Have an agentic AI mandate? Start with the outcome, not the agent

"Build 100 AI agents." 

For senior leaders, it is an increasingly familiar mandate. But as one technology leader interviewed for new Mastercard research observed, it can be like being handed a hammer and told to find a nail.

The organizations positioned to create value from agentic AI will not necessarily be those that deploy the most agents. They will be the ones that connect the technology to meaningful business outcomes and make the operational changes required to deliver them. 

That opportunity extends beyond making existing work faster or less expensive. Agentic AI can help organizations reimagine customer and employee experiences, create new sources of value and support growth. But moving from activity to impact requires more than identifying where an agent could be deployed.

Mastercard’s recent research points to a dual transformation. Innovation defines the experience, workflow or outcome worth creating. Renovation makes that value deliverable and measurable by changing how the business works. 

Interviews with senior leaders across AI, strategy, innovation and transformation show these efforts often fall out of step. Some organizations deploy agentic AI without first defining how the outcome should improve, while others envision new experiences without making the operational changes required to deliver them.

As a result, critical questions may then surface only after significant time, resources and attention have been committed.

Innovation begins by looking beyond the agent itself.  

“Customers often mistake their agent as the strategy for the business,” says Apurv Bhatnagar, vice president, AI and Agentic Advisory at Mastercard. “Instead, their strategy should be the outcome they want to achieve, with the agent being the means to achieve it.”  

Rather than asking where an agent can be deployed, leaders should first consider:  

• Whose need are we addressing?  
• What should change for that person or group?  
• What evidence would show that the change is worthwhile?  
• What end-to-end experience could produce that result?  

This moves the conversation beyond automating an existing task. It encourages leaders to imagine what better result may now be possible across a complete customer or employee journey.  

“Deploy an agent” describes an activity. Helping a customer or employee reach a desired outcome sooner, with less friction and greater confidence, defines the value that activity should create.

Defining the outcome and reimagining the experience are only the beginning. Delivering that value also requires organizations to rethink how the work gets done. As Bhatnagar explains, “As much as agentic AI is about a technology shift, it’s also about an operating model shift.”

Consider merchant onboarding. For many banks and acquirers, moving a merchant from application to business readiness involves multiple systems, documents, teams and handoffs. An agent-enabled experience could preserve context throughout the process, identify missing requirements, explain next steps and involve a specialist when judgment is required. 

The value would extend beyond completing an onboarding task faster. A more continuous experience could help merchants begin operating and generating revenue sooner, reduce repeat contacts and create opportunities to introduce relevant products or services along the way. 

Capturing that value may require changes to the operating model surrounding the agent. Teams may need to improve access to information across existing systems, eliminate unnecessary handoffs, reconsider existing approvals and clarify where employees have the authority to act. 

Making these changes requires clear ownership. One business owner should be accountable for the outcome, aligning product, data, technology, finance and operations teams on how the work will change, while risk, legal and compliance teams define the controls. Domain experts can help design the new workflow and teach the agent how the business operates.

This is the renovation side of the transformation. It ensures that improvements within one task translate into a better end-to-end experience and a measurable business result. 

Why do promising agentic AI pilots stall before reaching production? 
Discover what enterprise leaders are learning in new Mastercard research with The Harris Poll. 
Download the report

Even a well-designed use case must demonstrate that it can work beyond a controlled pilot. 

Real operating conditions introduce incomplete data, unusual cases, system limitations and unpredictable user engagement. They also reveal whether employees can support the redesigned workflow and whether improvements in an individual task translate into meaningful business value. 

Organizations must also determine how much authority an agent should have. They can begin with agent-generated recommendations or lower-risk actions, then expand authority as evidence grows to confidently support that decision. The level of granted autonomy should reflect the task and its potential consequences, with clear limits, escalation points and human recourse for any action that falls outside defined boundaries.

Together, these tests help leaders determine whether to expand, redesign, pause or stop a use case based on credible evidence. The full Mastercard report explores how organizations can test capability, organizational readiness, production readiness and business value, while addressing the data, context, governance and controls required to expand agent delegation over time. 

As agentic AI becomes a more central part of enterprise planning, differentiation will come from how organizations apply their unique data, context and expertise to drive outcomes that matter. 

Innovation identifies the value worth creating. Renovation establishes how the business will deliver and measure it. Advancing them together can help organizations turn agentic AI from a mandate into a repeatable source of measurable value. 

Enterprises can begin with a simple roadmap:

1. Innovation
   1. Identify high-value business outcomes
   2. Prioritize agentic use cases tied to those outcomes
2. Renovation:  
   1. Assess data readiness, workflow dependencies and organizational capabilities
   2. Roll out operational changes, if needed
3. Experimentation:  
   1. Test use cases under realistic conditions
   2. Expand deployment when the evidence supports it

Download Moving agentic AI from promise to performance to explore the full framework for defining value, renovating operations, generating credible evidence and building trust as agent operating authority evolves. 

Ready to identify and evaluate your highest-value agentic opportunities? Learn more about Mastercard Agent Suite.