Greg Kihlstrom
Host of The Agile Brand podcastThe Future of Frictionless Loyalty: Navigating the Shift to an AI-Operations Layer
"The 'people problem' in AI adoption usually signals operational misalignment. When official tooling lags the work, teams improvise.”
The promise of AI-driven digital transformation often collides with a stark reality: well-funded rollouts stall not because the technology fails, but because the human operating model breaks down. As brands integrate AI across the Customer Experience (CX) journey, they face a dual challenge of navigating the “psychological debt” of displaced employee expertise while striving for an anticipatory, zero-friction customer relationship. In this discussion, renowned agile strategist Greg Kihlstrom breaks down why traditional CX dashboards are failing, what non-negotiable data foundations Fortune 500 companies need before deploying agentic AI, and how search intelligence must evolve into an invisible operations layer to achieve true customer loyalty by 2028.
Q & A
Greg, you’ve noted that AI adoption is being undermined by “psychological debt”—the cognitive and identity costs to employees. As brands integrate AI across the Customer Experience journey, how can leaders design these systems to augment human judgment rather than making teams feel like their professional expertise is being sidelined by an algorithm?
The “people problem” in AI adoption usually signals operational misalignment. When official tooling lags the work, teams improvise. That improvisation reflects a rational response to being asked to automate around contradictions the enterprise has yet to resolve. I’ve watched plenty of well-funded AI rollouts stall here.
Fortunately there are some design decisions you can make to keep humans-in-the-loop. This pays dividends not just in results but also in adoption and execution of the task at hand. Firstly, treat the ability to opt-out of intelligent-execution as a first-class capability. Make it as expected, fast, and respected as clicking submit. Make decision rights explicit. Record what the agent can do by themselves, what gets escalated, and who owns exception escalation workflow. Track direct adoption of new capabilities. Percent completion won’t tell you much about if your new capability is sticking in your customer-facing operations. If you want to build a knowledgeable culture that gets what you’re doing with your automations and knows how to execute alongside them, treat enablement like a Sustained Operation. Line it up with your platform change cycles.
The teams that adapt fastest move their expertise up the stack into system stewardship: designing prompts, governing exceptions, improving orchestration. Their authority grows in the process.
SearchUnify Lens:
Greg’s framing points to a real design failure: AI adoption stalls when systems sideline human judgment instead of reinforcing it. The fix is better system design, not better change management.
SearchUnify Governance makes decision rights explicit by defining what agents handle autonomously, what gets escalated, and who owns exceptions. Teams are not left guessing where the machine ends and the human begins.
As cases are resolved, the AI Knowledge Agent continuously captures and updates institutional knowledge, closing content gaps in real time. Teams that govern exceptions and refine this knowledge see their expertise grow alongside the system, rather than being replaced by it.
In your book The Agile Brand Revisited, you advocate for a model where brands are adaptive to change. How can organizations leverage Search Intelligence and Unified Data to move away from rigid, transactional interactions toward a living, breathing relationship that anticipates customer needs in real-time?
Anticipation requires shared state. Many “anticipatory” programs operate across systems that disagree about basic customer facts. The result is synchronized guessing. I’ve watched personalization engines fail because identity, status, and consent data were fragmented across 17 property-management systems with conflicting logic. The AI layer cannot anticipate what it cannot reliably identify.
There are three conditions of operation that have proven most important here. The first is a governed source of truth. Identity, status, consent and entitlement are unified across systems. Agents on brand-side can access that information in milliseconds. Second, a unified context window. Customer intent persists across marketing, service and sales handoffs. Third, real-time orchestration. Logic that consumes that context to determine the best next action, and take action, with transparent thresholds for human handoff.
Search Intelligence helps here because it gives the brand a way to read intent earlier in the journey and respond before the customer escalates. A booked trip, a service call, a category browse — these are intent signals already in front of most brands. The work is connecting what is already there into one operating layer.
SearchUnify Lens
Anticipation fails when systems disagree. Without a governed source of truth and shared context, anticipatory CX becomes coordinated guessing, with different systems acting on conflicting interpretations of the same customer.
SearchUnifyFRAG™ (Federated Retrieval Augmented Generation) unifies customer signals, intent, and interaction data across fragmented sources into a single, consistent retrieval layer, so identity, status, and context are no longer in conflict. Intent detected early in the journey persists across marketing, service, and sales handoffs rather than resetting at every touchpoint.
SearchUnify Governance encodes the third condition by enabling real time orchestration with transparent thresholds for human handoff, ensuring the system operates within defined boundaries and escalates when judgment is required.
You’ve said that “insight was never the point; execution is.” Why are traditional CX dashboards failing to drive change, and how can brands close the gap between knowing a customer is frustrated and solving the issue in real-time?
CX dashboards tell you what already occurred. A frustrated customer taking a survey creates value when the service rep on the phone knows what happened, has authority, and routing logic to remediate the situation at that very moment.
There’s a structural disconnect between knowing and solving. CX programs hoard insight in one area of the org and distribute execution to teams who lack the data, decision rights, or system access to do anything about it. Detection is fast. Reflection shows on the dashboard. Resolution speed is glacial within the operating model.
The gap is closed by plumbing real-time data all the way down to the point of execution, pushing decision rights to the person or system closest to the customer (with guardrails scaled to the risk involved) and building feedback loops that tell us whether we actually solved the problem. Ownership of end-to-end journeys should be centralized.
Insight that does not change a decision is reporting.
SearchUnify Lens
Dashboards reflect what has already happened. The gap Greg describes is not a visibility problem; it is an execution problem. Insight that does not reach the point of action is just reporting.
Model Context Protocol (MCP) closes that gap by pushing real time customer context, including intent, history, and status, directly into the systems and agents handling the interaction. The person or system closest to the customer has what they need at the moment they need it, without waiting for a report.
When execution happens at that speed, the AI Case Quality Auditor evaluates every closed case automatically, building the feedback loop Greg describes. Not quarterly reviews or sampled audits. Every interaction informs the next, so the system continuously improves rather than simply moving faster.
As Fortune 500 companies prepare for a future with 150,000+ AI agents, you’ve emphasized that governance is lagging. What are the non-negotiable data foundations a brand needs before they let AI agents interact with their customers?
Before any brand lets agents transact, negotiate, or speak on its behalf, four conditions need to be settled.
The first is a governed source of truth. Identity, status, consent, entitlement, and product information have to be consistent enough that two agents acting on the same customer arrive at the same answer. If your CRM and your billing system disagree about who is a premium customer, that contradiction will surface at machine speed.
The second is machine-readable structure across product data, pricing logic, refund windows, warranty terms, and sustainability scores. Anything currently held in PDFs or human-only UI needs a machine-readable version.
The third is governance-as-code: machine-readable rules for what an agent may offer, escalate, or refuse. A “Commercial Policy” the agent stays within. Once agents start negotiating with each other, this is the only thing standing between the brand and an unauthorized discount applied at scale.
The fourth is auditability and rollback. Leadership needs visibility into what an agent did, why, and the ability to reverse it when something goes wrong.
These are the same operational disciplines that should have applied to digital before agents arrived. Agents make the deferred work non-deferrable.
SearchUnify Lens
Greg’s four conditions are not aspirational; they are non negotiable. SearchUnify Governance addresses them directly, operating across five pillars: policy enforcement, access control, PII masking, guardrails, and auditability. Agents do not guess at boundaries; what they may do, escalate, or refuse is defined within the system from the start.
SearchUnifyFRAG™ (Federated Retrieval Augmented Generation) ensures a governed source of truth, so two agents acting on the same customer arrive at the same answer, regardless of which system holds the data.
The AI Case Quality Auditor completes the loop with full auditability of every agent action. Not sampled or retrospective, every case is continuously reviewed and flagged when something goes wrong.
When governance is encoded and auditability is automatic, AI agents become accountable, not just fast.
Finally, looking toward 2028, you suggest that the greatest predictor of loyalty is how easy a brand is to do business with. How does “Search Intelligence” evolve from a search bar on a website to a proactive, invisible layer that eliminates friction before the customer even feels it?
The strongest predictor of loyalty is effort. Customers remember what was easy. By 2028, Search Intelligence operates as a layer across the customer-brand relationship, much of it invisible to the human.
Applied at scale, that means agent-to-agent resolution, so a customer’s agent and the brand’s agent hash out a renewal, support ticket or reorder without humans ever opening tabs. It means pre-emptive response to intent, so booking a flight, browsing a category or making a service call automatically triggers the right action — before customers even ask. It means self-resolving friction points, so policy clarifications, exception approvals and returns are worked out by agents who can draw upon the same source of truth as their human counterparts. And it means earned inclusion in machine-mediated consideration sets, so brands win by being legible, verifiable and easy to buy from.
I’ve watched brands spend years on personalization theater — campaigns built on top of an operating model that cannot deliver on them. The brands that win the next chapter will have product, pricing, policy, and service logic coherent enough that any agent, human or machine, can act on them cleanly.
Stop optimizing the campaign. Make the brand easier to do business with.
SearchUnify Lens:
Loyalty is earned by removing effort, not by adding personalization on top of a system that cannot deliver it. By 2028, Search Intelligence is not a search bar; it is an invisible operating layer that eliminates friction before the customer feels it.
That means intent is understood before a question is asked, exceptions are resolved agent to agent without a human opening a tab, and every handoff carries full context rather than resetting it. MCP enables agent to agent resolution across enterprise systems without manual steps. Agentic RAG continuously refines the system through every interaction. SearchUnify Governance keeps policy and entitlement logic consistent, so any agent, human or machine, can act on it effectively.
Looking Ahead: The Architecture of Effortless CX
As organizations look toward 2028, the benchmark for brand loyalty will completely decouple from traditional marketing campaigns and rest entirely on a single metric: low customer effort. The transition from reactive search tools to proactive, invisible operational layers means enterprise architectures must adapt to handle machine to machine commerce and agent to agent resolution.
To thrive in this next era, leaders must pivot their focus away from surface level personalization theater and invest heavily in structural governance as code and machine readable business structures. The brands that secure long term market dominance will be those whose data foundations, commercial policies, and service logics are so radically coherent that any agent, human or machine, can instantly execute them without friction.
