If your contact center is still routing customer chat conversations separately from voice—different systems, different agent queues, different routing logic—you are fragmenting the customer experience while multiplying operational complexity. 

Most enterprise contact centers handle voice calls through a mature telephony infrastructure. But chat is new. Newer than the systems built to integrate it. So chat often ends up disconnected: a separate vendor, separate agents, separate workflows. A customer calls in, gets disconnected, then has to start over in chat. An agent handling voice cannot take a chat hand-off because the systems don’t communicate. A supervisor has visibility into voice queue depth but is blind to chat volume until customers complain. 

The decision to unify chat with voice through omnichannel infrastructure is where “build vs buy vs configure” becomes real. And the choice you make determines whether chat integration takes 2 months or 12 months, whether you own a solved problem or a forever maintenance liability, and whether your contact center actually improves or just gets more complicated. 

In this article, I want to walk you through the real decision framework for enterprise live chat: what build means in practice, what buy means beyond the headline cost, what configure-and-extend means as a middle path, and how to evaluate vendor AI chatbot capabilities that are increasingly central to chat deployment. 

The Live Chat Problem: Why Enterprise Contact Centers Treat It Separately 

The central problem with enterprise live chat is not that chat is complex. It is that chat arrived after voice infrastructure was already built, and retrofitting is harder than building from scratch. 

When voice systems were built (15-20 years ago), the architecture was simple: calls arrive, routing rules send them to queues, agents answer, calls end, records close. This architecture scaled. It still powers most enterprise contact centers. 

Chat doesn’t fit this model. Chat is asynchronous. A customer starts a conversation. The agent responds minutes later. The customer goes away. Hours pass. The customer returns with a follow-up. An agent (possibly different) responds. 

This asynchronicity breaks voice infrastructure. Queues assume synchronous connections. Routing rules assume agents will be continuously available. Agent idle states assume the agent is waiting for the next inbound contact, not managing five conversation threads simultaneously. 

So chat gets bolted on: a separate platform, separate agents (or blended agents), separate reporting, separate routing. The result is operational fragmentation: 

• Customers experience disconnected interactions (voice context doesn’t transfer to chat; chat context doesn’t transfer to voice) 
• Agents cannot own full customer relationships (specialists in voice, different specialists in chat) 
• Supervisors manage voice and chat as separate operations 
• Routing becomes manual or rule-based rather than optimized 
• Compliance requirements vary across channels (voice has call recording; chat has transcript rules; different vendor = different implementations) 
• Analytics fragment (voice metrics in one system, chat metrics in another) 

The enterprise live chat decision is whether to accept this fragmentation or invest to unify it. 

Build vs Buy vs Configure: The Real Trade-offs 

When evaluating enterprise live chat, organizations typically face three paths: build a proprietary system, buy from a vendor, or configure an existing platform. Each has real trade-offs that most vendors won’t explain clearly. 

Build: Full Control, Full Liability 

Building proprietary live chat means engineering a system that: 

• Handles asynchronous conversations across multiple concurrent threads per agent 
• Routes conversations based on agent skills, queue depth, and customer context 
• Maintains conversation history and transfers context between agents 
• Scales to thousands of concurrent conversations 
• Integrates with existing CRM, knowledge base, and backend systems 
• Includes AI-powered chatbots for pre-qualification and routine questions 
• Provides real-time supervisor visibility and intervention capabilities 
• Tracks compliance and produces audit logs 

Estimated effort: 12-18 engineering months for a functional system, 24-36 months for a production-hardened system that handles edge cases and enterprise reliability requirements. 

Cost: 8-12 senior engineers × $150K-$200K fully loaded = $1.2M-$2.4M per year, for 2-3 years = $2.4M-$7.2M total investment. Plus infrastructure, plus maintenance. 

The advantage of build: complete control over architecture, no licensing fees post-development, ability to embed contact center-specific logic. The contact center becomes a competitive differentiator if you embed it in a product. 

The disadvantage of build: you are not a chat software company. You cannot scale engineering investment like Zendesk or Salesforce can. Your system works for your use cases but breaks on others. You own 100% of the bugs. When your engineers leave, knowledge walks with them. 

Build is rational only if you are a software company trying to embed chat in a broader product platform, or if you have persistent engineering capacity you cannot deploy elsewhere. 

Buy: Speed, No Control, Vendor Lock-in 

Buying a live chat platform (Zendesk, Intercom, Freshdesk, etc.) means: 

• Deploying a mature, battle-tested system in weeks 
• No engineering investment on core platform 
• Out-of-the-box integrations (with major CCaaS systems, CRMs, knowledge bases) 
• Vendor-managed infrastructure, security, compliance, scale 

Estimated implementation timeline: 4-8 weeks for basic deployment, 8-12 weeks for integration with voice systems. 

Cost: Per-agent licensing ($50-$150/agent/month) + implementation ($20K-$50K) + customization/integration ($30K-$100K) + annual subscription. 

The advantage of buy: speed to market, vendor handles maintenance, vendor handles scaling, integrated security and compliance, professional support. 

The disadvantage of buy: you accept the vendor’s architecture and design decisions. Want a feature they don’t build? You wait (or pay for custom development). Want to change vendors? Your workflows, integrations, and customizations are stranded. Licensing costs scale with agent count, creating cost pressure as you expand. 

Buy is rational when you want live chat fast, when you are not a software company, and when you accept that chat integrates alongside voice rather than within the same system. 

Configure: Configurable Platform as Middle Ground 

Configurable platforms (like ETS Labs ICE Communication) sit between build and buy: 

• Pre-built core platform (async conversations, routing, escalation, AI chatbots) 
• Designed for omnichannel from day one (voice, chat, email, social all native) 
• Highly configurable workflows, routing logic, and integrations 
• Can be extended with custom code if needed 
• Vendor handles core platform maintenance; you configure the implementation 

Estimated implementation timeline: 8-16 weeks (longer than pure-buy because configuration is involved; shorter than build because the platform already exists). 

Cost: Platform licensing (seat-based or volume-based) + implementation services ($50K-$150K for enterprise) + optional custom development if you need functionality outside configuration. 

The advantage of configure: faster than build, more control than buy, true omnichannel integration, vendor handles core platform evolution, implementation is reversible (if you want to change configuration, you don’t rewrite systems). 

The disadvantage of configure: requires discipline in planning (you must define requirements clearly before implementation), requires experienced platform team on vendor side, not a plug-and-play solution. 

Configure is rational when you want control and true omnichannel integration, when you can invest 8-16 weeks in implementation, and when you want to avoid build’s engineering burden. 

The Decision Framework 

Choose BUILD if: 

• You are a software company and chat is a core product feature 
• You have persistent engineering capacity you cannot deploy elsewhere 
• You need chat embedded in a proprietary system 
• You are willing to spend $2M-$7M over 2-3 years 

Choose BUY if: 

• You want live chat with minimal implementation effort 
• You are comfortable with chat as a separate system from voice 
• You can accept vendor architecture and design decisions 
• You value vendor maintenance and support 

Choose CONFIGURE if: 

• You want true omnichannel integration (voice + chat unified) 
• You can invest 8-16 weeks in careful implementation 
• You want control over workflow logic and routing 
• You want to avoid build’s engineering cost while maintaining integration 

AI Chatbot Capabilities: What Actually Matters 

Enterprise live chat increasingly means AI chatbots handling initial interactions. Customers expect to chat with a bot first, get routed to an agent only if needed. 

Most vendors make chatbot claims that sound similar: “AI-powered,” “natural language understanding,” “learns from interactions.” The actual capabilities vary significantly. 

What AI Chatbots Can Actually Do (Today) 

Rule-based bots: Match incoming messages to predefined patterns. If customer says “password reset,” trigger password reset workflow. Fast, reliable, limited. Cannot handle variations or context. 

Intent classification bots: Use machine learning to categorize what a customer is asking, then execute workflows. “Customer wants to reset password” → trigger reset workflow. More flexible than rule-based. Can handle variations. Still cannot understand nuanced context. 

Contextual understanding bots: Understand customer history, account status, and context. “Customer wants reset AND has high account value AND contacted 3 times this month” → escalate to agent instead of automating. More sophisticated. Requires training data. 

Generative bots: Use large language models to generate responses in natural language. Can handle open-ended questions, provide explanations, and engage in conversation. Powerful but risky (can generate inaccurate information, leak sensitive data). 

The Practical Reality 

Most enterprise deployments use a combination: 

• Rule-based and intent classification for high-confidence, high-volume interactions (password reset, billing inquiries, order status) 
• Contextual understanding to avoid automating interactions that need human judgment 
• Escalation to agents when confidence drops or when customer emotion is high 
• Generative language models for specific, controlled use cases (knowledge base queries with citation) 

The bot quality depends more on training and configuration than on the underlying technology. A well-trained intent classifier beats a poorly-configured large language model. 

What to Evaluate 

• Training data: Has the vendor trained the system on contact center data, or on generic internet text? Contact center-specific training is critical. 
• Escalation confidence: Can the bot decline to handle interactions where it is uncertain? This prevents bad experiences. 
• Context awareness: Can the bot access customer history and account data to inform decisions? 
• Human override: Can agents easily take over bot conversations or correct bot responses? 
• Learning feedback: Does the system improve based on agent feedback on bot performance? 
• Audit trail: Can you see what the bot said and why? (Critical for compliance) 

Avoid vendors who claim their bot handles “any question” or who emphasize the size of their language model. Enterprise contact centers need controlled, reliable automation, not impressive-sounding generative language. 

Integration Complexity: The Hidden Cost 

The largest hidden cost in enterprise live chat is integrating with existing contact center systems. 

Chat needs to connect to: 

• CCaaS platform (for agent assignment, reporting) 
• CRM (for customer context, history) 
• Knowledge base (for FAQ resolution) 
• Workforce management (for scheduling, staffing) 
• Quality assurance system (for chat monitoring, compliance) 
• Billing/backend systems (for account lookups, transaction execution) 

Each integration requires: 

• Authentication between systems (OAuth, API keys, service accounts) 
• Data mapping (chat customer ID = CRM customer ID; sometimes non-obvious) 
• Real-time data flow (customer types a message → route to agent → agent sees it within 2-3 seconds) 
• Error handling (what happens if CRM is slow or unavailable?) 
• Testing across integration scenarios (does chat work if CRM is down? How does the agent experience degrade?) 

For a typical enterprise with 5-8 integrated systems, integration work consumes 30-50% of implementation timeline. 

Vendors handling this well (like ETS Labs ICE Communication) have pre-built connectors to major platforms and mature integration patterns. This compresses timelines by months. 

Vendors without integration depth require custom development, which creates long timelines and permanent maintenance burden. 

Omnichannel Unification: The Real Value

The decision to unify live chat is ultimately about whether contact center operations improve. 

Unified omnichannel means: 

• One agent queue (voice, chat, email, social) 
• One routing engine (same logic evaluates all channel types) 
• One customer view (agent sees conversation history across all channels) 
• One supervisor dashboard (visibility into all channels) 
• Consistent workflows (compliance requirements apply uniformly) 

The value is operational: fewer agents needed to handle the same volume (agents can switch between channels; don’t need voice specialists and chat specialists), better customer experience (context transfers between channels), simpler operations (one routing system, one QA process, one training program). 

The cost is implementation complexity: unification requires careful architecture, vendor maturity, and clear planning. 

The Decision Framework For Your Organization 

For enterprise contact centers evaluating live chat: 

If chat is new: Start with BUY or CONFIGURE, not BUILD. Speed to market matters more than control at first. 

If you want omnichannel: CONFIGURE is almost always better than BUY (pure-buy platforms rarely unify chat and voice well). 

If you have specific customization requirements: CONFIGURE with ability to extend is better than pure BUY (which limits customization). 

If you are already building contact center software: BUILD might make sense. 

Cost is not the primary driver: Configure and buy have similar lifecycle costs. Build saves licensing fees but costs massive engineering investment. 

Timeline is critical: Buy is fastest (4-8 weeks), configure is moderate (8-16 weeks), build is slowest (24-36 months). 

ICE Communication: Omnichannel Designed For Enterprise 

ETS Labs ICE Communication is a configurable omnichannel platform designed to unify voice, chat, email, and social within a single contact center infrastructure. Unlike bolt-on chat platforms, ICE treats chat as a first-class channel alongside voice—same routing logic, same agent experience, same reporting, same compliance framework. 

Key differentiators: 

• True omnichannel: Chat and voice unified in the same platform; agent can switch between channels; routing logic applies uniformly 
• AI chatbot native: Pre-built chatbot capabilities for common use cases; configurable without custom development 
• Enterprise integration: Pre-built connectors to major CCaaS platforms, CRMs, and business systems 
• Configurable workflows: Define routing, escalation, and automation without engineering 
• Built in production: Designed inside Etech Global Services’ own enterprise contact center operations 

Learn more: https://etslabs.ai/products/ice-communication/ 

Frequently Asked Questions 

Should we build proprietary live chat? 

Only if you are a software company and chat is core to your product. Enterprise contact centers should buy or configure, not build. 

What is the difference between a chatbot and live chat? 

Live chat is the communication channel (customer and agent exchanging text messages). A chatbot is software handling some of those conversations automatically. Modern platforms combine both: bots handle routine questions; agents handle complex ones. 

How do we integrate chat with our voice system? 

This depends on your platform. Omnichannel platforms like ICE Communication integrate natively. Bolt-on chat platforms require custom integration via APIs, which takes weeks of development. 

How much does enterprise live chat cost? 

Buy: $50-$150/agent/month licensing plus integration. Configure: Licensing plus implementation services ($50K-$150K). Build: $2M-$7M over 2-3 years. 

Can AI chatbots replace agents? 

No. Chatbots handle 20-30% of interactions (routine, high-confidence questions). Agents handle 70-80% (complex, emotional, contextual interactions). 

From Disconnected Chat To Unified Omnichannel 

The choice between build, buy, and configure is ultimately about whether your contact center operates as a unified system or a collection of disconnected tools. 

Unified omnichannel improves customer experience, reduces operational complexity, and enables agents to own customer relationships across channels. Disconnected chat remains a separate operation, separate cost, separate training, separate headaches. 

If you are ready to unify chat with voice and create a true omnichannel contact center, the question is not whether to build (you shouldn’t), but whether to buy a bolt-on platform or configure an omnichannel system designed from the ground up for integration. 

The answer to that question depends on your timeline, your integration requirements, and whether you want chat as an addition to contact center operations, or chat as part of a unified customer communication strategy. 

Manu Dwievedi

Manu Dwievedi is Vice President of Product Strategy & Innovation at ETSLabs and Etech Global Services, where he leads the development of AI-powered interaction analytics platforms including QEval®, Real-Time Agent Assist, Voice AI, and Process Automation. These platforms process over 2 billion interactions annually across Fortune 500 environments.