Manifesto: Stop Letting AI Vendors Experiment on Your Contact Center

The Real AI Problem in Contact Centers

The underlying AI capabilities exist. Speech recognition now handles noisy, real-world calls with a level of accuracy that supports automation and analytics. Large language models generate human‑like text at speed and scale. Pattern detection across millions of historical interactions is feasible and financially accessible. The technology base no longer represents the primary constraint.

The breakdown appears in how the industry packages, positions, and deploys that technology into live operations. Vendors frequently ship general models wrapped in contact center language, without putting in the work to understand routing logic, regulatory constraints, legacy system behavior, and frontline realities. Buyers often accept broad claims instead of demanding specific commitments tied to operational metrics. Pilots start without defined baselines or success thresholds, which makes every outcome debatable. Agents and supervisors receive new tools that add steps, screens, and alerts, yet they rarely see a direct reduction in workload or stress. The gap does not come from lack of AI power. It comes from lack of discipline in how we apply it.

Why We Backed a Hard Standard

When Justin Robbins and I aligned on the CX AI Reality Standard, we agreed that it needed to function as a filter, not as a set of inspirational guidelines. The standard states that AI in a contact center must be purpose‑built, proven and explainable, and people‑centered before it earns access to customers and agents. Anything that fails one of those tests does not qualify, regardless of how impressive it looks in a sales presentation.

From a CEO perspective, soft guidance leads to predictable patterns. AI portfolios expand while measurable impact stays flat, because organizations layer tools instead of retiring underperforming ones. Governance arrives late, after tools already process customer data and influence decisions, which forces legal and security teams into defensive posture. Frontline teams grow cynical, because each new AI introduction brings more complexity without a corresponding lift in the quality of their workday. A hard standard shifts that dynamic. It introduces clear criteria that every vendor and internal team must meet before the organization invests time, money, and trust.

Using the Standard as a Mirror for Vendors

Vendors, including Diabolocom, need to treat the CX AI Reality Standard as a mirror. If purpose‑built design feels restrictive, that raises a real question about how much of the product depends on generic models with light customization instead of deep integration into contact center workflows and data. If proof and explainability feel like an excessive burden, that signals a lack of investment in measurement, governance, and observability. If the idea of “people‑centered” design sounds cosmetic, that exposes how little time anyone has spent sitting with agents and supervisors to watch them work with the product in real time.

I understand the pressure to ship features and to show rapid progress to investors and early customers. I understand the temptation to rely on slideware and highlight reels when deep case studies require patience and partnership. The standard forces a more honest conversation. It reveals which vendors have invested in real‑world performance and which ones lean on generic technology, polished demos, and aspirational claims. Vendors that take this seriously benefit as well. They compete on depth, reliability, and operational alignment instead of competing against promises that no one can deliver.

Helping Buyers Set AI on Their Own Terms

Contact center and CX leaders hold the budget and carry responsibility for what happens to customers and employees. They decide which AI enters production. The CX AI Reality Standard gives those leaders a practical tool to run that selection on their own terms, rather than on vendor terms.

In practice, that means asking direct, specific questions in every conversation and insisting on concrete answers. When a vendor says that its AI trains on massive data sets, leaders should ask to see the proportion and relevance of true contact center data, and how closely that data mirrors their own channel mix, languages, and acoustic conditions. When a vendor describes performance with broad claims such as “up to X percent improvement,” leaders should demand before‑and‑after results from operations similar in scale and complexity to their own. When a vendor leans on “continuous learning,” leaders should ask for a detailed description of the feedback loops, the model re-training, the human roles in those loops, and the guardrails that govern model updates. Specific answers enable smart risk decisions. Vague responses signal that the tool is not ready for a live environment with real customers.

Contact center and CX leaders hold the budget and carry responsibility for what happens to customers and employees. They decide which AI enters production. The CX AI Reality Standard gives those leaders a practical tool to run that selection on their own terms, rather than on vendor terms.

In practice, that means asking direct, specific questions in every conversation and insisting on concrete answers. When a vendor says that its AI trains on massive data sets, leaders should ask to see the proportion and relevance of true contact center data, and how closely that data mirrors their own channel mix, languages, and acoustic conditions. When a vendor describes performance with broad claims such as “up to X percent improvement,” leaders should demand before‑and‑after results from operations similar in scale and complexity to their own. When a vendor leans on “continuous learning,” leaders should ask for a detailed description of the feedback loops, the model re-training, the human roles in those loops, and the guardrails that govern model updates. Specific answers enable smart risk decisions. Vague responses signal that the tool is not ready for a live environment with real customers.

Why Diabolocom Accepted Public Evaluation

For Diabolocom, supporting the CX AI Reality Standard required a decision to stand under the same spotlight that we recommend for others. We agreed from the outset that Justin would apply the standard directly to our platform, in public view. That choice triggered focused internal work.

Our teams reviewed where we already demonstrate contact center specificity through training data, architecture, and workflow design, and where we needed to deepen that alignment. We examined our customer results and identified where we hold strong, publishable evidence of impact on QA coverage, after‑call work, handle time, and customer satisfaction, and where we must invest further in instrumentation and joint measurement with clients. We scrutinized our features from an agent and supervisor perspective and looked for places where we still created unnecessary steps or cognitive load. These reviews sharpened our roadmap and clarified where we must improve. Our customers deserve that level of introspection and accountability from us. The standard gave us a precise framework for it.

What Needs to Change in the Next Year

If contact center leaders and vendors adopt this standard with conviction, the AI landscape inside operations will look different over the next twelve months. Organizations will run fewer pilots, but each pilot will have a sharper scope, quantified baselines, defined success criteria, and clear timelines. Vendor shortlists will shrink because early questions about data, governance, and people impact will eliminate unprepared options before they consume attention and budget. Collaboration between operations, IT, legal, security, and HR will improve because everyone will share a structured way to discuss AI risk, value, and readiness.

Most importantly, frontline teams will feel the difference. Agents and supervisors will be able to describe in straightforward terms how AI changed their work. They will talk about fewer clicks, clearer guidance, faster resolution, and better coaching. Our research points to a broader truth: the future of customer experience will not be defined by autonomous AI replacing people. It will be defined by organizations that successfully combine human judgment, empathy, and creativity with AI-powered capabilities. The goal is not to build fully autonomous contact centers. It is to create exceptional human agents—what I often think of as “super-human” employees—equipped with the knowledge, context, and support to deliver experiences neither humans nor AI could achieve alone. When that happens, AI stops being a corporate narrative and becomes an operational asset. 

At Diabolocom, we will continue to build under the CX AI Reality Standard and invite customers and prospects to evaluate us with it. We expect critical questions and detailed scrutiny. We design for that level of examination. Artificial intelligence belongs in the contact center. The standard exists to ensure that only systems with the discipline, transparency, and operational fit required for this environment earn that place.