For most organisations, the answer is no!

Don’t get me wrong; organisations are listening to the conversation and are keen on the potential long-term benefits. And some may indeed be ready for to implement agentic AI properly.

Ambitious, but underprepared: why most organisations aren’t ready for agentic AI

The reason I doubt that organisations are adequately prepared is that for agentic AI to function optimally, you require good clean data, seamlessly automated workflows that are well defined, and comprehensive flexible integration across all systems intended for agentic AI operations.

We see the importance of clean, integrated data in the acquisition of data management companies by other tech vendors, notably Salesforce and Informatica. Reconciled, normalised data gives AI agents the best possible intelligence to work with so they can provide the best outcomes and most accurate information.

You should always consider: How will AI agents benefit you and your organisation? Can your journey to autonomous work be achieved through workflows? Even Anthropic, a leader in the AI space talks about using the simplest solution possible and only implementing AI agents for complex requirements. In some cases, agents are not only more costly but also less efficient than pre-built workflows.

Preparation today, agentic AI tomorrow

As you consider pursuing agentic AI, there are several factors to examine. With planning and preparation, you can be successful, and this checklist will help guide you on your journey.

Planning

When you start your journey to agentic AI, clearly outline your goals for introducing agentic AI into your service management environment while ensuring that they align with your overall organisation’s goals. Some common objectives are:

  • Reduce the workload on IT staff by automating repetitive and time-consuming tasks.
  • Improve the speed and accuracy of incident resolution.
  • Enhance the end-user experience with seamless automated request fulfillment.

With your goals defined, what does success look like, and how will you measure it? These will vary based on your expectations and defined goals, but some common metrics could be:

  • % of tasks successfully automated by AI.
  • Decrease in errors within tasks handled by AI compared to manual tasks.
  • User satisfaction score (CSAT).
  • % of users opting for AI-driven self-service solutions.
  • Reduction in the average cost of resolving a ticket with AI.
  • % of incidents escalated to human agents after AI intervention.

Related content: AITSM: How AI Is Redefining IT Service Desk Automation, part of Ivanti’s Digital Employee Experience Research Report Series

Data preparation

Within the context of service management, there are many sources of data that may be required, including sources that are purely ITSM-related, such as a CMDB or asset repository. As you move towards the wider enterprise, you may also need data from CRM, HR, facilities, finance or other departments.

Regardless of where you are getting the data from, you need to consider the following:

  • Is your data quality reliable? Is the data consistent? How accurate is the information? Is there a more trusted source (or master) that should be considered as the default? Often data is inconsistent across multiple systems, making it harder for an AI agent to make well-informed decisions.
  • How accessible is your data? How many different data sources are there, and will integration via APIs be required? Are the data sources up to date?
  • Is your CMDB ready to be used by agentic AI? Do you have critical CIs in your CMDB, and more importantly, how up to date are the relationships and the information?
  • Are you required to adhere to data privacy compliance? Do you need to adhere to data protection regulations such as GDPR, CCPA or others? It is possible that regulations may demand that you anonymise sensitive data and keep it within country.

Defined workflows and automation

Not only is having high-quality data vital, but having good, functioning workflows to drive tasks is a critical component of successful agentic AI. In some cases, your existing workflows can be repurposed for use within an agentic AI process; however, some workflows will need to be modified to meet the request/response requirements of an AI agent.

Some things to consider for workflows:

  • Are your workflows able to be used by AI agents?  Your existing workflows may not be structured in the most effective way for AI agents to utilise the entire workflow to deliver accurate results.
  • Will the workflows only be working with data within your organisations, or will it need to access external data sources? When accessing external data sources, what security measures are in place to ensure the security and accuracy of the data within the responses?
  • Ensure all agent AI and related workflows have audit trails enabled, so you can log and monitor AI actions and decisions. This ensures the correctness of the AI and can be used in the event of compliance or security issues.

Governance and compliance

Earlier we discussed compliance requirements for data and the need to meet regulatory or statutory requirements. Governance and compliance must also apply to the entire AI ecosystem within your organisation to deliver responsible, secure and trusted responses and actions.

Here are a few things to consider:

  • Define policies for how AI is to be used within your organisation and assign people or teams who are accountable for protecting the data and evaluating ethical considerations.
  • Ensure that regulatory or statutory compliance standards are understood across the entire AI workflow and met. For example, if an AI agent requires access to health-related data and data collected from European users, then you would need to consider both HIPAA and GDPR.
  • Identify potential risks throughout your environment and meticulously mitigate those linked with the AI deployment.

Testing and validation

Start your journey to agentic AI in a controlled environment. Set up a pilot program to test the functionality, ensuring you measure against your goals and gather feedback from users. It is important to have a continuous improvement program in place for adjustments and enhancements. Be patient, as good results may take time as the AI learns and adjusts. Always monitor and adjust as required and define a process for handling issues or failures.

Agentic AI is an evolution, not a revolution

Discussion around AI is everywhere, and while there is a lot of interest from organisations, many are still rightly cautious about how to proceed and what is best for their organisation. As we have seen over the last few years, AI has progressed very quickly from generative to agentic. Keeping up with emerging technologies and planning for future enhancements is vital.

Ivanti is well positioned with AI capabilities that can enable you to grow. Find out more about what Think Tank can offer for your AI journey.

Key Takeaways
  • Most organisations are underprepared to implement agentic AI for IT service management.
  • The most important factors for agentic AI adoption are clean, integrated data, well-defined workflows and compliance with data privacy regulations.
  • Governance and compliance are critical, including policies, accountability and risk management to ensure responsible AI use.