The Biggest Challenges to Generative AI Adoption Success

Generative AI Adoption Success

Where does one start? There’s a long list of challenges when using generative AI to leverage and disrupt current IT service management (ITSM) and enterprise service management (ESM) operations. However, I can sum that list up in five words, “Get your house in order!” These are the top five ways you can achieve this.

There’s a long list of challenges when using generative AI. However, that list can be summed up in five words, “Get your house in order!” This article lists the top five ways you can achieve this. #ITSM #GenAI #ServiceDesk Share on X

Generative AI Foundations and Infrastructure

Prioritizing budgets around how generative AI will positively impact your ITSM or ESM operations necessitates preparing a good foundation.

For instance, let’s say you want to insert a closet organizer into a closet. First, you must go through item by item and inventory what you want to keep. Second, you’ll need to choose what you no longer need and determine how every item will serve you conveniently and efficiently.

You won’t leave the closet jam-packed full of stuff and start building your structure around it. You’ll clean it up, organize it, prioritize it, determine the most efficient access for you, and decide on what to store and track long before you add a single piece of wood or choose to hammer in the first nail.

The same applies to embracing and incorporating generative AI or other AI automation into your corporate ITSM and ESM operations. Design the infrastructure thoughtfully, with careful planning, data based on what you wish to achieve, and focus on how those solutions will feed the overall business outcomes.

Executive Support

A top priority for generative AI adoption is to get traction from the senior executive to invest in cleaning up what’s already there and readying the organization for your implementation. Communicating infrastructure priorities is tricky because the details appear complicated and complex to share, so getting ample budget is a challenge that pushes up against the pace at which the organization expects generative AI implementation.

'A top priority for generative AI adoption is to get traction from the senior executive to invest in cleaning up what’s already there and readying the organization for your implementation.' #GenAI #ServiceDesk #ITSM Share on X

Things are moving fast – something has to give. To get that budget, your team must be very good at conveying the priorities that must be in place, and depending on the maturity of your ITSM operation, the list can be long and extensive, maybe even seemingly unmanageable.

Start with a sponsor who is digitally aware enough to understand the complexity of the underlying infrastructure and the risks involved. We’ve all heard the phrase, “garbage in, garbage out,” which is particularly relevant to applying AI solutions. The senior level must understand that starting small and working toward a comprehensive solution is beneficial for getting your house in order.

Identify:

  • The problem generative AI must solve or the business goals it must achieve.
  • The required data is well-defined and cleaned up to achieve those goals
  • The investment in required infrastructure, segmentation, and security
  • The required collaboration needed across the organization
  • The necessary talent for continuous improvement and training on new data
  • The risks involved with bias, inaccuracy, privacy, and security
  • The use cases for starting small and increasing investment toward strategic outcomes
  • The organization’s readiness for adoption and the ability to fill gaps where generative AI is weak.

Convey these foundations with clarity and tie them to the organization’s business goals with a clear understanding of the effort involved.

Key Generative AI Considerations

Let’s start with some of the most critical areas: data quality, privacy, security, configuration management, and knowledge management. These areas need significant attention to ensure the completeness on which an organization will train its generative AI model. Ideally, you will start by identifying and defining your infrastructure’s relationships and dependencies. An organization’s data sources must be reliable, relevant, and represent/align with the use cases and business objectives.

Avoid letting the metrics become the goal. Meaningful metrics are targets for improvement, not the goals themselves. You must know what matters most to the company to get that right. Metrics must contribute to a detailed plan for increasing business and customer experience value.

Ticket times ( tickets completed, total time to close, etc.) aren’t a performance measure if you are unaware of what truth those metrics reveal. They’re incredibly ineffective when not combined with other metrics to understand the overall performance toward business goals.  

 Automating routine tasks or personalizing user interactions are dynamic and evolving activities. Without good data quality, feedback mechanisms, and privacy and security measures, you will struggle to make generative AI a beneficial addition to your ITSM or ESM. You will lose the analytical ability to predict future incidents or contribute to the reduction of repeated problems.   

Accountability and Ethics

Just because you can converse with generative AI tools, don’t believe for one minute that generative AI modeling or development will result in reliable outputs replicating human qualities. It can only spit out what you feed it because generative AI uses historical data. Any inaccuracies or errors in your data will now show up at scale and with more speed. The last thing your organization needs is for your bot to go rogue and recommend suicide or swear at users.

You must ensure there’s a human hand to assess output quality for AI accountability. This act means looking at the bias and fairness in training data and what the model development team prioritizes—putting a generative AI analyst in place to continually test and evaluate outputs and respond with human intervention where necessary. These evaluations of outputs delivered to internal and external customers are critical to continuous improvement.

Without mechanisms for users to escalate quickly, your generative AI might turn into a black hole of death; it’s imperative that a human can intervene and continuously set or reset ethical and accountable parameters.

Even if generative AI can simulate empathetic language and responses using natural language processing (NLP), it lacks emotions, consciousness, and personal motivation. It operates on algorithms, statistical patterns, and predefined objectives, so leverage it for those strengths. Any training data that’s biased or unethical can result in AI amplifying these biases or ideas in its output.

Both leadership and those leveraging generative AI must understand that it lacks causal understanding and is weak in cause-and-effect relationships or the ability to make sound judgment calls. The recent Open AI o1 Model does include chain-of-thought reasoning with some math science and coding tasks, surpassing other models. Where it shines is in creative writing and natural language processing, not necessarily good judgment.

Despite the remarkable outputs of the latest models of generative AI and other learning models, we must be on guard for wild hallucinations, safety issues, manipulation of task data, and the sharing of sensitive data.

Generative AI Governance

That last paragraph leads us directly to governance. A company with weak overall governance will find poor governance a significant roadblock to successful generative AI adoption in ITSM.

At the organizational level, you will need well-defined ethical guidelines. Companies must stay atop the ever-changing global and local regulatory changes surrounding the use of AI and be transparent by informing users of the information used in their model and the fact that they are dealing with a generative AI agent. Governance must include mechanisms to minimize risks due to changes in industry standards and regulatory requirements.

Companies must establish clear AI policies and guidelines early on within their organization for the use of data, as well as procedures. Organizations must develop well-followed processes to protect the organization, users, customers, and other stakeholders.

The last consideration a company must understand is the cost of the talent needed to develop, train, and oversee your generative AI model. Specialized data science and machine learning skills and expertise shouldn’t be wholly offloaded to the vendor if you leverage their tools. Good governance indicates the need to know what data, its use, and the quality around the automation and outputs associated with a successful generative AI addition within your organization.

Oversight, critical thinking, and judgment are valuable human qualities that become clear when we read news articles about generative AI blunders and recognize that those qualities are lacking in many generative AI models. Remember, generative AI is there to complement your services and augment the work, not replace entire service teams. We need humans to deal with the complex, nuanced, unforeseen issues users and customers encounter.

What do you think? Or what’s your experience? Please let me know in the comments.

Further Reading

Please use the website search capability to find more helpful ITSM articles on topics such as AI application (and the role of technical expertise with data scientists and data engineers), customer service, data collection and data integration, open-sourced software, regulatory compliance, pre-training with existing data, machine learning algorithms, foundation models, and improving business processes.

Patti Blackstaffe
Patti Blackstaffe
Author, Keynote Speaker, Consultant at GlobalSway

Patti partners with clients to solve systemic leadership challenges that reduce effectiveness so that they save time and money, reduce risk, increase adaptability, and advance strategic priorities.

Want ITSM best practice and advice delivered directly to your inbox? Why not sign up for our newsletter? This way you won't miss any of the latest ITSM tips and tricks.

nl subscribe strip imgage

More Topics to Explore

Leave a Reply

Your email address will not be published. Required fields are marked *