Yes. You do not need to switch helpdesks to get an AI agent: several AI layers sit on top of Zendesk or Intercom, read and reply to tickets in your existing queue, and hand off to your team there. Lorikeet, Decagon, Maven AGI and Forethought publish connectors for both, Ada for Zendesk, and Fin (formerly Intercom, now part of Salesforce) is built into Intercom and also sold as a layer for Zendesk. The alternative is the AI that comes with your helpdesk, such as Zendesk AI agents.

Key takeaways

• A layer is a delivery model, not a niche. Fin is sold for "any helpdesk", and Zendesk says its Forethought agents work off-platform too, so the choice is between agents, with your helpdesk held constant.

• Built-in AI is the cheapest route to answers. Zendesk includes AI agents in every current Suite and Support plan, billed per automated resolution (Zendesk pricing).

• A layer earns its fee on actions. If tier 1 means changing a payment date or checking a claim in another system, compare layers on those tickets.

• Lock-in moves, it does not vanish. Built-in AI ties you to the helpdesk; a layer ties your workflows to the layer. Keep your SOPs and APIs as your own.

• Decision rule: if more than half your tickets need an action outside the helpdesk, trial a layer; otherwise switch on the built-in AI first.

Which tools sit on top of Zendesk or Intercom?

Lorikeet works inside the ticketing system you already run and takes action through your own APIs, with built-in connectors for Zendesk, Intercom, Salesforce, HubSpot, Front and more, so there is nothing to migrate (Lorikeet integrations). Decagon syncs tickets from Salesforce, Intercom and Zendesk and handles email through Zendesk and Intercom (Decagon integrations). Maven AGI connects to Intercom workspaces and installs from the Zendesk Marketplace (Maven AGI for Intercom). Fin is the reverse case: built into the Intercom helpdesk, and added to Zendesk over API (Fin for Zendesk).

How we chose and checked these tools

We included tools that resolve customer conversations autonomously and either connect to Zendesk or Intercom as a separate layer or come built into one of them. Every vendor fact was checked on the vendor's own public pages on 30 Oct 2026 and is linked in its entry. Where a vendor does not publish something, the table says "Not published". This compares what vendors publish, not a hands-on benchmark, and it is ordered by fit for a team asking this question while keeping its current helpdesk.

Lorikeet publishes this guide and is one of the tools listed. We have tried to describe every product, including our own, using only what each vendor publishes, and we link to the source for each claim.

Do I need an AI layer or a helpdesk with built-in AI?

You need a layer when the hard part of your tickets happens outside the helpdesk; otherwise built-in AI is usually enough. Built-in AI is one product: the AI is a feature of the ticketing platform, reads the same help centre and reports in the same dashboards. A layer is a second vendor that works the queue like a teammate, acting in billing, identity, order or claims systems and handing over with the investigation done. Pick on the job, not the label: sample 200 recent tickets and mark which ones a correct answer would close and which need something changed.

What does each model cost?

Built-in AI usually costs seats plus a per-resolution rate; a layer costs a plan or platform fee plus a per-resolution rate, with your helpdesk seats unchanged. Zendesk plans include an allowance of automated resolutions and list pay-as-you-go resolutions at US$2.00. On Zendesk's resolution tiers, introduced on 18 May 2026, assisted escalations and contained resolutions do not count against the allowance; only verified resolutions do (resolution tiers). Intercom helpdesk Essential seats list at $29 a month ($19 for new customers), with Fin at $0.99 per outcome; Fin on another helpdesk carries a 50 outcomes a month minimum (Fin pricing). Unit prices look similar. What differs is which tickets each tool can close, so model cost on resolved tickets. See our total cost framework for the full calculation.

Where does lock-in sit in each model?

With built-in AI, your AI is tied to your helpdesk: leaving one means rebuilding the other. With a layer, your workflows and guardrails live in the layer, but your tickets, reporting and agents stay in the helpdesk. Either switch is survivable if you keep SOPs in your own documents and expose actions through your own APIs. Our guide to switching AI support vendors covers what moves and what does not.

AI layers and built-in AI compared

1. Fin (now part of Salesforce): built into Intercom, and a layer for Zendesk

Salesforce completed its acquisition of Fin, formerly Intercom, on 10 September 2026, and says Fin will continue to serve its customers as part of Salesforce's AI Labs (Salesforce). Fin is built natively into the Intercom helpdesk and also "works on any helpdesk", including Zendesk, where it connects over API, syncs your Help Center, and says most teams go live in under an hour (Fin for Zendesk). Best for Intercom teams that want one vendor, and Zendesk teams that want the fastest published setup. Limits: the self-serve plans run on the Intercom helpdesk; Fin on another helpdesk is a "Contact sales" card with a monthly minimum.

2. Lorikeet: a layer for complex and regulated support on either helpdesk

Lorikeet is an AI customer support platform for complex and regulated businesses. It connects to Zendesk, Intercom and other helpdesks through an admin-approved authorization flow, learns from your help docs and policies, and is tested on replays of your own historical tickets before it goes live (how it works). Payments and identity checks run as code the agent can invoke but never alter. It is SOC 2 Type II audited, ISO 27001:2022 certified, HIPAA compliant and signs BAAs (Trust). Linktree plugged it into Intercom alongside its human agents and cut first response time to 1 minute; UK car finance lender Carmoola resolves 60% of inbound on WhatsApp, chat and email end to end. Pricing: Lorikeet publishes its pricing: the Start plan is $2,100 a month and Scale is $5,100 a month (both billed annually), with resolved chat, email and SMS conversations at $0.99 (Start) or $0.90 (Scale) and voice at $1.50 or $1.20. Unresolved tickets and handoffs to a person are not charged. Signature is custom, with a $75K platform fee. Full details are on the pricing page. A 30-day free trial is available. Limits: two vendors to manage, and the plan fee is more than a small team with simple tickets needs.

3. Zendesk AI agents: the built-in option for Zendesk teams

Zendesk AI agents work on messaging, email and voice and can "perform actions in authorized systems autonomously", from knowledge answers to generative procedures and API integrations (About AI agents). They are included in every current Suite and Support plan, with an allowance of 5 to 15 automated resolutions per agent per month depending on plan, up to 10,000 a year. A resolution is only billed when an LLM verifies the customer's request was resolved without a human. Best for Zendesk teams whose tier 1 is mostly answerable from the help centre. Limits: older AI agent features, including AI agents Essential and bot builder answers, stop being available on 10 December 2026, so check which generation you run before you compare.

4. Decagon: an enterprise layer with a supervisor model

Decagon syncs tickets and customer data from Salesforce, Intercom and Zendesk, takes actions across Zendesk Sunshine and handles email through Zendesk and Intercom (integrations). Its security page describes a supervisor model that revises ungrounded replies before they are sent and lists SOC 2 Type II, HIPAA, ISO 27001 and PCI DSS 4.0.1 (security). The same page says Decagon enforces zero data retention with its AI providers, redacts PII with Google's DLP service, and runs Watchtower, an always-on QA system that reviews every conversation against your criteria. Beyond the helpdesk it connects to other systems over MCP and custom APIs, and to telephony through call forwarding and SIP trunking. Best for large enterprises running high volumes across chat, email and voice that want QA built in. Limits: no published pricing or trial; evaluation is sales-led, so you cannot test it on your own Zendesk or Intercom before a contract conversation.

5. Maven AGI: a layer with Marketplace installs

Maven AGI installs from the Zendesk Marketplace or over API (Zendesk) and connects to Intercom workspaces, ingesting help centre articles and conversation history (Intercom). It runs either as a frontline agent or as a copilot that shows suggested replies and articles to human agents, and it says the same configuration covers chat, email, messaging and voice. It lists ISO 27001:2022, ISO 42001, PCI DSS 4.0 Level 1 and SOC 2 Type II, plus independently assessed HIPAA/HITECH controls (trust). Best for teams that want a Marketplace install on both helpdesks and a long certification list. Limits: no published pricing or trial, and its "up to 93%" resolution figure is customer-reported on its own page.

6. Ada: a layer for Zendesk's full channel set

Ada integrates with Zendesk Guide, Talk, Support, Chat and Messaging and transfers customers to human agents on any channel (Ada integrations). It also connects to Salesforce, Kustomer, Gladly, Genesys and Amazon Connect, among others, and offers open APIs and SDKs. Best for Zendesk teams using Talk, Chat and Messaging together who want one agent across them. Limits: Intercom is not in the platforms its integrations page names, and pricing is not published.

7. Sierra: a standalone agent platform with outcome-based pricing

Sierra deploys one agent across chat, SMS, WhatsApp, email and voice and prices on outcomes (Sierra); it integrates "with your systems of record" (financial services) and lists SOC 2, HIPAA, PCI, FedRAMP High and ISO 42001 (trust). Sierra holds PCI DSS Level 1 Service Provider certification and keeps cardholder data on separate infrastructure that never touches its LLMs, which matters for payment-heavy support, and teams build agents either in code with its Agent SDK or without code in Agent Studio. Best for large consumer brands that want an outcome-priced agent across voice and digital channels. Limits: the product pages we checked do not name Zendesk or Intercom connectors, so confirm how handoff works in your helpdesk.

8. Forethought AI agents by Zendesk: Zendesk's own layer

Forethought lists Zendesk, Intercom and Salesforce among its integrations (integrations) and prices with a platform fee plus an outcome-based cost via sales (pricing). Zendesk describes it as an independent omnichannel suite for support teams, "even those that don't use the Zendesk platform" (Zendesk). Best for Zendesk customers who want a deeper agent while staying with one vendor family. Limits: it is Zendesk-owned, so it is not an independent second vendor.

What still needs a human?

Neither model removes the need for people on disputes, complaints, vulnerable customers and anything resembling clinical, legal or financial advice. The CFPB found over 98 million US users engaged with a bank chatbot in 2022 and warned about "doom loops" with no route to a human (CFPB). The FCA's Consumer Duty requires firms to put retail customers' needs first (FCA), and the EU AI Act requires people be told when they are talking to a machine, with transparency rules effective August 2026 (EU AI Act). Use NIST's voluntary AI Risk Management Framework as the review checklist (NIST AI RMF), and test for prompt injection, which OWASP ranks first among LLM risks (OWASP).

Is Lorikeet the right fit?

Lorikeet is built for complex, regulated support: fintech, healthtech, insurance and telehealth teams where the AI has to take actions in backend systems safely and leave an audit trail. It is not the right choice for everyone:

• If most of your tier 1 is answerable from the help centre and you want AI bought with your Zendesk seats, Zendesk's own AI agents fit that.

• If you want a self-serve, published-price agent you can switch on inside Intercom (or alongside your helpdesk) within days, Fin fits that.

• If you handle a few hundred tickets a month, the Start plan (from $2,100 a month) will be more than you need.

• Expect more setup than switching on an FAQ bot: workflows, guardrails and simulations are configured per use case.

Questions to ask before you pick

1. Which of my last 200 tickets need an action outside the helpdesk?

2. Can the vendor show that action on a test account, and replay every step afterwards?

3. What counts as a billable resolution, and are escalations charged?

4. Can I scope the agent to specific tickets with triggers or tags, and pause it instantly?

5. Can I trial it on my own helpdesk without a sales call?

For setup steps, see adding an AI agent to Zendesk and adding one to Intercom. To test a layer on your own tickets, start a 30-day Lorikeet trial.