On this page · 9 sections
Summary. The buy option is now priced per outcome: Intercom's Fin agent charges $0.99 per resolution and Zendesk charges about $1.50 to $2.00 per automated resolution plus a $50 per agent monthly add-on, as of July 2026. Building a custom agent in India runs roughly ₹8,00,000 to ₹20,00,000 for a startup-grade support bot, plus monthly LLM API costs from about ₹4,000 to ₹25,000 at 5,000 conversations a month. The decision hinges on one number most teams get wrong: resolution rate, where the 2026 enterprise median deflection sits at 41.2% for tier-1 queries and purpose-built systems reach 85%. Gartner expects over 40% of agentic AI projects to be cancelled by the end of 2027, and generative AI cost per resolution to pass $3 by 2030. This guide gives the real costs on both sides and a framework to choose.
What you are actually paying for
An "AI support agent" is not one product. On the buy side you rent a vendor's agent that plugs into your helpdesk and bills per resolved conversation. On the build side you own a system: a large language model, a retrieval layer over your knowledge base and past tickets, integrations into your order, billing and CRM systems, plus the guardrails and evaluation harness that keep it safe. The sticker price hides the real driver, which is how many tickets the agent actually closes without a human.
That distinction matters because deflection and resolution are not the same thing. A platform can show 90% deflection, meaning the conversation ended without escalation, while solving only 40% of the underlying problems. Deflection measures containment; resolution measures outcome, as CX benchmarking group ClarityArc puts it. You pay for one and your customers care about the other.
Buy: what SaaS AI support agents cost in 2026
Vendor pricing moved to outcomes over the last two years. You no longer pay mainly per seat; you pay when the agent resolves something.
| Platform | Pricing model | Indicative price (2026) | Billed on |
|---|---|---|---|
| Intercom Fin | Per outcome | $0.99 per resolution | Resolution, handoff, or disqualification |
| Intercom Fin | Per outcome | $9.99 | Qualified lead |
| Zendesk advanced AI | Per automated resolution | $1.50 to $2.00 | Automated resolution, plus $50 per agent monthly |
| Zendesk | Included tier | 5 to 15 free resolutions | Per agent per month since May 2026 |
| Typical market range | Per resolution | $0.99 to $2.00+ | One charge per resolved conversation |
The trap in outcome pricing is that cost scales with success. At 3,000 resolved conversations a month and $1.50 each, a bought agent runs about $4,500 a month before add-ons, and the bill grows every time volume grows. Gartner's own forecast that generative AI cost per resolution will exceed $3 by 2030, higher than many offshore human agents, is a warning that per-unit economics get worse, not better, as models do more work per ticket.
Build: what a custom agent costs in India
Indian engineering rates make building viable earlier than in the US or UK. Market pricing for 2026, drawn from Indian development studios, spans a wide band by complexity.
| Build type | One-time build (India) | Monthly run (LLM API) |
|---|---|---|
| Rule-based FAQ bot | ₹1,00,000 to ₹4,00,000 | ₹500 to ₹3,500 (500 conversations) |
| LLM support agent (startup grade) | ₹8,00,000 to ₹20,00,000 | ₹4,000 to ₹25,000 (5,000 conversations) |
| RAG agent with actions, multichannel | ₹5,00,000+ (add ₹50,000 to ₹3,00,000 for RAG) | ₹18,000 to ₹1,00,000 (25,000 conversations) |
| Enterprise custom platform | ₹50,00,000 to ₹3 crore+ | ₹60,000 to ₹5,00,000 (1,00,000+ conversations) |
Two costs get underestimated. Each system integration, into your order tracker, payments, or CRM, adds about ₹15,000 to ₹80,000 in development, and an omnichannel agent that behaves identically on your website, WhatsApp, and Instagram can nearly double the build because each channel has its own API quirks. Maintenance on an LLM agent built at ₹15,00,000 runs about ₹2,25,000 to ₹3,00,000 a year before API usage. The build number is mostly fixed; the run cost scales gently with volume rather than per resolution.
The number that decides ROI
Neither path pays back if the agent cannot resolve enough. The 2026 benchmarks are sobering and useful.
- Enterprise median deflection for tier-1 queries is 41.2%, with the top quartile at 58.7%, per ClarityArc production data.
- AI resolution benchmarks sit at 65% to 70% for standard deployments and 85% or higher for purpose-built platforms, per 2026 resolution benchmarks.
- A mature autonomous agent handling routine tickets, such as order tracking, refunds, cancellations, and policy questions, lands around 50% to 80% resolved end to end with no human reply.
The lesson: resolution rate depends far more on knowledge-base quality and system integration than on which model or vendor you pick. A cheaper agent that resolves 70% beats a premium one that resolves 45%. This is also why our work on pgvector versus a dedicated vector database and on building agents on Amazon Bedrock AgentCore matters to the cost line: retrieval quality is the resolution lever.
Build vs buy: a decision framework
Gartner and CX analysts converge on a practical rule. Buy when your workflow is standard and already fits a helpdesk or CRM. Build when it depends on multiple systems, proprietary data, regulated processes, or long-term ownership.
"Most agentic AI projects right now are early stage experiments or proof of concepts that are mostly driven by hype and are often misapplied," said Anushree Verma, Senior Director Analyst at Gartner. The read-across for support is plain: pick the path that ships a working, well-integrated agent, not the one that looks most advanced on a slide.
| Factor | Lean buy | Lean build |
|---|---|---|
| Workflow standardisation | Standard tickets fit a helpdesk | Custom, multi-system journeys |
| Data sensitivity and DPDP | General support data | Regulated or highly personal data |
| Systems to integrate | One or two, off the shelf | Many proprietary or legacy systems |
| Escalation and logic | Simple routing | Complex, conditional escalation |
| Time to launch | Weeks | Months, with a real team |
| Cost at high volume | Rises per resolution | Mostly fixed, scales gently |
Volume is the quiet decider. Outcome pricing is cheap to start and expensive at scale; a built agent is expensive to start and cheaper per ticket once volume is high. Somewhere between a few thousand and a few tens of thousands of resolved conversations a month, ownership starts to win, which is exactly the range where a growing Indian SaaS business lands.
India-specific considerations
Support conversations are personal data, so an agent that reads order history and identity details falls under the Digital Personal Data Protection Act, 2023 (DPDP). Design for consent, data minimisation, and clear retention from the start; our DPDP engineering playbook covers what that means in code. Multilingual support is the second factor: an agent that handles Hindi and regional languages resolves more tickets in India, and that raises both build effort and token cost. WhatsApp is often the primary channel, which adds a channel integration rather than a web-only widget. None of these change the framework; they shift where the build number lands.
How eCorpIT can help
eCorpIT is a Gurugram technology organisation, founded in 2021 and assessed at CMMI Level 5, with senior-led teams that build and evaluate AI support agents end to end. We help you run the build-versus-buy maths for your ticket volume, design retrieval over your own knowledge base and tickets, integrate order, payment and CRM systems, and set up the evaluation and guardrails that hold resolution quality. We design systems aligned with DPDP requirements and can start from a bought platform, a custom private LLM deployment, or a hybrid. To scope a support agent for your volume and workflows, talk to us.
FAQ
References
- Fin AI, "Fin AI Agent Pricing" - https://fin.ai/pricing
- Intercom, "Pricing" - https://www.intercom.com/pricing
- Gleap, "Intercom Fin AI Pricing Explained: Evaluating $0.99 Per Resolution in 2026" - https://www.gleap.io/blog/intercom-fin-ai-pricing-2026
- Zendesk, "Pricing Plans" - https://www.zendesk.com/pricing/
- eesel AI, "Understanding Zendesk AI pricing: a complete pay-per-resolution guide" - https://www.eesel.ai/blog/understanding-zendesk-ai-pricing-a-complete-pay-per-resolution-guide
- ClarityArc Consulting, "AI Support Ticket Deflection 2026" - https://www.clarityarc.com/
- Richpanel, "AI Customer Service Statistics for 2026" - https://www.richpanel.com/learn/ai-customer-service-statistics-2026
- Gartner, "Gartner Predicts GenAI Cost Per Resolution for Customer Service Will Exceed Offshore Human Agent Costs by 2030" - https://www.gartner.com/en/newsroom/press-releases/2026-01-26-gartner-predicts-genai-cost-per-resolution-for-customer-service-will-exceed-offshore-human-agent-costs-by-2030
- CX Today, "How CX's Build vs Buy Debate is Changing" - https://www.cxtoday.com/ai-automation-in-cx/ai-agents-cx-build-buy-platform-strategy/
- Codingclave, "AI Chatbot Development Cost India 2026" - https://codingclave.com/blog/ai-chatbot-development-cost-india-2026
- Zethic, "AI Chatbot Development Cost in India (2026)" - https://zethic.com/ai-chatbot-development-cost-in-india-2026-full-price-breakdown/
- Ministry of Electronics and IT, "Digital Personal Data Protection Act, 2023" - https://www.meity.gov.in/data-protection-framework
- Gartner, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," June 25, 2025 - https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027
_Last updated: July 18, 2026._
