New arXiv paper argues autonomous "agentic" personalisation can sustain CRM performance over time

Lead: autonomous agents, sustained gains?

Can autonomous, learning marketing agents replace constant human oversight? A new preprint on arXiv, "Sustained Impact of Agentic Personalisation in Marketing: A Longitudinal Case Study" (arXiv:2604.08621v1), reportedly finds that agentic personalisation systems can deliver sustained performance gains in customer relationship management (CRM) without continuous manual intervention. The study frames agentic personalisation as adaptive, goal-directed systems that make messaging and targeting decisions autonomously, and it presents a longitudinal case study in consumer applications to test whether those gains persist.

What the paper claims

It has been reported that the authors tracked campaign outcomes over an extended period and compared autonomous policies with traditional, rule-based approaches and human-in-the-loop optimisation. Reportedly, agentic systems maintained improved engagement and conversion metrics while requiring less day-to-day oversight. The paper positions this as evidence that scalable, autonomous learning can overcome some limitations of static CRM rulesets — a major selling point for marketers wrestling with personalization at scale.

Why it matters — and why to be cautious

If replicated, the findings matter for brands and platforms that want to automate personalization across millions of users. But the result arrives amid growing regulatory scrutiny of automated decision-making. Data privacy rules such as the EU’s GDPR, emerging AI rules like the proposed EU AI Act, and sector guidance from agencies such as the US FTC are forcing marketers to balance performance with transparency, fairness and user consent. It has been reported that the paper discusses operational safeguards and monitoring; still, the technology raises questions about bias, auditability and cross-border data governance.

Caveat: preprint, not yet peer reviewed

As a reminder, the work is hosted on arXiv as a preprint and has not (yet) undergone peer review. Independent replication and fuller disclosure of methods and data will be crucial for firms and regulators to assess whether agentic personalisation can be safely and reliably deployed at scale.