The ICX blog covers conversational AI, prompt engineering, and CX strategy. It is written for teams that build AI customer service.
Editor’s desk
Anthropic's Skills feature lets teams package an AI capability as a small folder of instructions that loads only when relevant. It ends the mega-prompt era, makes capability portable across Claude surfaces, and turns AI governance into a file-review problem. For CX leaders, Skills are the first real reusable building block of enterprise AI.
On June 11, 2026, Anthropic named Tata Consultancy Services its first Global Premier Partner and committed Claude to 50,000 TCS associates. The deal is not a vendor announcement. It is a quiet admission that enterprise AI scaling needs more than a frontier model. CX leaders should read it as a new playbook.
Sample-based quality review was built for a few human agents handling a few hundred calls a day. It cannot keep up with a fleet of AI agents handling thousands. Modern AI agent evaluation is continuous, rubric-driven, and tied directly into deployment gates. CX leaders who build it now will catch failures before customers do. Those who keep sampling will find their failures on social media.
AI evaluation is the discipline of testing whether AI systems actually behave the way you intended. Most enterprises still treat it as an engineering afterthought. In 2026 that mismatch is the leading reason agent rollouts stall and governance reviews fail. Here is what real eval looks like and why CX leaders own a bigger share of it than they think.
Browse by practice
Each topic matches a real choice CX and AI teams face in their work.
New to ICX?
These three articles show how to judge and improve an AI customer experience.
A conversational AI maturity model maps where your system sits on the path from rule-based bots to adaptive, governed AI. It names five stages: Scripted, Assisted, Grounded, Orchestrated, and Adaptive. It is not a vendor checklist. It is a way to see the one trap stalling your current stage and the single change that moves you to the next.
The 30-minute AI CX audit is a five-dimension framework that finds what is actually broken in a chatbot. Spend 5 minutes on setup. 15 minutes having real conversations as if you were a customer. 10 minutes scoring the five dimensions: language quality, error handling, escalation design, trust signals, and resolution. Dashboards miss most of what this audit catches.
Intelligent CX combines conversational AI, prompt engineering, and CX strategy to design customer experiences that resolve issues, build trust, and scale across channels. It is not a new technology category. It is a discipline that treats language, escalation, and measurement as design problems instead of vendor configuration tasks.
The latest
Showing 1–15 of 76 articles
The AI industry has named five engineering terms in four years. Prompt engineering came in 2022, context engineering in June 2025, then harness, loop, and graph engineering through 2026. The layers of work are real. Some of the names are marketing. Context engineering has frontier-lab backing. Harness engineering is gaining credible adoption. The two newest terms are still contested. A business leader should map problems to layers, not chase job titles.
OpenAI launched Presence on July 22, 2026. It is a managed enterprise platform for AI voice and chat agents, deployed by OpenAI's own engineers. There is no self-serve tier, no published pricing, and no small-business entry point, so a small business cannot buy it today. The useful move is to build the same readiness Presence requires. That means mapped intents, clean knowledge, clear policies, a designed human handoff, and resolution metrics.
Most AI agent pilots never reach production. Gartner expects over 40 percent of agentic AI projects to be canceled by the end of 2027. The cause is rarely the model. It is unclear goals, weak evaluation, poor workflow design, and missing governance. The teams that succeed treat these as engineering work, not afterthoughts.
AI tools cut the time from idea to prototype from months to an afternoon, so building is no longer the scarce skill. Systems thinking, seeing how a business's parts connect and predicting the ripple effects, is what companies now need. The World Economic Forum, Figma, LinkedIn, Harvard Business Publishing, and Forbes Tech Council all point in the same direction. ICX argues customer experience work fails at those same connection points, not the tool itself.
Effective enterprise customer service chatbots are designed, not just deployed. The best practices: build from real intents, design the conversation before the prompts, set guardrails, plan fallback and human escalation, and measure resolution instead of containment. The model matters far less than the design underneath it.
Claude thinking frameworks are simple prompt patterns that push the model to think differently. ELI5 turns complex ideas into plain speech. Premortem surfaces risks before launch. Steelman builds the strongest counter argument. Red Team attacks your plan. Used well, they help leaders make sharper decisions and ship better AI experiences.
Intent architecture is the structured map of everything a conversational AI is meant to handle. It groups what customers actually say into intents, links each intent to an action and a handoff, and adds fallback paths for the unexpected. Good intent architecture is what makes a bot resolve issues instead of guessing.
A conversational AI maturity model maps where your system sits on the path from rule-based bots to adaptive, governed AI. It names five stages: Scripted, Assisted, Grounded, Orchestrated, and Adaptive. It is not a vendor checklist. It is a way to see the one trap stalling your current stage and the single change that moves you to the next.
AEO (answer engine optimization) is shaping content so AI features pick it as the direct answer. GEO (generative engine optimization) is shaping content so AI chat tools cite it inside their generated replies. Both aim for the same prize: being the trusted source an AI shows users. This guide explains both in plain English.
On every turn, an AI assistant should pick one of seven moves: answer, ask, verify, repair, stop, escalate, or hand off. The move it picks shapes the whole experience. This guide gives a plain trigger for each one, so a bot acts well when it is unsure, limited, or asked to do something risky.
Loop engineering is the practice of designing AI agents that act, observe the result, decide what to do next, and repeat until a real goal is met. It is the skill replacing one-shot prompting. This guide explains the agent loop in plain words, shows real examples, and maps the new jobs it is creating.
A chatbot can run on rules, on a large language model, or on a hybrid of both. The right choice depends on the real system, not the demo. This guide shows how to decide per intent using six tests: determinism, compliance, latency, channel, memory, and backend access. Most strong systems are hybrid.
On June 11, 2026, Anthropic named Tata Consultancy Services its first Global Premier Partner and committed Claude to 50,000 TCS associates. The deal is not a vendor announcement. It is a quiet admission that enterprise AI scaling needs more than a frontier model. CX leaders should read it as a new playbook.
Sample-based quality review was built for a few human agents handling a few hundred calls a day. It cannot keep up with a fleet of AI agents handling thousands. Modern AI agent evaluation is continuous, rubric-driven, and tied directly into deployment gates. CX leaders who build it now will catch failures before customers do. Those who keep sampling will find their failures on social media.
AI evaluation is the discipline of testing whether AI systems actually behave the way you intended. Most enterprises still treat it as an engineering afterthought. In 2026 that mismatch is the leading reason agent rollouts stall and governance reviews fail. Here is what real eval looks like and why CX leaders own a bigger share of it than they think.
