- Agentic AI is shifting the customer experience from navigating provider-defined journeys to delegating outcomes. Customers will increasingly ask AI agents to research, compare, transact and service on their behalf.
- This creates growth and efficiency opportunities, but it also raises new questions about agent identity, delegated authority, consent, fraud, accountability and the risk of losing the direct customer relationship.
- PwC and Ping Identity bring together the two capabilities enterprises need to respond: business and operating-model transformation, combined with an identity control plane that governs every agent action at runtime.
What Are Agentic Channels and Why Do They Matter
For more than two decades, organizations have competed to create better digital channels. Banks, insurers, and retailers moved services online, then connected websites, apps, marketplaces, contact centers, and physical locations into omnichannel experiences.
Agentic AI changes the premise. Instead of navigating each provider's process, a customer can express an intended outcome: find insurance, switch a savings account, resolve a claim, or make a purchase within budget. An AI agent can interpret those preferences, coordinate across organizations, and act, returning when a material decision or approval is needed.
This is more than a smarter chatbot. An agentic channel lets an AI agent plan and act for a person or organization within defined authority. Research commissioned by the UK Financial Conduct Authority (FCA) found that one in five consumers (around 11 million UK adults) would likely use autonomous AI within preset goals.1
The shift changes three things at once:
How Agentic Channels Change Growth and Competition
Agents can gather information, pre-populate forms, monitor conditions, and coordinate work across providers. By reducing friction, they can improve conversion, lower service costs, and create more continuous relationships. They may also surface relevant products that traditional advertising or search would miss.
The same shift introduces disintermediation. A bank may become one offer in an agent's comparison, an insurer anonymous capacity behind a marketplace, or a retailer a fulfillment endpoint selected on price and availability. The provider may complete the transaction yet lose visibility into the customer and the opportunity to build loyalty.
The strategic response is not simply to block agents or accept them as anonymous automation. It is to create a trusted agentic channel: one that recognizes the agent, connects it to a consenting customer, understands its mandate, and supports useful action without surrendering accountability.
Why Trust Is the Hardest Agentic AI Problem
Most digital controls assume a human signs in and remains present. Agents break those assumptions. They may act for hours, call application programming interfaces (APIs) and tools, collaborate with other agents, or adapt their plan. A credential issued at the start cannot prove that every later action is appropriate.
An agent may look like a malicious bot. It may reuse customer credentials, carry broad consent, or delegate work without preserving authority. Even a legitimate agent can be compromised, overprivileged, or misaligned with intent. At machine speed, a small control failure can become a large business, conduct, or security event.
The governing questions therefore become continuous: Is this agent legitimate? Whom does it represent? What was it authorized to do? Does that authority cover this action, value, resource, and moment? Is new consent or human approval required? Can the complete chain be reconstructed?
What an Identity Control Plane Does for AI Agents
An identity control plane turns those questions into enforceable policy. It moves the security boundary from a one-time login to every material action at runtime.
Each managed agent receives its own identity instead of hiding behind a human account. That identity is related to the principal it represents. Delegated authority becomes a specific, revocable mandate. Fine-grained authorization evaluates each request against identity, intent, resource, risk, and policy.
The control plane should also protect APIs, tools, and Model Context Protocol resources, and detect anomalies. It should govern lifecycle, privilege, and credentials, trigger human review for high-impact decisions, and record the chain from principal to agent, mandate, policy decision, action, and outcome.
This is the role of Ping Identity's Identity for AI capabilities within a converged identity platform. It enables organizations to secure agents alongside customer, workforce, and partner identities, avoiding another isolated control stack while supporting agent recognition, delegation, runtime authorization, threat detection, and traceability.
Why Business Transformation and Identity Architecture Must Move Together
Trusted agentic channels are not solely an AI initiative or identity and access management (IAM) program. They change strategy, product design, distribution, data use, operations, risk, and architecture. Strategy without enforceable control cannot scale. Control without a clear customer and commercial purpose adds friction.
PwC helps organizations decide where and how to compete: which journeys create value, whether to build provider agents or support external agents, how disintermediation may reshape the value chain, and which operating-model, governance, and regulatory changes are required.
Ping Identity provides the identity control plane that makes those choices governable. Together, PwC and Ping connect business ambition to trusted execution from journey design through AI agent identity security, delegated authority, runtime policy, and auditability.
Where Leaders Should Start
Organizations should not open every journey to agents at once. A measured path expands autonomy as identity, policy, and operational confidence improve.
Start by discovering and prioritizing. Inventory provider, personal, partner, and workforce agents, including those embedded in software. Map the intent, data, tools, and systems involved. Choose a high-volume, bounded journey that solves a real problem.
Next, establish the trust foundation. Give agents identities, define mandates and revocation, centralize runtime policy, protect APIs and tools, and specify when human approval is mandatory. Measure value and control together: completion and time to outcome alongside policy denials, anomalies, escalation, and customer understanding.
Then scale deliberately. Prove provider-owned journeys before opening appropriate services to personal, comparison, and ecosystem agents. Support relevant interoperability standards, publish machine-readable capabilities, and define certification, liability, dispute, and recourse processes for activity that crosses organizational boundaries.
Who Will Win the Next Customer Interface
The agentic-channel race will not be won by the organization that deploys the greatest number of agents. It will be won by the organization that makes its products and services easy for trusted agents to discover and use, while preserving customer choice, context, and control.
That requires a new strategic asset: converged identity able to secure human and non-human actors across customer, workforce, partner, and agentic channels. It also requires a transformation model that aligns technology with commercial value, customer outcomes, and regulatory responsibility.