Fast track your AI-powered business

[ComputeExperts]

Get a performance-driven jumpstart on your AI initiatives or improve existing projects with these three steps. - By guest blogger Anne Taylor, independent technology and business writer

Somewhere between 60-80% of AI projects are currently failing, according to news sources, analysts, experts, and pundits.[i] In addition to infrastructure issues, organizations are struggling with business case justification, lack of in-house expertise for design and deployment, and IT integration, among other challenges.

Yet, AI has great potential to drive business value. The vast majority (89%) of IT decision makers are actively researching, piloting, implementing, or upgrading AI deployments, according to the Foundry 2023 AI Priorities Study. They expect their investments will:

• Improve employee productivity
• Enable innovation
• Help gain a competitive edge
• Improve customer support or services
• Develop new products or services

So, while it may be tempting to jump into the AI fray, harnessing these solutions requires three key steps to put you on the fastest route to business value:

1. Understand and prioritize your use case – and link it to your business case.
2. Determine which type of AI application or solution is right for your organization.
3. Choose your infrastructure approach: a pre-packaged route or a build-your-own solution.

Step #1 – Understand the use case and its relation to the business case

AI can address multiple manual or time-consuming tasks, such as optimizing customer-service workflows, summarizing enormous volumes of text, rapidly sifting through video footage files, and more.

To decide which use case makes sense for your business, start by answering this question: What is the business objective we are trying to meet, or the business problem we are we trying to solve?

For example, retail organizations seeking to improve risk mitigation could use AI to address product theft, employee safety, and property protection. There are also revenue-enhancing opportunities with AI solutions, such as improving customer service and demand forecasting, as well as optimizing pricing models.

Organizations must align desired outcomes with associated tasks. For example, to accelerate employee productivity, the business case should include processes like optimized workflows, rapid data analysis, or summarized video or text files. Narrow your objectives as much as possible to remain focused and better positioned for success. In other words, don’t try to boil the ocean; look for a focused use case with a strong chance of success as a starting point.

Step #2 – Choose an AI application to implement

AI is a sweeping term that includes technologies such as machine learning (ML), deep learning, large language models (LLMs), and intelligent automation. That said, AI can be broadly categorized into three groups:

• Vision AI. Also known as computer vision, in which digital devices identify and process information within images and videos. Examples include rapidly sifting through surveillance footage on cameras or reviewing medical images from healthcare equipment and analyzing input for recommendations.
• Speech AI. Uses voice-based technologies such as speech recognition, speech-to-text, and text-to-speech. It is a common way to power chatbots and digital assistants, which can augment customer service functions, sentiment analysis, and virtual meetings, among others.
• Generative AI (genAI): Creates text, images, audio, and other content based on data inputs or prompts into LLMs. GenAI algorithms can assist with data analysis, market research, fraud detection, and content generation.

The common thread among these AI groups is inferencing. They all require an engine that can rapidly apply logic and process information. Think about the size and volume of video files, slide presentations, webinar scripts, MRI images, and audio recordings.

The key thing to remember is that AI is performance driven,” commented Aaron Lamond, Worldwide Compute Product Marketing Manager at HPE. “To efficiently deploy and scale these projects, you’ll need inferencing systems that are designed for AI from the ground up.

Step #3 – Determine your AI implementation approach

There is choice when it comes to AI deployment: pre-packaged AI solutions or build-your-own. The fastest route to business value will depend on current resources, including budget and staff expertise.

Let’s say your organization plans to overlay a vision AI solution onto an existing camera infrastructure. “There’s a bit of work to configure and fine-tune it, but you can procure a pre-packaged vision AI software solution and have it implemented by an independent software vendor if your business lacks sufficient internal expertise,” Lamond said.

For example, having to manually review video footage for potential incidents is time-consuming and prone to inaccuracies. An off-the-shelf AI solution with a performance-driven inferencing model can be trained on existing footage to automatically scan for, detect, and identify risk-based situations – such as slick surfaces on factory floors or in assisted-living facilities that might lead to an accident.

Another option is to build your own AI inferencing solution. This approach requires leveraging dedicated high-performance compute infrastructure, using AI technologies such as LLMs, ML, or automation, and tapping enterprise data to create your own specific business tools, such as chatbots. For instance, your teams could adapt a LLM to develop generative AI tools that create sales and marketing content.

This do-it-yourself (DIY) path gives your business greater flexibility and results, but it requires assembling a plan, people, and tech partnerships for rapid success. Considerations include the use of pretrained models, frameworks, security, and support, as well as lifecycle management of the data, AI tools, and infrastructure.

“This is a customizable approach in which you would use your own IT ecosystem and, depending on how many building blocks you need, integrate LLMs, your intellectual property, and infrastructure to scale at your pace,” Lamond added.

No matter if you choose the pre-packaged or build-your-own approach, the underlying infrastructure matters.

How HPE powers AI today and tomorrow

There is a great deal of excitement around AI”, stated Aaron Lamond, “but also a reasonable measure of apprehension. We know enterprises want to deliver business value fast by learning from those who’ve gone before them. They will leverage best practices.

And that is where HPE can help. Working with other leaders in the AI industry such as Intel, NVIDIA, and VMware, we have developed performance-driven components that integrate, at a foundational level, “for an optimized, end-to-end stack that is performance driven,” Lamond reported.

For example, HPE is uniquely positioned to help organizations achieve their desired AI outcomes. The latest HPE ProLiant Gen11 servers are optimized to meet AI requirements, including:

• Optimized AI tuning and inference solutions
• Expert software ecosystem
• Open scalable acceleration
• AI-Native architecture
• 4^th and 5^th Generation Intel® Xeon® Scalable Processors
• NVIDIA GPUs including L4 and L40S

Also, enterprises can ensure a seamless integration of AI across their hybrid environments with HPE GreenLake. It’s an edge-to-cloud platform that offers a complete compute stack with layered security protection, optimized performance, and lifecycle management.

HPE offers AI workshops and consulting services to help fill knowledge or expertise gaps.

AI should be exciting, and not a drain to manage or protect,” remarked Lamond. “One of our goals with the HPE ProLiant AI solutions has been full-stack performance. We want our customers to get to value fast – and successfully.

Accelerate your path to AI business value with performance-driven AI solutions. Discover more at HPE.

Meet guest blogger, Anne Taylor. Anne is an independent technology and business writer with 20+ years of experience. She strategizes and creates content — including blogs, webinars, white papers, research surveys, and infographics — across a wide range of companies and industries. Her background is in both journalism and content marketing.

[i] https://www.forbes.com/sites/cognitiveworld/2022/08/14/the-one-practice-that-is-separating-the-ai-successes-from-the-failures/?sh=147515fe17cb