<   Back to Blog Articles

Managing risks and delivering successful AI-driven outcomes at the Service Desk

5 Dec, 2024•

Jim Hirschauer

AI is at or approaching the peak of its Hype Cycle™ — a graphic modeling of the typical progression of an emerging technology from initial excitement through disillusionment and eventually to practical usefulness.

Now more than ever, it’s imperative that IT leaders carefully balance the risk vs. reward of all AI projects within their organizations.

A recent Gartner research ‘How to Achieve Success With AI at the Service Desk’ uncovered many eye-opening AI-related trends. Two that jumped out at us were:

1. 45% of respondents indicate that “I&O [Infrastructure & Operations] will use GenAI to engage more efficiently with employees in the helpdesk through virtual assistants or chatbots.”
2. 44% of respondents indicate that “GenAI will automate repetitive and manual tasks in I&O [Infrastructure & Operations] workflows, increasing efficiency and reducing human effort.”

Given these revelations, it’s critical that I&O leaders mitigate risk to achieve the desired positive business outcomes.

But there are roadblocks.

Top barriers to successful AI-driven business outcomes

In addition to the metrics listed above, the top three reasons for AI not meeting ROI expectations are:

1. Cost
2. Security
3. Hallucinations

Fortunately, all 3 of these risks are manageable.

Note: Gartner also states, “Return on Investment Is Key, but Should Not Be the Only Driver.” Download a complimentary copy of this Gartner research for more details.

Reason #1: AI costs must be controlled at the Service Desk

At Xurrent, we’ve incorporated many productivity-enhancing AI features, offered to customers without any additional licensing or usage charges.

How is that possible?

We’ve been very intentional and deliberate with our AI implementation, using Amazon Bedrock as our AI backend, a fully managed service from Amazon Web Services (AWS) that makes it easy for developers to build and scale generative AI applications. Amazon Bedrock has features we leverage to minimize costs.

A few of the key strategic cost savings measures we’ve adopted include:

• Pay-as-you-go pricing model: We only pay for the computing and storage we actually use, making it easier to manage and predict AI expenses.
• Scalable infrastructure: Resources are allocated based on demand, preventing overuse and saving money by avoiding unnecessary expenses.
• Choosing the right models: We can select from different AI models that vary in capability and cost. By picking models that fit our specific needs, we keep costs down without sacrificing performance.
• Cost tracking tools: Integration with tools like AWS Cost Explorer allows us to monitor our spending in real-time. We set budgets and receive alerts if usage patterns become unusual, helping us stay on top of costs.

This complimentary Gartner report details how to build a cost model that reflects both direct and indirect costs of implementing AI at the Service Desk.

Reason #2: Maximizing AI security

Amazon Bedrock has the following capabilities we utilize. If you use a different AI provider, be sure to follow similar patterns to secure the AI implementation.

• Data encryption: All data handled is encrypted when stored and during transfer, protecting sensitive information from unauthorized access.
• Compliance with standards: Compliance with various industry and regulatory standards like GDPR, HIPAA, and SOC 2 means the AI service adheres to strict security and privacy requirements for handling sensitive data.
• Access control: The ability to set and enforce detailed access permissions guarantees that only authorized people can access or change AI models and data.
• Data isolation: Keeping each customer’s data and workloads separate from others prevents data from different users from mixing or leaking.
• Continuous Monitoring: Providing ongoing monitoring and real-time threat detection of the AI platform helps quickly identify and address potential security issues.
• Secure Development Practices: Following AWS’s strict software development practices, including regular security checks, code reviews, and vulnerability scans, keeps Amazon Bedrock safe from new threats.

There is clearly a lot to consider around AI security.

At Xurrent, over 90% of our customers have enabled AI — a testament to the security measures that we have implemented.

Reason #3: Minimizing hallucinations

As Gemini Advanced (AI) says, hallucinations refer to an instance where an AI model generates incorrect or misleading information that isn’t grounded in reality or the data it was trained on. It’s like the AI is making things up, similar to how humans might hallucinate under certain conditions.

Whether you’ve dipped your toes into AI or dove into the deep end headfirst, you’ve likely encountered AI hallucinations.

Their impact can range from annoying to funny to disastrous for a company. Here are a few tips on how to avoid AI hallucinations:

• High-quality models: Consider using top-notch AI models from leading providers like AI21 Labs, Anthropic, Stability AI, and Amazon. These models are trained on large and diverse datasets, making them more reliable and less likely to make errors.
• Customization: Fine-tune AI models with your own data, making the AI’s responses more accurate and relevant to your specific applications.
• Effective prompt design: Create well-designed prompts. Clear and specific prompts guide the AI to produce more accurate and appropriate responses, lowering the chance of mistakes.
• Content filtering: Use content filtering to review AI-generated outputs before they reach users. This extra step ensures that any incorrect or inappropriate information is caught and corrected.

The Gartner report How to Achieve Success With AI at the Service Desk provides a framework for improving data quality by implementing a “distrust and verify” policy. Getting to a point where there are never AI hallucinations is out of the realm of possibility — for now — but they can be minimized to the point where they are statistically irrelevant.

Moving AI from hype to measurable ROI

Successfully implementing AI in your service desk operations isn’t about jumping on the hype train – it’s about taking a measured, strategic approach that addresses the core challenges head-on.

By implementing scalable cost management through pay-as-you-go models, maintaining robust security measures that earn customer trust, and developing effective strategies to minimize hallucinations, organizations can move beyond the 47% of CIOs reporting insufficient ROI from their AI initiatives.

The key is finding the right balance between innovation and control, ensuring that AI implementation delivers tangible business value while managing potential risks. Whether you’re just starting your AI journey or looking to optimize your current implementation, you need a comprehensive framework for success.

Click below to download the essential Gartner report “How to Achieve Success With AI at the Service Desk” to access detailed insights and practical strategies that will help you maximize the value of your AI investment while minimizing potential pitfalls.

Resources and disclaimer:

1. Gartner, How to Achieve Success With AI at the Service Desk, Mark Cleary, Rich Doheny, 29 August 2024
2. Gartner, Hype Cycle for Artificial Intelligence, Afraz Jaffri, Haritha Khandabattu, 17 June 2024
GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally, and HYPE CYCLE is a registered trademark of Gartner, Inc. and/or its affiliates and are used herein with permission. All rights reserved.