The critical distinction between failed and successful AI implementations in customer service centres on whether organisations treat technology adoption as isolated efficiency projects or as fundamental operational redesign. Zendesk's Sarah Al-Hussaini articulates this clearly: teams that layer disconnected automation initiatives—chatbots here, workflow tools there—end up with fragmented systems that underperform precisely because they lack the data integration that makes modern AI systems intelligent. The real transformation separates companies that start with a coherent long-term vision from those chasing quick wins. This raises an immediate question for teams already running Agentforce or similar platforms: are you optimising isolated use cases, or have you mapped how each automation feeds into a unified knowledge architecture that compounds AI capability over time? The answer determines whether your 12-month roadmap yields sustainable competitive advantage or technical debt.
The cultural barrier outweighs technological constraints entirely. Al-Hussaini identifies AI literacy as the decisive factor—not access to tools, but organisational willingness to hire for it, embed it into hiring processes, and fundamentally shift how problems are solved. This manifests in a specific operational risk: teams that default to rule-based, deterministic systems under the guise of risk management are actually limiting their AI's ability to handle real-world variation. Agentic systems that make independent decisions within defined parameters deliver exponentially better outcomes, but only if leadership accepts the observability and audit requirements that come with that autonomy. For support leaders, this means the next 18 months will separate organisations with high AI literacy from those without—not through feature parity, but through compounding operational advantage.
The practical roadmap Al-Hussaini outlines—starting with knowledge base automation to achieve 30% deflation, then building end-to-end procedures for specific use cases, then upskilling human agents with AI-assisted tools—is deliberately sequenced to prove ROI quickly whilst building toward sustainable transformation. The overlooked step is the knowledge health phase: most organisations discover their help centres are fragmented and outdated only after deploying an AI agent. This creates a critical dependency: your automation ceiling is your knowledge quality ceiling. For teams managing outsourced support or distributed agent pools, the handoff problem becomes acute—AI agents become more capable whilst human agents lack access to the same information, creating brand inconsistency and rework. The implication is stark: you cannot scale AI-powered resolution without simultaneously modernising how human agents access and apply information, which requires change management investment that many teams underestimate.
Lessons from the front lines of AI-powered service transformation Raconteur