AI agents are being pitched as the answer to rising customer expectations, but many early deployments are not yet delivering the experience lift marketers hoped for. Paul Meinshausen, co-founder of Aampe and head of agentic at MoEngage, says the gap is less about model capability and more about what we ask the agent to optimise.
The AI agent adoption curve
Marketing teams are moving through a typical adoption curve: high hopes followed by a satisfaction gap. Leaders expect agents to improve CX, while practitioners see uneven results in early pilots. Meinshausen sees that gap as a signal to fix measurement, not simply add more automation.
Why this matters now
For CX and martech teams, this moves the conversation from “which agent should we buy” to “what should the agent optimise for.” If the feedback signal stays stuck at clicks and opens, the agent will automate the old funnel rather than improve the customer experience. The opportunity is to use agents as an always-on decision layer that gets better with every real-world interaction.
CX has outgrown the click
The legacy CX playbook leaned on top-of-funnel click metrics. But consumer expectations and touchpoints now span the full shopping process, from discovery and comparison through purchase and support. An agent rewarded mainly for clicks will optimise for an incomplete view of the journey.
- Map the full shopping process, not isolated touchpoints.
- Treat every interaction as a signal for the next best action.
- Let agents learn from ongoing outcomes instead of static segments.
Governance and the “biggest lie”
AI agents can also reduce organisational friction. They can help navigate complex internal trade-offs, resolve cross-departmental messaging conflicts, and cut operational failure points. But the real shift is philosophical. Meinshausen points to a widely held belief he calls one of marketing’s biggest lies: that the answers are simply locked inside static data stores.
The alternative is dynamic, iterative feedback loops and reinforcement learning. Instead of mining historical dashboards, agents get continuous signals about what works and adjust their next action.
What marketers should do next
Before scaling agents, define journey-level outcomes, align guardrails across marketing, sales and support, and run small pilots with a clear CX scoreboard. The goal is not a perfect model but a feedback loop that improves with every interaction. Brands that close the loop early will get more from the same AI investments.
Source: MarTech
