Integrating AI into Customer Experience starts with outcomes, not algorithms - Hunor Kovacs, Managing Director at Graia
Customer Experience Romania
Empowering businesses to deliver unique experiences. Transform your customers, employees and partners into your friends.
Graia develops technology solutions for contact centres and CX departments. In today’s context, how do you see the balance between automation and human interaction in customer relationships?
The right balance is simple to state and hard to do: automate everything repetitive, keep humans one click away for anything emotional, complex, or high-risk. In practice this means two design rules. First, self-service and AI should resolve routine tasks instantly, across every channel. Second, every automated flow must include a visible, seamless handoff to a human, with full conversation context carried over so customers do not repeat themselves. Modern CCaaS can already do this. For example, Graia Agentic CCaaS supports bot-to-agent escalation with auto-summaries, and even lets agents place a voice call from an active webchat when empathy matters more than speed. It also offers live voice translation so language never blocks understanding.
The outcome is not “bots instead of people,” it is “humans focused where they add human value.” Our own platform vision holds the same line: humans stay decisively in control, supported by agentic AI for pre-call, in-call and post-call tasks, with governance, audit, and memory across the journey.
AI is moving fast. The goal is not to choose between automation and people, but to set a new balance that delivers both efficiency and a human experience customers trust.
Automate every repeatable step. Keep humans one click away for moments that need judgement, empathy, or brand protection. When agentic AI and human expertise work together, each makes the other better. AI removes friction and surfaces next best actions. People carry tone, trust, and final decisions.
Done this way, you get the best of both worlds: operational efficiency at scale and a customer experience that feels personal and accountable.
Artificial intelligence is becoming increasingly present in support and communication processes. What are the most important criteria a company should consider when integrating AI into its Customer Experience strategy?
Integrating AI into Customer Experience starts with outcomes, not algorithms. Map the few journeys that drive most of your volume and define what “good” looks like for both the customer and the agent. Then automate the repetitive parts and keep a human checkpoint wherever empathy, judgement, brand risk or regulation demands it. This is the operating model we use at Graia: agentic AI handles the heavy lifting, while humans stay decisively in control through built-in human-in-the-loop steps, shared knowledge, memory and a clear path to escalate any interaction to a person.
AI governance must sit alongside CX governance from day one. Treat AI governance as a companion practice, not a later add-on. In practice that means adopting accepted guardrails and language so your C-level, Risk and Compliance teams can steward the program. The NIST AI Risk Management Framework gives a practical backbone with its Govern, Map, Measure and Manage functions, while ISO/IEC 42001 provides a management-system standard for running AI responsibly across the lifecycle. Using both keeps decisions observable and auditable while you scale.
Trust is the everyday currency. Customers accept AI when they have transparency, control and a visible human fallback, especially on complex or sensitive tasks. Current research on customer sentiment reinforces this point, and our platform design follows it closely with RBAC, cost ceilings, immutable audit and a model-agnostic approach so you avoid lock-in.
Finally, measure both CX and EX. Containment and AHT tell only part of the story. Quality, CSAT and agent effort complete it. In Graia we operationalise this with a single governance spine across the “infinity loop,” so the same policies and telemetry cover live interactions in CCaaS and the back-office processes that fulfil the promise made to customers.
You’ve worked with clients in over 20 countries. What cultural differences have you observed in the approach to CX, and how do these influence the design of technology solutions?
Culture does not decide your channels, but it shapes expectations about speed, tone, and control. In the United States, customers expect autonomy and fast self-service with clear status visibility. In some countries like Israel for example, direct communication and time sensitivity reward concise flows and immediate escalation when nuance is needed. In places like Greece, stronger uncertainty avoidance means reassurance matters, so a visible human path and clear policy explanations build trust. Closer to us, in Poland, Hungary or Romania, clarity and predictability count, so structured troubleshooting and formal confirmations help. In Romania, a relationship-oriented tone, named ownership of cases, and strong messaging support go a long way. Across all these examples, language support and a seamless handoff to a person are universal expectations.
Empathy is the common thread. My own multicultural work in the USA, Israel, Greece, Poland, Hungary, and Romania taught me that the best CX feels both global and local. Global means efficient, consistent, and data-driven. Local means speaking the language, matching the tone, and honouring norms. We can do both because Graia has tens of colleagues with deep experience across many countries, so design choices are informed by lived context, not guesswork.
Language should never be a blocker. Our platform supports live voice and chat translation with configurable language profiles and multiple providers, so we can operate in almost any language and adapt to regional variants, including those used in multilingual countries like Switzerland. Admins can select providers such as Microsoft, Google and DeepL, tune formality, and add domain phrase lists to improve accuracy. The webchat widget can also be localised through API for a growing list of languages. These controls make multilingual service practical at scale.
The takeaway for enterprises is simple: choose systems that let you customise at every level, from channel language and tone to workflows, SLAs, and policy thresholds. Local tweaks on a common platform keep you culturally precise without fragmenting your stack.
From your perspective, what are the most common challenges companies face when trying to combine technological efficiency with empathy in customer interactions?
A common criticism of technology in service is that it feels cold and transactional. Modern agentic AI changes that. Empathy in AI does not mean the system has feelings. It means the platform can recognise sentiment and context, adapt tone and pacing, and choose the right next step in a way that feels natural and builds trust. Done well, the bot handles the routine, senses when a human should step in, and carries the full context forward so the conversation continues, not restarts.
The first pitfall is local optimisation that breaks the journey. A company automates identity checks, then makes customers repeat everything after transfer. The fix is to keep context end to end. When a handoff happens, the agent sees a clear summary, the sentiment so far, and suggested next actions. The customer feels heard because nothing is repeated and the tone matches the situation.
The second pitfall is the “no escape hatch” bot. Containment looks good in a report, but if a customer cannot reach a person, they will leave. The fix is a visible path to a human on every screen, skill-based routing, and a rule that rising negative sentiment or vulnerable situations trigger a live handover. Empathy becomes operational, not a slogan.
Third, tone and language mismatches. A fast flow that uses the wrong words or level of formality will still feel wrong. Real-time translation and configurable tone help agents keep warmth while staying efficient. For sensitive cases, the system can slow the cadence, avoid cross-sell, and surface reassurance scripts that match the customer’s history and preference.
Fourth, fragmented governance. Speed without guardrails erodes trust later. One governance spine across channels and back-office tasks keeps efficiency and empathy aligned. Put approvals on high-impact steps, log decisions, and cap costs. Define when AI can act, when it must ask, and when it must hand over.
Here are simple examples that bring this to life. An insurance intake bot recognizes frustration during a claim, simplifies its questions, pauses optional data capture, and offers a warm transfer with a one-paragraph summary for the agent. A telecom billing assistant detects anxiety, suppresses upsell, presents a payment plan that respects tenure, and books a callback with the same agent if the customer prefers. A healthcare appointment bot proposes alternatives around school hours, confirms transport options, and escalates exceptions to a coordinator with all notes attached. In each case, the system is fast, the experience is kind, and a human is always one click away.
Measure what matters. Track containment and handle time, but also sentiment improvement, successful empathetic handoffs, complaint rates, and agent coaching lift. The goal is simple. Make automation quick and transparent, design empathy into the flow, and keep people in the loop wherever judgement and care are required. That is how you scale technology without losing the human warmth customers remember.
Given that this year’s edition of CX Conference Bucharest explores how companies are integrating AI into their processes without losing empathy and human warmth, what message would you share with participants undergoing digital transformation in their organisations?
Treat AI as an enterprise program, not a tool rollout. Start from your signature journeys and remove friction where it moves the needle fastest, then connect these wins under one fabric that unites CX governance with AI governance. Keep humans in the loop by design. Make the human path obvious to customers and insert review or empathy steps wherever judgment or brand tone requires it. Build once and reuse everywhere so front-office interactions and back-office execution share the same policies, audit, and cost controls. Measure trust and quality alongside containment and handle time. That is how you scale AI without losing the human warmth customers remember.
Key takeaways are clear:
North star: use AI to make human interaction more meaningful, more strategic, and more empathetic. This is the standard we hold ourselves to at Graia.