An AI customer support agent uses AI to interpret requests, respond conversationally, retrieve information and, where integrated and authorised, perform actions or escalate cases. Capabilities range from knowledge-based answering to multi-step workflow automation.
Capabilities differ across products and integrations. Handling a multi-step refund request requires more policy context and permissions than answering a routine order-status question.
The comparison examines pricing, capabilities and practical limitations using published product documentation.
Quick answer
- Outcome-based support shortlist: Intercom Fin - standard outcomes $0.99; existing-helpdesk deployments require a 50-outcome monthly minimum. Intercom helpdesk starts at $29/seat/month plus outcomes. Its 76% resolution figure is vendor-reported, with vendor-defined measurement.
- Entry trial: Tidio Lyro - 50 non-renewing free conversations; paid conversation packages have separate terms.
- Existing Zendesk teams: Zendesk AI agents - plan-dependent included resolutions, with additional usage charges and channel entitlements.
If you want the blunt version: the deciding factor isn’t model quality. It’s whether your data is clean enough for the agent to act on.
Evaluation methodology
This comparison covers 15 platforms selected for documented capabilities, integrations, pricing transparency and business fit. Public pricing is included where available; enterprise vendors without published rates are marked quote-only. The evaluation uses product documentation and vendor information rather than controlled hands-on testing.
Ada remains included based on its documented product; quote-only pricing does not disqualify a platform. Other products such as Cognigy, Sierra and Decagon may be relevant beyond this non-exhaustive selection. A failed pricing URL does not establish poor suitability.
Evaluation considers authorised actions, knowledge and data integrations, billing definitions, hosting and contractual safeguards, languages, channels and human handoffs. Vendor resolution figures use different samples and definitions; no directly comparable independent dataset is cited.
The 15 agents at a glance
| Agent | Best for | Standout feature | Starting price | Free plan / trial |
|---|---|---|---|---|
| Intercom Fin | Support workflow automation | Outcome types include configured handoffs | $0.99 standard outcome; 50/mo minimum with existing helpdesk; Intercom seats from $29/mo | 14-day trial |
| Zendesk AI agents | Existing Zendesk estates, enterprise | 80 languages, included in Suite and Support plans | $55/agent/mo (Suite Team, annual) | 14-day free trial |
| Salesforce Agentforce | Salesforce-centric support | Foundations access; usage-priced production | Flex Credits or conversation pricing; eligibility varies | Confirm included allowance |
| Freshdesk Freddy | Freshworks support teams | 500 one-time sessions on eligible accounts | $19/agent/mo annually; Omni from $29; sessions extra | One-time allowance and trial terms |
| Help Scout AI support | Knowledge-based support | AI Answers resolution billing; confirm broader Agent scope | Applicable subscription + $0.75/qualifying AI Answers resolution | New-account three-month AI Answers trial |
| Tidio Lyro | Ecommerce support | Approved-content grounding and handoff | Paid package and quotas; confirm effective rate | 50 one-time conversations |
| Chatbase | Configurable support agents | Knowledge, actions, integrations and handoff | $40/mo (Hobby); limits apply | Free plan |
| Zoho Desk AI Agents | Zoho helpdesk teams | Model integrations; edition/region limits | Express ₹420/user/mo annually, India; confirm AI entitlements | Free Edition: 3 users; AI limits separate |
| Genesys Cloud CX | Large contact centres, voice-first | Deep telephony heritage plus AI agents | $90/user/mo (CX Cloud 2) | No free plan |
| NICE CXone | Regulated enterprise, contact centres | Enterprise governance; compliance depends on deployment | $110/agent/mo (Omnichannel Suite) | No free plan |
| Parloa | European enterprise, voice automation | Voice automation; verify deployment safeguards | Quote only (consumption-based) | No free plan |
| Yellow.ai | Multilingual enterprise automation | Conversational AI and integrations | Quote; resolution definition and included usage unconfirmed | Confirm current availability |
| Ada | Large consumer-facing brands | ACX platform with playbooks for regulated sectors | Quote only (no public list price) | Demo only |
| Kustomer | Retail and commerce operations | Built around the customer timeline, not the ticket | Quote only (ROI-calculator model) | No free plan |
| Forethought | Mid-market teams, now Zendesk-owned | Self-improving agents; deploys across any platform | Quote only (no public list price) | Demo only |
Quote-only products remain eligible. Request the relevant plan, included usage, billing definitions, implementation costs and commercial terms before comparing total cost.
The 15 agents, reviewed
Intercom Fin
Intercom Fin is a standalone AI support agent that Intercom built specifically for customer service, rather than bolting a general-purpose chatbot onto a helpdesk.
It’s the most aggressively opinionated product in this list. Fin runs on Intercom’s own Apex models, custom-trained on customer-interaction data, and Intercom is explicit that it is building vertical AI for support rather than wrapping a frontier model.
Fin’s pricing FAQ distinguishes standard $0.99 outcomes from $9.99 sales qualifications. Billable standard outcomes can include resolutions, successful configured procedure handoffs, disqualifications and self-service routing. A routine transfer without a qualifying outcome is not billed. Existing-helpdesk use requires at least 50 outcomes/month; Intercom helpdesk subscriptions start at $29/seat/month plus outcomes. Pro is an optional $100/month add-on and Fin Voice uses separate pricing.
Intercom reports an average 76% resolution rate across 12,000+ customers and roughly two million weekly resolutions. These use its own sample and metric. Its claimed monthly improvements do not establish an ongoing future improvement rate for a new deployment.
Fin integrates with Intercom and other helpdesks, with channel availability dependent on configuration. Hosting, certifications, audit reports, retention, HIPAA-related contractual safeguards and SLAs must be checked for the selected service, region and agreement; HIPAA is not a security certification.
The limitation. Fin’s economics only work if your conversation volume is meaningful and your knowledge base is genuinely maintained. At low volume you’re paying per resolution on a product whose advantage is scale. And Intercom’s own copy leans hard on the Apex model story - if you want model optionality, that runs against the grain of this product.
Zendesk AI agents
Zendesk AI agents are autonomous, self-improving support agents built into the Zendesk Resolution Platform, capable of working messaging, email and voice.
Zendesk includes AI agent access in Suite and Support subscriptions with plan-dependent automated resolution allowances. Additional committed resolutions are listed at $1.50 each. Messaging, email and voice access differ by plan, and voice AI requires suitable voice functionality.
Pricing from Zendesk’s own page, paid yearly:
- Support Team - $19/agent/month
- Suite Team - $55/agent/month
- Suite Professional - $115/agent/month
- Add-ons: Copilot $50/agent/mo, Contact Center $83/agent/mo, action credits $200 per 100K actions per subscription term
Zendesk advertises multilingual support across more than 80 languages. Test quality and coverage using the customer’s actual languages, channels and support workflows; multilingual availability does not guarantee native fluency.
The most interesting development is consolidation. Zendesk’s pricing page and AI agents page both now carry “Zendesk + Forethought,” and the FAQ confirms Forethought is an acquisition that brings self-improving agents deployable across platforms beyond Zendesk itself.
The limitation. Seat costs and resolution usage must be modelled together. Following the Forethought acquisition, confirm published integration, support and migration commitments rather than assuming a specific migration risk.
Salesforce Agentforce for Service
Agentforce for Service is Salesforce’s agentic layer for customer service, priced either per conversation or per action through Flex Credits.
The pricing page is unusually transparent compared to most enterprise agent platforms, and it publishes working cost examples.
- Salesforce Foundations - eligible-customer access to specified capabilities and builders; production customer-facing usage may incur Flex Credit or conversation charges.
- Flex Credits - $500 USD per 100,000 credits. Covers customer-facing agents, employee-facing agents, Agentforce Voice, Slackbot.
- Conversations - $2 USD per conversation, for customer-facing agents.
- Agentforce add-ons - $125/user/month, unmetered employee usage.
- Agentforce Industries add-ons - $150/user/month.
- Agentforce Max Editions - from $550/user/month.
- Agentforce User License - $5/user/month (requires Flex Credits).
The unit economics are the interesting part. An Agentforce action costs 20 Flex Credits; Agentforce Voice actions cost 30. Salesforce’s illustrative vendor example, subject to the current rate card and action classification: an order-status self-service flow uses 2 actions = 40 credits = $0.20 per request, or $120/month for 20 such requests a day.
The limitation. Credit-based metering makes forecasting genuinely difficult, and Salesforce explicitly notes unused Flex Credits do not roll over into subsequent terms. Flex Credits and Conversations cannot be used in the same org, so choosing wrong means a migration.
Freshdesk Freddy
Freddy provides Freshworks AI support capabilities, with product- and channel-specific access. Freshdesk annual subscriptions start at Growth $19/agent/month, Pro $55 and Enterprise $89. Eligible new accounts receive 500 complimentary AI Agent sessions once per account, not every month. Additional packs cost $49 per 100 sessions (session FAQ).
Freshdesk Omni is a separate offering advertised from $29/agent/month annually. Chat AI Agent access is associated with Omni; Email AI Agent availability differs. The $19 Freshdesk price should not be presented as including all omnichannel functionality.
The limitation. Model recurring sessions, product edition, Copilot and other add-ons after the one-time allowance is consumed. Sessions are not necessarily resolved cases.
Help Scout Compass
Help Scout offers AI-powered support, including AI Answers with resolution-based billing alongside a Help Scout subscription. Its pricing lists $0.75 per qualifying AI Answers resolution and a three-month unlimited-resolution trial for new accounts. A qualifying resolution involves no escalation, further knowledge-base search or indication that more help is needed.
Compass branding and broader Agent capabilities should not be treated as automatically identical to AI Answers; confirm the product named in the quote and its integration scope. Help Scout’s 73% resolution and 49-check figures are vendor claims, not independently replicated results.
The limitation. Validate action-taking, account-specific retrieval and integrations for the purchased product. Knowledge-based answering and write-enabled workflows have different requirements.
Tidio Lyro
Tidio Lyro is designed for grounded responses and escalation using approved business information.
Lyro is designed to ground responses in approved business information and escalate unsupported requests. Grounding can reduce unsupported answers but does not eliminate hallucinations or guarantee accuracy.
The 50 free Lyro conversations are a one-time, non-renewing allowance. Paid packages have recurring quotas and plan-specific prices; an advertised effective $0.50 rate should not be interpreted as unrestricted pay-as-you-go billing. Confirm package commitment and integration access.
Tidio claims customers average a 75% resolution rate. Note the contrast in approach: Tidio states it’s powered by Claude (Anthropic) plus in-house models, not ChatGPT. Where Lyro can’t answer, it creates a ticket in your support system and learns from the human’s reply.
Custom enterprise work - custom model training, unlimited AI Actions and Guidances, priority SLA, SOC 2 and HIPAA compliance - is contact-sales only.
The limitation. Test unsupported questions, policy ambiguity and escalation behavior. Knowledge grounding does not guarantee that novel or multi-system requests always become correct human handoffs.
Chatbase
Chatbase is an AI support and agent-building platform with document knowledge, configurable procedures, integrations and supported actions.
Its integrations directory documents connections including Freshdesk, Zendesk, Intercom, Shopify, Salesforce, HubSpot and Stripe. Depending on configuration, agents can retrieve order data, execute supported actions and escalate to human support; voice channels are also advertised.
Pricing from Chatbase’s own pricing page:
- Hobby - $40/mo
- Standard - $150/mo
- Pro - $500/mo
- Free plan available
The limitation. Validate each connector’s permissions, action scope, human handoff and voice entitlement for the selected plan. Broader capabilities do not establish parity with every enterprise contact-centre workflow.
Zoho Desk AI Agents
Zoho Desk AI Agents let you bring your own large language model key, or use Zoho’s, to build and deploy support agents inside a helpdesk that prices aggressively.
Zoho supports model-provider integrations within a broader helpdesk. Availability, regional hosting and data processing depend on edition, data centre, provider and contractual terms; regional presence alone does not establish privacy compliance.
Pricing is shown per user/month, billed annually, with currency switching across USD, INR, EUR, GBP, JPY, SGD, AUD, AED, SAR, MXN, ZAR, NGN, KES and EGP:
- Express - ₹420/user/month, India-region annual billing; confirm AI agent access and complimentary tokens by edition/region
- Standard - ₹800/user/month
- Professional - ₹1,400/user/month
- Enterprise - ₹2,400/user/month
- Free Edition - 3 users, free forever
- Light agents - ₹345/light agent/month on Professional and Enterprise (Enterprise includes 50 free light users)
Beyond price, Zoho’s coverage is genuinely broad: instant messaging across WhatsApp, Instagram, Facebook Messenger, Telegram, WeChat and Line, plus a multilingual help centre in over 40 languages and a 15-day free trial.
The limitation. The AI layer is a feature inside a helpdesk, not a purpose-built agent product - specialist AI products place autonomous support workflows more centrally in their positioning. Zoho’s own pricing claims it can be implemented “50% faster than competitors,” which is marketing, not a benchmark.
Genesys Cloud CX
Genesys Cloud CX is an enterprise contact-centre platform with AI agents and voice AI attached, aimed at organisations where the phone still matters.
Genesys publishes per-user pricing across four tiers: CX Cloud 2 at $90/user/month, CX Cloud 3 at $115, CX Cloud 3 Voice at $155, and CX Cloud 4 at $240, billed annually. There is no free plan and no free trial.
The strategic relevance is worth flagging: Genesys has joined Sierra and Meta in an open standard for personal AI agents - a standards initiative developing interoperability specifications; participation does not establish implemented cross-platform compatibility for how agents interact with business systems. Genesys also appears as one of the vendors in that agent-commerce working group.
The limitation. Genesys suits substantial contact-centre and voice requirements. Smaller-team fit depends on interaction volume, channels, integrations, automation scope and implementation costs. Timelines vary with deployment complexity; no validated 30-agent break-even threshold is cited.
NICE CXone
NICE CXone is an enterprise contact-centre suite whose Ultimate tier bundles AI Agents for Self-Service, aimed at heavily regulated and large-scale operations.
NICE publishes the most granular per-agent pricing in this list, billed annually:
- Omnichannel Suite - $110/agent/month
- Essential Suite - $135/agent/month
- Core Suite - $169/agent/month
- Complete Suite - $209/agent/month
- Ultimate Suite - $249/agent/month, plus $0.25 per session
- AI Agents for Self-Service - included in the Ultimate Suite tier
No free plan, no public trial. This is the category’s expensive end, and it’s priced that way deliberately.
The limitation. Compare suite subscriptions with the selected AI/session charges and implementation scope. The pricing page lists $0.25/session for Ultimate; confirm its billing definition and contract applicability. Enterprise governance does not automatically establish regulatory compliance.
Parloa
Parloa is a voice-agent platform for European enterprise, where agents handle phone conversations rather than deflecting text tickets.
Pricing is consumption-based and quote-only - Parloa does not publish list pricing. That’s the honest state of things; the pricing page routes to sales.
Parloa has a significant European market presence and a voice-focused product. Confirm hosting regions, contractual safeguards, subprocessors and certifications for regulatory or residency requirements; vendor location is not proof of compliance.
The limitation. No public pricing means no way to estimate cost without a sales process, which is a real barrier for smaller teams. And committing to a voice-first platform is a bet that your support volume will be phone-heavy enough to justify it.
Yellow.ai
Yellow.ai provides enterprise conversational AI and customer-service automation. Current pricing and free-tier availability should be confirmed with the vendor, including the billable resolution definition, included usage and subscription commitments.
The limitation. No verified public entry rate is included here. Evaluate language, channel, integration, hosting and support coverage against the intended deployment.
Ada
Ada is an “agentic CX” platform focused on large consumer-facing brands and regulated sectors, with a distinct ACX (agentic customer experience) positioning.
Ada publishes no verified list price for the relevant package. Pricing remains quote-only; product and industry documentation support inclusion without a public subscription rate.
Ada publishes industry-focused solutions and guidance for insurance, financial services, ecommerce, travel and other sectors. Evaluate these against actual policies, regulatory requirements and permitted actions rather than assuming the guidance is a production-ready SOP.
Ada claims 350+ businesses as customers. Its demo pages show vendor-reported outcomes including a 42% reduction in average handle time and 75% CSAT on AI agent interactions - both customer-specific vendor claims, not audited figures.
The limitation. You cannot budget from the public site, full stop. Budget this category only after a sales conversation. And Ada’s vertical playbooks are its own framing of your industry - validate against your actual SOPs before assuming they map.
Kustomer
Kustomer is a customer-operations platform built around a unified customer timeline rather than around the ticket queue, aimed at retail and commerce.
Pricing is quote-only via an ROI calculator model - Kustomer publishes no per-seat or per-resolution list price. The pricing page is built to funnel you into a savings conversation rather than publish a rate card.
The core architectural difference is the thesis: rather than treating support as a queue of discrete issues, Kustomer models the whole customer relationship over time and puts every channel against that history. For commerce brands with complicated post-purchase flows, that’s a genuinely different design.
The limitation. Request an itemised Kustomer quotation and evaluate projected savings using the buyer’s own interaction volumes, staffing costs, integrations and baseline support metrics. The timeline model also requires workflow and migration evaluation.
Forethought
Forethought builds self-improving AI agents that learn from resolution outcomes, and since March 2026 it has been part of Zendesk.
Zendesk acquired Forethought in 2026 and positions its agents for deployment across platforms. No unsupported acquisition-value estimate is included.
Forethought’s own pricing page publishes no list price - it’s an ROI and demo-gated motion, which is standard for the mid-enterprise segment.
The technology positioning is about closed-loop learning: agents that detect failing procedures, identify knowledge gaps, and refine workflows based on outcomes rather than being hand-tuned in advance.
The limitation. Confirm the published integration roadmap, support commitments and migration requirements following the acquisition. Future packaging may change, but specific architectural merger or migration risks should not be assumed.
What actually separates an agent from a help-center search box
The useful distinction is the workflow supported: retrieval, troubleshooting, triage or authorised actions, with a reliable failure path.
An integrated workflow may authenticate a customer, check cancellation or refund eligibility, execute an authorised action and report the result. These steps require suitable permissions and policy checks; not every agent or request supports them.
That capability has a concrete architectural marker: write access to your back-end systems. Look for API connectors, MCP servers, and explicit permissions to modify records. Write access expands operational capability, while read-only systems can still support information gathering, guided troubleshooting and triage.
The second marker is clarifying-question behaviour. Real conversations are messy. An agent that asks “do you mean the June invoice or the July invoice?” is reasoning about the case. An agent that immediately guesses is guessing at scale.
The third marker is the failure path. Ask any vendor what happens when the agent is wrong. If the answer includes automatic escalation with full context preserved, that’s a mature product. If the answer is vague, test it in the trial by deliberately asking something unanswerable.
Vendor deflection claims versus measured reality
Vendor-reported resolution rates reflect the vendor’s measurement definitions and customer sample. Escalation rules, billable outcomes, case complexity and deployment maturity make direct comparisons unreliable. Without underlying datasets and sampling methods, representativeness cannot be independently established.
Fin billable outcomes include more than fully resolved conversations. Help Scout and Zendesk also have distinct resolution definitions. Intercom’s 76%, Help Scout’s 73% and Tidio’s 75% are vendor benchmarks, not guaranteed outcomes or proof of survivorship bias.
Run a pilot against the organisation’s baseline. Measure correctly resolved cases, recontacts, inappropriate actions, escalation quality, customer satisfaction and total cost - not just a vendor dashboard’s resolution percentage. Do not project continuing gains from a claimed monthly improvement statistic.
Where agents still fail - and how to design the handoff
Agents fail on long-horizon tasks, unlisted edge cases, and emotional escalation - and the fix for all three is a deliberately designed exit, not a better prompt.
Failure mode one: the multi-step request. A customer wanting a refund for an order placed across three systems is asking the agent to do something nobody scripted. This is where the agentic architectures - Fin’s Procedures, Zendesk’s self-improving loops, Parloa’s voice flows - genuinely differentiate, and where hallucination risk concentrates, because the agent is inferring the procedure rather than looking it up.
Failure mode two: missing or ambiguous information. Test whether the system asks for clarification, avoids unsupported policy claims and preserves context when escalating. Knowledge grounding reduces risk but does not guarantee correct answers or an infallible handoff.
Failure mode three: emotional escalation. Test realistic outage, refund and complaint scenarios. Reliability depends on policy context, escalation design and the deployment; universal claims about what every agent can or cannot handle are unsupported.
Failure mode four: the language gap that isn’t. Zendesk advertises more than 80 languages; Help Scout claims over 50. The honest test is whether the agent can answer a policy question - not greet in Spanish - in that language, and whether escalation survives the transition. Ask for a demo in your language and your policy domain.
Data residency and privacy safeguards depend on service, region, plan, subprocessors and data-processing agreements. Distinguish certifications and audit reports from HIPAA-related contractual commitments. Verify the applicable DPA, retention, training settings and permitted data flows before deployment.
How to design the handoff properly. Four rules, in order of impact:
- Preserve context, always. The customer must never repeat themselves. If your trial shows a customer re-explaining their issue, the handoff is broken regardless of the agent’s resolution rate.
- Set exclusions before ambitions. Define what the agent must never touch - refunds over a threshold, legal complaints, anything regulatory - before launch, not after the first incident.
- Test the failure path deliberately. Run scenarios the agent cannot answer during the trial. A product that degrades gracefully beats one that confabulates confidently.
- Instrument before you automate. Measure baseline first resolution rate, first response time and ticket volume for a month. Without that baseline you’ll never know whether the agent helped.
Who should skip this category entirely
A purchase may be premature when reliable information is unavailable, integration permissions are undefined, or human review and escalation cannot be staffed. Knowledge can come from maintained help content, historical tickets, CRM records, databases and APIs; a standalone knowledge base is not the only source.
Low support volume can reduce financial benefit, but “under 50 tickets/month” is a heuristic, not a validated break-even threshold. Model the actual subscription, minimum commitment and usage. Fin’s $100 Pro add-on is optional. Highly bespoke or high-risk requests often require human involvement.
Buy on workflow fit, authorised action scope, hosting, handoff quality and total cost. Acquisitions create continuity considerations rather than proof that independent vendors will disappear. A standards initiative does not establish implemented interoperability; confirm actual connectors, export access and migration options.
Sources
- Intercom - Intercom Pricing (accessed October 10, 2026) - https://www.intercom.com/pricing
- Fin (Intercom) - 22 reasons to hire Fin (accessed October 10, 2026) - https://fin.ai/
- Zendesk - Zendesk Pricing Plans (accessed October 10, 2026) - https://www.zendesk.com/pricing/
- Zendesk - AI Agents for Customer Service (accessed October 10, 2026) - https://www.zendesk.com/service/ai/ai-agents/
- Zendesk - Copilot (accessed October 10, 2026) - https://www.zendesk.com/service/ai/copilot/
- Zendesk - Getting started with AI agents (accessed October 10, 2026) - https://support.zendesk.com/hc/en-us/articles/6478272212506-Getting-started-with-AI-agents-Essential
- Zendesk - Enter your resolution era with Zendesk’s agentic AI (accessed October 10, 2026) - https://www.zendesk.com/blog/zendesk-insights/innovation/enter-your-resolution-era-with-zendesks-agentic-ai/
- Zendesk - Resolution is the only customer service outcome that matters (accessed October 10, 2026) - https://www.zendesk.com/blog/zendesk-insights/innovation/resolution-is-the-only-customer-service-outcome-that-matters/
- Salesforce - Agentforce Pricing (accessed October 10, 2026) - https://www.salesforce.com/agentforce/pricing/
- Salesforce - Agentforce Rate Card (accessed October 10, 2026) - https://www.salesforce.com/agentforce/rates/
- Salesforce - Agentforce Digital Wallet (accessed October 10, 2026) - https://www.salesforce.com/agentforce/digital-wallet/
- Salesforce - New Study of 2,025 Agentic AI Leaders (August 27, 2026) - https://www.salesforce.com/news/stories/agentforce-study-2026/
- Freshworks - Freshdesk Pricing - https://www.freshworks.com/freshdesk/pricing/
- Help Scout - Help Scout Pricing (accessed October 10, 2026) - https://www.helpscout.com/pricing/
- Help Scout - Agent: AI Customer Support Agent (accessed October 10, 2026) - https://www.helpscout.com/agent/
- Help Scout - AI Policies & Procedures (accessed October 10, 2026) - https://www.helpscout.com/company/legal/ai-transparency/
- Help Scout - Seasonal product updates 2026 H2 (accessed October 10, 2026) - https://www.helpscout.com/seasonal-release-2026-h2/
- Tidio - Tidio Pricing (accessed October 10, 2026) - https://tidio.com/pricing
- Tidio - Lyro AI Customer Service Agent (accessed October 10, 2026) - https://tidio.com/lyro
- Tidio - Custom AI Agents (accessed October 10, 2026) - https://tidio.com/custom-ai-agents/
- Chatbase - Chatbase Pricing (accessed October 10, 2026) - https://www.chatbase.co/pricing
- Zoho - Zoho Desk Pricing & Editions (accessed October 10, 2026) - https://www.zoho.com/desk/pricing.html
- Zoho - Zoho Desk free plan details and pricing FAQs (accessed October 10, 2026) - https://www.zoho.com/desk/pricing.html#pricing-faq
- Genesys - Genesys Cloud CX Pricing (accessed October 10, 2026) - https://www.genesys.com/pricing
- NICE - NICE CXone Pricing (accessed October 10, 2026) - https://www.nice.com/pricing
- Parloa - Parloa Pricing (accessed October 10, 2026) - https://www.parloa.com/pricing
- Yellow.ai - Yellow.ai Pricing (accessed October 10, 2026) - https://www.yellow.ai/pricing
- Ada - Book a Demo / agentic CX overview (accessed October 10, 2026) - https://www.ada.cx/demo/
- Ada - Ada industry playbooks index (accessed October 10, 2026) - https://www.ada.cx/platform/
- Kustomer - Kustomer Pricing (accessed October 10, 2026) - https://www.kustomer.com/pricing
- Forethought - Forethought Pricing (accessed October 10, 2026) - https://forethought.ai/pricing
- Zendesk - Zendesk + Forethought AI agents product page (accessed October 10, 2026) - https://www.zendesk.com/service/ai/ai-agents/
- Salesforce - CX Trends 2026 (accessed October 10, 2026) - https://cxtrends.zendesk.com/
- Intercom - 2026 Customer Transformation Report (accessed October 10, 2026) - https://intercom.com/customer-transformation-report
- Fin - Trust and reliability certifications (accessed October 10, 2026) - https://fin.ai/trust-reliability
- Fin - How We Built a World-Class Reranker for Fin (accessed October 10, 2026) - https://fin.ai/research/how-we-built-a-world-class-reranker-for-fin/
- Intercom - Fin API platform documentation (accessed October 10, 2026) - https://fin.ai/api-platform
- Fin - Fin helpdesk integrations (accessed October 10, 2026) - https://fin.ai/integrations/helpdesk
Verification notes. The selection is non-exhaustive and uses publicly available documentation. Quote-only pricing is a commercial limitation, not evidence of product unsuitability. No controlled hands-on test or independent cross-vendor resolution ranking is claimed. Additional primary sources: Fin outcome pricing, Freshworks session FAQ, Freshdesk pricing, Chatbase integrations, Lyro one-time allowance.
Help Scout - AI Answers billing and resolution definition.
