Getting Started with Agentic AI

As I’ve mentioned, finance groups and other parts of the organization are increasingly using agentic AI. Tax groups just getting started with AI may have questions about AI governance and AI-related risks. Leaders must ask -- and address -- these important questions. It also helps to keep in mind that many aspects of the “AI trust journey” resemble the transition from on-premises systems to cloud.

In our development work and customer collaborations, we see opportunities for AI to evolve from a tool to a “co-worker” that tax professionals interact with to complete tasks. We also see opportunities for AI agents to support tax processes across the end-to-end lifecycle, co-innovation activities and more seamless integration among Vertex products and our ecosystem partners.

As tax leaders and professionals consider agentic AI uses, they should recognize that C-suites and boards may ask related questions regarding AI governance and risk management. Tax leaders planning to deploy AI tools should prepare to discuss their responses to these types of questions with the C-suite.  Questions that could arise are:

• How are we preparing for potential future developments in AI technologies and adapting them to our business as our strategy evolves?
• What specific business programs, processes or opportunities are we addressing with agentic AI?
• What governance structures are in place to oversee the responsible deployment and use of agentic AI within the organization?
• Have we or our partners addressed the risks associated with agentic AI deployments, including the risk of potential bias, hallucination and other operational risks?

A Protiviti bulletin to board members and C-suite leaders lists those and several more related questions.  While you do not need to answer the questions by yourself, the best practice is to be inclusive of several different groups, including your vendor partners, service partners and internal cross functional groups. Bringing together these different perspectives will go beyond answering the questions and foster a deeper understanding of opportunities to drive value and mitigate risk.

A KPMG report describes a three-phased approach to developing an agentic AI capability. 

• Phase 1 – Measure your AI readiness
• Phase 2 – Conduct an opportunity assessment
• Phase 3 – Implement an AI operating model designed for value

While IT may be leading your company through these phases, it is important for all functional areas to participate. These steps help to ensure that a solid foundation is in place, gaps are mitigated, the most promising areas to focus on are identified and execution is holistic with a clear measurement of progress. Additional value and acceleration can be achieved by partnering with vendors – either by leveraging your existing vendors’ AI roadmaps or identifying new ones that align with your needs – enabling faster progress across all phases and a quicker time to value.

The first two phases focus on establishing a strong foundation by evaluating the presence and quality of AI strategies, the availability and readiness of data to support AI use cases and the supporting technologies. They also assess critical enablers such as risk management, trust and governance processes, workforce skills, project management capabilities and mechanisms for tracking value (e.g., measuring ROI). Organizations should identify and address any gaps across these foundational areas. The second phase specifically involves an opportunity assessment to prioritize potential agentic AI use cases based on their expected value, associated risks and ROI targets.

The third and final phase of KPMG’s playbook tackles a big challenge: implementing an AI operating model. This capability requires these enabling components:

• Functional processes: clear goals, technology investment priorities and explicit ROI targets
• People and culture: cross-functional collaboration, reskilling/upskilling activities and communications (often related to employee concerns about AI’s impact on their roles)
• Service delivery models: The adoption of “enterprise-grade AI platforms”
• Technology and tooling: data governance and management, tight integrations between AI solutions and existing systems and security considerations
• Performance insights and data: KPIs, dashboards for monitoring AI agent performance and AI metrics linked to business outcomes
• Trusted AI: Ethical AI guidelines and risk management structures, data privacy and other regulatory compliance mechanisms (including the monitoring of emerging AI regulations), the formation of an AI ethics committee and regular audits of AI decision-making
  The above bullets can also serve as an outline of key areas to work with your vendors to understand their AI roadmaps and how they can complement your organizations’ AI operating model.

KPMG also stresses that time is of the essence: “organizations must begin transformation immediately as AI agents are disrupting business models faster than previous technology shifts.”

Tax groups don’t bear the sole responsibility for creating agentic AI playbooks and operating models for their companies. Yet, they should understand these plans thoroughly and prepare to stress-test them when needed -- and adhere to them as they advance their own agentic AI journeys.