Live Chat Support Outsourcing
Dedicated chat agents with concurrency set for your traffic and complexity, not for a platform default.
Page guide
On this page
Chat works until it does not
Live chat is the lowest-cost channel per contact and the easiest to ruin. It looks like a small commitment - add a widget, staff it from existing headcount - and then the queue arrives. Response times slip past the point where anyone waits, agents juggle six conversations and give short answers to all of them, and the channel that was supposed to reduce phone volume starts generating it.
The number that decides whether chat works is concurrency: how many conversations one agent can hold before quality degrades. Get it wrong upward and every conversation suffers. Get it wrong downward and you have paid for idle agents.
RHI provides dedicated full-time chat agents with concurrency set for your traffic and your complexity, not for a platform default.
What the team handles
- Pre-purchase questions - sizing, availability, compatibility, delivery timescales, the specific thing blocking a decision.
- Order and account queries - status, changes, cancellations, billing questions.
- First-line troubleshooting - the faults resolvable in a conversation, with clean escalation for the rest.
- Proactive chat - triggered on defined behaviour such as a stalled checkout, where you want it and only where it helps.
- Complaint intake - captured properly first time so it does not have to be reconstructed.
- Handover to phone or email - when a conversation has outgrown the channel, which some do.
Concurrency, set properly
Concurrency is a quality setting disguised as a capacity setting. Simple, high-volume queries support more simultaneous conversations. Diagnostic or emotionally loaded conversations support fewer - sometimes one.
We set it per queue rather than per agent, monitor whether resolution quality holds as it rises, and adjust. Platform defaults are usually optimistic because they are set to make the licence look efficient rather than to protect the conversation. Where you have a mixed queue, we route the simple contacts to higher-concurrency handling and keep the difficult ones on lower, which is more effective than one compromise setting.
How a chat is handled
- Greeting inside the target response time. A chat unanswered for two minutes has usually already lost the customer.
- Identification against your verification rules before anything account-specific is discussed.
- Context. Prior contacts across every channel are visible, so the customer is not asked to repeat themselves.
- Resolution using your knowledge base, with a plain statement of what happens next.
- Disposition recorded against defined reason codes, because free-text-only notes make reporting worthless.
- Follow-up where a promise was made, scheduled and owned rather than left to memory.
Canned responses, used honestly
Templates keep response times down and wording consistent. They also produce the single most common chat complaint: an answer that is clearly pre-written and does not address the question asked.
Our rule is that a template is a starting point an agent must adapt, never a reply sent unchanged unless it genuinely fits. Quality review scores this specifically - a technically correct canned answer to a question the customer did not ask loses the point.
Escalation
We agree before launch what an agent resolves, what goes to a senior agent, what goes to you, and what triggers an immediate alert. Some contacts should never queue: a vulnerable customer, a safety issue, a regulatory complaint or a data protection request is flagged and routed straight away.
Where chat is the wrong channel for the problem, the agent moves it to phone or email with the context attached rather than persevering in a medium that is not working.
Languages and coverage
Written support is a distinct skill from voice, and we assess it separately. From India we staff English, Hindi, Hindi-English bilingual, Bengali, Nepali and Punjabi chat support, with further Indian languages on request. See Hindi customer support for how a single language is recruited and reviewed - chat in particular requires reading messages typed in Roman script with inconsistent spelling and English mixed in, then replying in whichever script the customer chose.
Chat volume rarely follows office hours. Coverage is built from shift units - eight hours is one, twelve is two, weekends need enhanced staffing, round-the-clock is three daily teams plus a holiday plan.
Systems
Agents work in your chat platform and help desk wherever possible. Our teams are experienced across the common platforms including Zendesk, Freshdesk, Zoho Desk, Intercom and Gorgias, alongside commerce platforms such as Shopify, WooCommerce and Adobe Commerce. Those are platforms we are competent in, not partnerships. See our technology page for the full picture.
Recruitment, training and quality
We hire for written clarity, reading comprehension under time pressure, and the ability to sound like a person while working quickly. Training covers your product, systems, tone, policies and escalation rules, and ends in assessment before anyone takes a live chat.
Chats are sampled and scored against a scorecard agreed with you, with reviewers calibrated against each other. Reporting covers first response time, average response time within a conversation, resolution rate, concurrency actually achieved, customer satisfaction and quality score. Figures published elsewhere on this site are historical or representative and campaign-dependent.
Security
Access is least-privilege and role-based. Agents verify before disclosing. Chat transcripts frequently contain personal data pasted by the customer, so retention and redaction are configured to your requirements as part of scoping. RH International Ltd. is registered with the Information Commissioner's Office under reference ZB879055; we do not claim certifications we do not hold.
Going live, and what it costs
Scope, design, recruit, train, monitored pilot, stabilise, scale. Chat is usually quicker to stand up than voice, because the recruitment pool is wider and there is no accent dimension - system access remains the slowest step and it sits on your side.
Chat is priced per dedicated agent per month with a five-agent minimum, and sits below voice because concurrency means one agent covers more contacts. See the published rates and the nine things that move them.
Bring your chat volume by hour, current first response time and abandonment rate, and we can tell you what concurrency and headcount your traffic actually needs. Related: email and ticket support, omnichannel support, phone support.
Frequently asked questions
It depends on the complexity of your queue, and it is lower than most platform defaults suggest. Simple high-volume queries support more; diagnostic or emotionally loaded conversations sometimes support only one. We set concurrency per queue, monitor whether resolution quality holds as it rises, and adjust.
Yes, as a starting point an agent must adapt - never sent unchanged unless it genuinely fits. Quality review scores this specifically, because a technically correct template answer to a question the customer did not ask still loses the point.
They can, and cross-trained agents improve utilisation. But blending voice and chat degrades quality if the concurrency and interrupt rules are wrong, so we set those explicitly rather than letting the platform decide.
The agent moves it to phone or email with the context attached, rather than persevering in a medium that is not working. Which contact types trigger that is agreed before launch.
English, Hindi, Hindi-English bilingual, Bengali, Nepali and Punjabi as standing capability, with further Indian languages on request. Written support is assessed separately from voice, because they are genuinely different skills.
Yes, per agent per month, because concurrency means one agent covers more contacts. Both carry a five-agent minimum. The published rates and the variables that move them are on our pricing page.