AI Hype Vs Reality: CDW Tech Leaders Break Down Key AI Trends at BCNET CONNECT

When researchers subjected the Claude 3 AI model to the needle-in-a-haystack test, not only did it pass the test, but the model was also able to, almost as if consciously, decipher that it was being tested.

Such instances may spark emotional responses. “All of a sudden such occurrences can lead to opinions about this optimistic world or this pessimistic world, depending upon whether AI is going to solve all of our problems, or is it going to lead us into a dystopian future?” Burke said.

But in this case, the model was able to understand the test because it happened to be included in the data that the model was trained on. So, naturally the model was aware of the test. 

Without this key piece of information, one may form emotional opinions about AI systems. You might overestimate their capabilities with extreme optimism or let fear and uncertainty creep into the decision-making process.

Therefore, it’s important to understand how these AI models really work to ensure decisions are based on factual knowledge.  

The successful adoption of AI systems fundamentally relies on two things:

• Data: The integrity, cleanliness and usability of the data on which an AI system is trained.
• Tools: How organizations select and utilize various AI-powered tools for their use cases. 

 “Our computers these days can actually run a lot of these LLMs on-premises and give access to private data, which makes me a little scared. Because now we have people that are out there creating apps on their own without any IT knowledge,” Beauchamp said. 

Until there is a deep understanding of what kind of LLM tools you’re working with and how they use your data, it’s risky and difficult to take full advantage of AI systems. Beauchamp stresses the need for data governance and guardrails to prevent misuse and inadvertent damage.  

Based on the customer conversations and new solution architectures being built by our experts, here are six key trends that the market is expected to observe in 2024.

1. AI transformation 

AI can transform how individuals achieve daily repetitive tasks. AI-powered tools and “as a Service” offerings will continue to rise. Organizational deployment of AI tools will depend largely on data governance maturity.

“It’s important to understand that you’re looking at small percentage gains that are out there. Right now, we’re in a bit of a primitive age of AI. We have LLMs with capabilities like text-to-speech, voice-to-image, etc., but those are going to be the building blocks,” Beauchamp said.

“As of today, AI applications can’t shoot a business four years into the future. It’s about small incremental gains and level setting your expectations from a business perspective.” 

2. Data management and governance 

Organizations will look to integrate AI and advanced data analytics into their businesses and will need to focus on better ways of managing and curating data. There will be a need for data strategy that lays down policies on how data is acquired, processed and governed within the organization.

“The need to take on formal data management, data governance and have teams put guardrails around user-based controls, zero-trust frameworks – that’s really where a lot of organizations would have to deep dive and build those muscles within,” Burke said.

3. Insights at the edge 

Insights from data gathered at the edge will help organizations in retail, manufacturing, healthcare and finance innovate their business and take advantage of AI faster than other verticals. These industries have a higher tendency to use data insights that can be fused with AI applications for rapid transformation.  

4. Multicloud investment and expansion

In the coming years, we will see renewed investment in infrastructure to not only prepare for AI innovation but also invest in building the hybrid multicloud. Organizations will need to better standardize their tools and technologies across the public cloud, modern data centre and the edge.

5. Cybersecurity integration

Cybersecurity started as a separate entity within many organizations, but many roles have begun to combine traditional operations teams and cybersecurity teams. Risk to the business needs to be shared across the entire organization and coordination and shared tooling will be required.

6. Business focus

As businesses digitize services and deliver value using applications, their budgets are increasingly directed to projects around application development and away from IT operations. It is critical for IT teams to align with business objectives alongside IT operations to stay relevant as well as ensure the business follows well-defined technology practices.

It’s expected that AI adoption will follow a top-down path. Managers will expect their teams to be more productive using AI-powered tools that can help them use data in whole new ways.

• Coworkers and teams: Teams will be able to access and use data to streamline workflows, achieve higher productivity and do more within their organization. This would have required IT expertise a few years ago, but today, it can be boiled down to writing a simple sentence to get the job done.  
• Organizations: More LLMs will be focused on business units and are more likely to be developed in-house. Organizations would prioritize areas that can drive innovation, bring more value and increase revenue.
• Platforms: AI platforms such as NVIDIA will play a key role in enabling data services, compute power and integrated offerings to make AI innovation accessible for all.

“For organizations, it’ll be about not putting roadblocks in the way, but building guardrails and ways for people to use AI tools within the security policies,” Burke said.

AI management can be broken down into four major disciplines: 

• Data governance: Be aware of how your private data is leveraged by an AI system, make sure to remove biases and build policies that disallow misuse of sensitive data housed in your organization. 
• Cybersecurity: Work with your IT teams to understand the security aspects of AI systems you want to deploy. Test new pilots rigorously before they go public from a cybersecurity perspective and move towards more secure ways of deployment.
• Platform management: Adopt modern AI deployment practices such as AIOps to improve the manageability of AI systems while leveraging the scalability of cloud. 
• Model management: Build strategies around how you select models, manage their lifecycle, migrate to new models and benchmark their performance for the intended results.

As Burke explains, most organizations eyeing AI investments often run into discovery problems. There are heaps of unorganized data within their enterprise systems that they don’t know how to streamline. Without a clear roadmap, there is fear, uncertainty and doubt about the feasibility of the AI systems.

They are not fully sure of the use cases that would make the most sense to their business workflows. And lastly, they are uncertain how to start building AI capabilities.

We suggest the following measures to overcome the struggles associated with AI adoption.

• Data curation: Start by identifying what is important in your data. Segregate datasets that could meet organization objectives and better support the tools you want to use for maximum value creation. 
• Demystify AI: Promote AI education within your enterprise and be on the lookout to improve your understanding of AI systems you wish to work with to reduce the fear and uncertainty that comes with limited knowledge. 
• Start simple: Focus on easy implementations that are less resource intensive and still drive value for your users. Identify use cases that don’t require boiling an ocean. 
• Identify partners: Work with expert partners that can help you reduce risk, chart a clear roadmap and test out novel technology with greater confidence. 

CDW Canada is home to several distinguished technology experts who have enabled Canadian organizations with ground-breaking solutions in the past and continue to push the envelope in the AI age.

Our experts can help you navigate the unknowns around AI adoption and help you build a definitive pathway to successful implementation. We specialize in solutions that span hybrid cloud, digital workspace and cybersecurity, all of which have massive potential to support AI innovation across industries and domains. Get in touch with our experts to discuss and brainstorm your next AI pilot.