Experts tell all: How to avoid 4 common enterprise AI adoption mistakes

Reading Time: 8 minutes

“By the end of this decade, there will be two types of organizations: those that fully utilize artificial intelligence (AI) and those that are out of business.” – Peter Diamandis, American entrepreneur.

While Diamandis’ line of thinking reveals a certain level of fear, pressure, and uncertainty surrounding AI, it shouldn’t come as a surprise. We’re still just beginning to understand the impact of generative AI, and yet it has already dramatically changed work, research, and communication with capabilities like document summarization, meeting recaps, live audio translations, and routine task automation.

And as use cases and platforms continue to grow, it’s only a matter of time before customers and employees define your organization by your AI deployment.

This is the opinion of Cory Peters, SHI’s Vice President – Product & MSP. Peters and other industry experts recently spoke at the 2025 SHI Spring Summit, where we, OEMs, and hundreds of organizations from across North America explored how to maximize productivity and security in the AI world.

During our summit, SHI’s Director – Product Management, John Moran, revealed a statistic that deeply resonated with our attendees:

When deploying AI, organizations should focus 70% of their efforts on people and processes, 20% on technology and IT, and 10% on AI algorithms, but most organizations focus disproportionately on algorithms.

In other words, they’re deploying AI all wrong (or, at least sub-optimally).

Let’s dive into common AI adoption mistakes and how you can better spend your time and effort where it really counts.

1. Not investing enough in AI literacy

Many organizations focus so much time and energy building out and testing their AI algorithms that they overlook two important truths:

1. Generative AI’s output can only be as strong as the prompts you give it.
2. You’ll see few benefits and may even introduce risks without training your employees to effectively and responsibly use AI.

This is why your organization should spend 70% of your efforts on people and processes. Without investing in your teams, you all but doom your adoption and productivity goals. To right the course, focus on your users’ AI literacy, which you can measure using frameworks like the University of Würzburg’s Meta AI Literacy Scale (MAILS).

MAILS measures users’:

• Confidence and effectiveness using AI.
• AI anxiety and willingness to use AI.
• Understanding of and ability to weigh risks when using AI.
• Ability to detect AI-generated media.

To build strong AI literacy, we recommend focusing on three user groups: senior leaders, employees, and practitioners.

AI-literate senior leaders

When your senior leaders have a deep understanding of and willingness to use AI, they strengthen the alignment between your organization and the AI platform you adopt. AI-literate senior leaders should be able to identify high-impact AI use cases that will increase productivity, eliminate manual tedium, and bolster employee satisfaction.

AI-literate employees

Simply put, AI-literate employees are more likely to adopt AI than those who aren’t. Inexperienced employees have a much higher chance of writing poor prompts, being frustrated with the unideal results, and quitting out of frustration.

Training them to effectively engineer prompts and understand data privacy implications, as well as overcoming their anxieties and fears around AI, will be essential for successfully driving widespread adoption.

Your AI-equipped workforce should have no trouble automating tasks via prompt engineering. As many of the tasks we performed even just a year ago can be automated by AI, achieving ROI with AI means supporting employees to focus their efforts on workflows that can’t (or shouldn’t) be automated.

And as an added benefit, investing in employees’ AI literacy can help you hire and retain workers! Gen Z knowledge workers expect access to AI tools from the get-go and may even skip over open positions from organizations that have yet to realize them.

AI-literate practitioners

And of course, where would your AI deployment be without the literacy of your AI practitioners? These are the specialists commandeering the technicalities of your AI solution. With a high AI literacy rate, your practitioners can solve AI use cases at a minimum cost, complexity, and risk.

AI-literate practitioners should be able to securely integrate your organization’s data into AI systems via retrieval-augmented generation (RAG). This helps output relevant responses while mitigating the risk of exposing sensitive information to public AI models.

2. Building AI use cases from the top down

While it’s important for your C-suite and executives to be as AI-literate as your employees and practitioners, you can’t fall into the trap of building AI use cases that primarily benefit the leaders approving your AI deployment. It happens far too often: leadership ideates and are excited about use cases that benefit their daily workflows while inadvertently overlooking the needs or pain points of teams that may not have a seat at the table during initial strategy sessions.

According to a survey of our summit attendees, 89% of IT leaders believe AI and AI-powered devices will improve productivity. And yet, research by Upwork found 77% of employees say AI tools will decrease productivity, while 47% have no idea how to achieve productivity gains through AI. Building AI use cases from the top down will only exacerbate this divide and hinder true value creation.

Consider this example: The events team suffering in silence

IT, C-suite, and sales leaders have been working for months to test and validate their organization’s AI platform. IT has outlined the security and infrastructure initiatives they need to address prior to deployment. C-suite leaders have projected AI’s ROI and outlined their change management strategy. Sales leaders are already training employees how to draft emails, summarize meetings, and automatically pull data from their CRM platform.

But unbeknownst to them, a team that could greatly benefit from AI is suffering in silence. In fact, they’re barely aware their organization is even planning an AI deployment.

This organization’s events team conducts attendee and sponsor surveys after every event and manually analyzes the data for weeks at a time. AI could expedite the process with automatic data analysis, but the team doesn’t have a seat at the table – and thus their needs go unaddressed, and leadership misses a valuable opportunity to demonstrate AI’s value.

When planning your AI use cases, the conversation can’t be siloed to just a few leaders and departments. Implement cross-functional AI discovery workshops or use case clinics that bring IT, business units, compliance, and frontline employees. Together, you can identify and prioritize opportunities and align with real operational needs.

As a tool intended to automate tedium and redirect productivity across your organization, AI needs to strategically solve problems from the bottom up, otherwise you risk investing in solutions with low adoption rates and minimal impact on employee productivity or satisfaction.

3. Overloading your data center with AI workloads

Unlike many traditional and legacy applications that rely mainly on your servers’ central processing units (CPUs), AI workloads draw heavily from graphics processing units (GPUs). If you intend to run multiple demanding AI workloads through aging servers configured for CPU-intensive work, you could set yourself up for data center slowdowns as well as soaring power, cooling, and management costs.

AI-powered devices are a great way to shift AI’s demands away from your data center. Built with neural processing units (NPUs) that can handle AI workloads while preserving device performance and battery life, AI PCs can unlock AI’s productivity gains without slowing down your servers.

Avoiding data center overload means building the right mix of GPU-equipped servers, cloud AI services, and NPU-based endpoint devices. And yet, nearly half (46%) of summit attendees are hesitant to adopt AI PCs due to device costs.

According to Dell’s General Manager, Workforce Solutions Group, Greg Haenggi, just a 12% boost in efficiency can yield a return on investment for AI PCs in as little as one month.

Here’s how you can start driving value and productivity from AI-powered devices:

• Offload AI workloads appropriately to AI PCs’ NPUs to spare constrained resources.
• Identify where AI makes the most sense for your organization and form use cases tailored to your users and capabilities.
• Build your own AI assistants with your data and processes via RAG.
• Train users on prompt engineering and responsible AI use, as well as how to make the most of Windows 11.

4. Being too slow to adapt to new cyberthreats

According to CrowdStrike Field CTO Christian Rodriguez, the average breakout time for a compromise to spread from one to multiple endpoints in 2016 was about eight hours.

Today, that time is just 48 minutes. And according to CrowdStrike’s 2025 Global Threat Report, the fastest breakout time on record was just 51 seconds.

This acceleration in threat penetration can largely be attributed to jailbroken large language models (LLMs), which bad actors use to weaponize AI in dubiously creative ways. This requires a shift toward faster detection and response that, by and large, many organizations simply aren’t adaptable enough to prepare for.

A new type of cyberthreat

At our Summit, Rodriguez shared an anecdote of a sophisticated social engineering attack that mimicked the voice of an organization’s helpdesk agent. The AI called an employee and, pretending to be the agent, asked the employee to give remote desktop permissions to threat actors who quickly took control of the device, stole valuable data, and breached the organization’s network.

How can you defend against threats in the age of AI?

As you strategize your AI deployment, it’s crucial to move beyond reactive defense and integrate security from the very beginning of AI development and deployment lifecycles.

So how can you bake cybersecurity into the core of your operations and adoption? Here are our tips:

1. Teach your users to recognize AI-generated text, images, and audio as part of your AI literacy initiatives.
2. Create a helpdesk practice that counteracts AI spoofing. For example, request employees call into your helpdesk, rather than having them await a call that can be mimicked.
3. Leverage AI-powered devices to monitor employees’ hardware for suspicious activity in real time.
4. Test and validate RAG architecture to ensure the data you feed AI remains secure and within your organization.
5. Establish clear AI governance policies covering acceptable use, data handling, model validation, prompt injection defenses, and incident response specific to AI-related threats.

Avoid AI adoption mistakes with SHI’s AI & Cyber Labs

At our Spring Summit, attendees had the opportunity to tour our new AI & Cyber Labs, which NVIDIA’s Vice President of the Americas Partner Organization, Craig Weinstein, described as having “the technologies and expertise needed to harness the transformative power of generative AI and drive innovation across industries.”

Here, your organization can test and validate leading AI platforms from Microsoft, AWS, Google, and NVIDIA in a secure, production-grade environment. Work with our trained engineers to develop your AI use cases, benchmark performance, stress-test solutions against your workflows, data, and security requirements, and adjust as needed to build and de-risk your ideal solution.

Coupled with expert-guided assessments and workshops from our Advanced Solutions Group (ASG), we’ll help overcome AI deployment pitfalls so you can:

• Drive AI literacy across your organization through AI literacy and adoption services.
• Build value-driven productive AI use cases via our AI Accelerator.
• Test and validate AI PCs in our Next-Gen Device Lab.
• Defend against evolving attack vectors and confirm your AI deployment readiness with our AI Readiness Assessment.

Don’t fall for common AI deployment mistakes. Build toward success with our AI & Cyber Labs or contact our experts today to learn more.