if they also want to use images of potentially defective products for quality-assurance purposes, image-generation capabilities become essential.

You’ll also have to assess the availability and quality of the data required for GenAI. Do you already have access to the right high-quality data, or will you have to collect, clean or augment existing datasets?

Then, knowing your ideal use cases will help you determine what level of expertise you’ll need in AI, machine learning and data science (and in maintaining the underlying IT infrastructure). You can then expand your in-house team or partner with an expert third party. For instance, NTT DATA has supported the Tour de France by delivering AI-driven insights in real time that benefit both the fans and the organizers.

Where will you run your GenAI applications?

As mentioned above, you’ll need to decide between private and public infrastructure for your GenAI strategy for optimal performance, scalability, cost-effectiveness and security.  

If your existing infrastructure lacks the computational power, storage capacity and network capabilities to run GenAI applications effectively, consider partnering with a cloud service provider that offers infrastructure and resources designed for AI workloads – including services and tools that can improve your GenAI workflows.

Working with a skilled service provider will also allow you to first evaluate the GenAI readiness of your infrastructure before deciding how to proceed. The service provider will also help you understand your implementation and usage needs and what type of cloud you’ll need.

Access to this level of expertise means you can implement GenAI swiftly and get ahead of your competitors. However, evaluating the scalability and flexibility options provided by different cloud providers remains crucial to ensure that your GenAI applications will handle varying workloads efficiently.

You may also need to integrate your new GenAI applications with your other tools, services or existing infrastructure, and working with a full-stack service provider will make such an integration far more painless than doing it in-house.

Importantly, running GenAI applications comes with serious cost implications. Cloud providers’ pricing models, such as pay-as-you-go or reserved instances, can help you optimize costs based on your usage. Compare pricing structures across cloud providers to find the most cost-effective option.

What are your cloud-based GenAI security considerations?

Corporate GenAI applications often deal with sensitive data, and you should prioritize data privacy and security.

Data protection and access control is key at all times. And, depending on your industry and geographic location, you may need to comply with specific data-protection regulations such as the European Union’s General Data Protection Regulation.

Also keep in mind that cloud-based GenAI applications rely on network connectivity to function, so network security measures such as firewalls, intrusion detection and prevention systems, and virtual private networks will help protect against breaches.

Cloud providers may differ in their security measures, compliance certifications and data-protection capabilities, so do your homework based on your use-case requirements to check that your GenAI applications will meet regulatory requirements and maintain data integrity.

How will you fine-tune your GenAI model?

GenAI models don’t necessarily arrive ready for duty on day one, although some of the latest iterations are more advanced in this regard. Fine-tuning these models – generally an iterative process requiring sufficient computational resources – is therefore still an important step in optimizing their performance according to your requirements.

This involves clearly defining your objectives and desired outcomes, ensuring that the training data is diverse, bias-free and of the highest possible relevance and quality, and adjusting several technical parameters and optimization algorithms.

It’s also essential to establish robust evaluation metrics and validation procedures to assess the performance of the fine-tuned model. Once the GenAI model is fine-tuned and deployed, its performance should be continuously monitored and adjusted over time. This involves techniques such as A/B testing, human feedback loops (where the outputs or results of actions are fed back into the model as learning material) and retraining the model with new data as required.

These tasks can be complex and time-consuming, so working with an expert service provider is likely to prove beneficial.

Service providers like NTT DATA are also combining cloud business services with underlying infrastructure and platform-as-a-service offerings in the form of industry clouds to make it faster and easier for clients to build new industry capabilities or support value chains.

As more of these capabilities become GenAI-centric, organizations can benefit from pretrained models and use cases rather than having to train generic AI models from scratch. 

• ALSO READ → How much can AI systems like ChatGPT really help your organization?

Need more input on GenAI infrastructure?

In my next blog in this series, I’ll go deeper into the best way to evaluate infrastructure options for GenAI, including the potential benefits and drawbacks of public, private and industry cloud.

In the meantime, we’re ready to help you explore the potential of GenAI in your organization.