No, Lorikeet is not a helpdesk. It is an AI layer: a concierge that works inside the helpdesk you already run, such as Zendesk, Intercom, Salesforce, Front, HubSpot, Help Scout or Kustomer, resolves tickets end to end by taking actions in your systems, and hands everything else to your team in the same queue.
The default advice in 2026 is to switch on the AI that came with your helpdesk. It is already in the contract, it already sees your tickets, and procurement is one email. For plenty of teams that is the right call. For teams whose tier 1 is really "change my payment date", "where is my claim" or "update my dose", it often is not, because the hard part of those tickets happens in systems the helpdesk never touches.
So the useful question is less "which category should we buy" and more "which AI can actually resolve our tickets, inside the tools our team already uses". This page lays out both models honestly, including where the answer is not us.
Key takeaways
Lorikeet is an AI layer, not a helpdesk. It connects to your ticketing system, replies, tags, reassigns and closes tickets, and escalates to your people with context. There is nothing to migrate.
A helpdesk with built-in AI is one product. The AI is a feature of the ticketing platform, billed alongside your agent seats.
The line is blurring. Fin, formerly Intercom and now part of Salesforce, sells its agent for other helpdesks and Zendesk says its Forethought agents work on any platform, so "layer" is now a delivery model as much as a vendor category.
Pick on the job, not the label. If your tier 1 is knowledge questions, native AI is usually enough. If your volume needs actions in billing, identity, claims or clinical systems, you need an AI agent that takes actions, and that is where a layer earns its fee.
Cost is about what gets resolved, not the unit price. Per-outcome prices look similar across vendors. The difference is which tickets each one can close.
Lock-in lives in two places. Native AI ties your AI to your helpdesk; a layer ties your workflows to the layer. Keep your SOPs and APIs as your own and either switch is survivable.
1. The two models
1.1 Helpdesk with built-in AI
Here the AI is a feature of your ticketing platform. You buy seats for your human agents, then pay for the AI on top, usually per resolution or per session. The AI reads the same help center, sits in the same admin console and reports in the same dashboards. Public pricing at the time of writing (October 2026) shows how the major helpdesks package it:
Intercom sells its helpdesk per seat, from $29 per seat per month on the Essential plan paid annually, and its Fin AI Agent at $0.99 per outcome (Intercom pricing).
Zendesk describes its pricing as primarily seat-based, with AI agents included in every Suite and Support plan and billed per automated resolution (Zendesk pricing).
Freshdesk includes the first 500 Freddy AI Agent sessions and charges $49 per 100 additional sessions (Freshdesk pricing).
Salesforce sells Agentforce with consumption pricing, through Flex Credits or per conversation, or with per-user licensing (Agentforce pricing).
The appeal is real: one vendor, one contract, one console, and an AI that already knows your macros and help center. If you are choosing a helpdesk anyway, getting the AI in the same purchase is efficient.
1.2 AI layer on top of your helpdesk
An AI layer is a separate platform that plugs into your helpdesk and works the queue like a very fast teammate. It reads incoming tickets, replies, tags, reassigns and closes them, and when a human needs to step in, it hands over with the investigation already done. Your agents keep working in the inbox they know. Your reporting stays where it is.
The layer's job is resolution, not deflection. It connects to the systems where the work actually happens (billing, loan servicing, identity verification, order management, pharmacy, claims) and does the thing the customer asked for. That is the difference between an AI agent that takes actions and one that tells the customer how to do it themselves.
Lorikeet is built this way. How Lorikeet works is four verbs: connect, train, test, improve. Setup is connecting, not migrating.
1.3 The line is blurring
The categories are less clean than they were two years ago. Fin now markets that its agent works on any helpdesk, including Salesforce and HubSpot, at the same $0.99 per outcome (Fin pricing). Zendesk's pricing page says its Forethought AI agents work on any platform as well as inside Zendesk. So several helpdesk vendors now sell their AI as a layer too.
That is useful for buyers. It means the choice is not "stay on our helpdesk and accept its AI" versus "migrate". It is a straight comparison between AI agents on the tickets you actually get, with your helpdesk held constant.
2. AI layer vs helpdesk with built-in AI, side by side
Dimension | Helpdesk with built-in AI | AI layer (for example Lorikeet) |
|---|---|---|
What you buy | Ticketing platform plus an AI feature | An AI concierge that works inside your existing ticketing platform |
Where your team works | The helpdesk | The same helpdesk, unchanged |
Migration needed | Yes, if you are not already on that helpdesk | No, it connects to the one you run |
Typical strength | Knowledge answers from your help center, native reporting | Multi-step workflows that take actions in backend systems |
Actions outside the helpdesk | Varies by vendor and plan; often configured per action | Core design: APIs, webhooks and MCP servers inside workflows |
Voice | Depends on the helpdesk's own voice product | Works with the telephony you already use, through call forwarding |
Pricing shape | Seats plus AI per outcome, resolution or session | Plan fee plus a rate per resolved ticket; helpdesk seats unchanged |
Vendors to manage | One | Two |
Where lock-in sits | AI tied to the helpdesk | Workflows and guardrails configured in the layer |
Best fit | Mostly FAQ and how-to volume, one helpdesk, simple actions | Regulated or action-heavy volume, several channels or systems, no appetite to migrate |
3. When a helpdesk with built-in AI is the right call
3.1 Your tier 1 is mostly knowledge questions
Password resets, shipping times, plan differences, "how do I export this". If a well-kept help center answers most of your volume, native AI trained on that help center will do a good job. You do not need a second platform to paste an article into a reply.
3.2 You are choosing a helpdesk anyway
New team, first real tooling, or a migration already underway for other reasons. Buying the helpdesk and its AI together is one evaluation, one security review and one bill. Revisit the question once you know what your queue actually looks like.
3.3 Your volume is small
Be honest about the floor. Lorikeet's Start plan is $2,100 a month, paid annually, and is built for teams under 5,000 monthly tickets (pricing). If you handle a few hundred tickets a month and most of them are simple, the AI inside your helpdesk will almost certainly cost less.
3.4 The actions you need already live in the helpdesk
If "resolution" for most tickets means changing a field, applying a macro or updating a ticket status, the helpdesk's own automation may cover it without a separate AI layer.
4. When an AI layer is the right call
4.1 Your hard tickets need actions in other systems
This is the big one. A payment date change touches your loan servicing platform. A claim status lives in the claims system. A dose change needs the clinical record and a policy check. If the helpdesk AI can only answer from knowledge, those tickets land back on your team, and they are usually the expensive ones. An AI layer is designed for exactly this: it calls your systems inside workflows, runs sensitive steps such as payments and identity checks as deterministic code the agent can invoke but never alter, and escalates with context when human judgment is needed. We covered the difference in more depth in AI agents that take backend actions vs FAQ bots.
4.2 Your AI has plateaued at FAQ answers
Many teams hit the same wall: the AI handles the easy questions well, the resolution rate stops climbing, and every multi-part or account-specific ticket still needs a person. Carmoola, a UK car finance lender, had automation built on a knowledge base and rigid, form-driven prompts that resolved about 30% of inbound questions. With Lorikeet, its concierge now resolves 60% of all inbound on WhatsApp, chat and email end to end (Carmoola's story).
4.3 You run more than one system
Chat in one tool, email in another, voice in a separate contact center platform. An AI layer can serve all of them with one set of workflows, so the customer gets the same answer whether they call or write in. Lorikeet trains workflows once and deploys them across chat, email, voice, SMS and WhatsApp, with voice available in the US, UK and Australia.
4.4 You are regulated
Fintech, healthtech and insurance teams need to know why the AI did what it did, test it before customers see it, and stop it saying things it should not. Lorikeet builds that in: simulations replay historical tickets before launch, guardrails check messages in both directions, and Coach scores every conversation afterward. Lorikeet holds SOC 2 Type II, ISO 27001:2022, HIPAA and GDPR, and signs BAAs for healthcare customers.
4.5 You do not want to migrate to get better AI
Helpdesk migrations are expensive in ways that do not show up in the license price: ticket history, macros, routing rules, reporting, integrations and agent retraining. If your helpdesk works for your people, changing it just to change the AI is the tail wagging the dog. A layer lets you upgrade the AI and leave the helpdesk alone.
5. Cost: what each model actually charges for
Both models increasingly price per outcome, so the headline unit prices look close. The real differences are in what else you pay for, and what counts as an outcome.
Cost line | Helpdesk with built-in AI | Lorikeet (public pricing, October 2026) |
|---|---|---|
Helpdesk seats | Per human agent, for example from $29 per seat per month on Intercom Essential | You keep paying for the helpdesk you already have; Lorikeet charges for resolved conversations rather than seats |
AI unit price | Intercom Fin $0.99 per outcome; Zendesk per automated resolution; Freshdesk per session | Chat, email and SMS resolutions $0.99 on Start and $0.90 on Scale; voice $1.50 and $1.20 (3 minute average) |
Platform fee | Included in the helpdesk plan | Start $2,100 a month and Scale $5,100 a month, paid annually, including the platform and a forward-deployed engineer support package |
Unresolved tickets | Check each vendor's definition | Unresolved or unsatisfactory tickets cost nothing, and escalations to a person are not charged |
5.1 The honest math
If both options resolve the same tickets, the AI inside your helpdesk will usually be cheaper, because you are already paying for the helpdesk and there is no second platform fee. An AI layer only earns its fee by resolving tickets the native AI cannot, and those are typically the ones your most expensive people handle today.
So compare total cost, not unit price. For each option, add the AI spend to the human cost of the tickets it leaves behind. A layer that resolves the account-change and billing tickets can cost more per month in software and still cost less in total. A layer that resolves the same FAQs your native AI already handles is money wasted. Our total cost framework for CFOs walks through the full model.
5.2 Check what "resolved" means
Per-outcome pricing is only as good as the definition of an outcome. A conversation that ended is not the same as a customer whose problem is fixed. Ask every vendor, us included, what triggers a charge, whether you can dispute one, and how they prove the ticket was actually resolved. More on this in outcome-based pricing for regulated support.
6. Lock-in: where it actually lives
6.1 Helpdesk-native AI ties the AI to the helpdesk
If the AI is a feature of your helpdesk, changing AI can mean changing helpdesk, unless that vendor also sells its AI standalone. Your prompts, procedures and reporting are all inside one system, which is convenient right up until you want to leave.
6.2 An AI layer ties your workflows to the layer
With a layer, your ticket history stays in your helpdesk and your systems of record stay yours. What lives in the layer is the configuration: workflows, guardrails and tone. That is real lock-in too, and any vendor who says otherwise is grading their own homework.
6.3 How to keep either switch survivable
Keep your SOPs as your own documents, not only as configuration inside a vendor. Expose actions through your own APIs, webhooks or MCP servers, so any AI can call them. Keep the helpdesk as the system of record for conversations. Do those three things and switching becomes reconfiguration, not a rebuild.
7. Do you still need a helpdesk if you use Lorikeet?
Usually, yes. If you have human agents working escalations, the helpdesk is where they work, where your SLAs and routing live, and where your reporting comes from. Lorikeet works inside it and in front of it, not instead of it.
There are three exceptions worth knowing:
No helpdesk yet. Lorikeet Chat deploys AI agents directly on your website without needing a ticketing system.
AI inside your own product. The Lorikeet SDK lets your app send customer messages to Lorikeet and get responses and outcomes back, while your team owns the interface.
A helpdesk without a built-in connector. Flexible Ticketing connects any ticketing system that sends webhooks and accepts replies and updates through an API. It is set up with our team, not self-serve.
8. How Lorikeet works inside your helpdesk
Lorikeet works inside the ticketing system you already run and takes action through your own APIs, so there is nothing to migrate. The integrations page lists the current connectors:
Ticketing systems: Intercom, Salesforce, Zendesk, Zendesk Messaging, Front, HubSpot, Slack, Help Scout, Kustomer and Salesforce Messaging, plus Flexible Ticketing for any system that sends webhooks, and Lorikeet Chat and the Lorikeet SDK where there is no helpdesk in the path.
Telephony: any platform that supports call forwarding, including Amazon Connect, Twilio, Genesys Cloud, Dialpad, Vonage, Five9, AirCall and Talkdesk, plus the Lorikeet API for starting a voice conversation with customer context.
Knowledge bases: Confluence, Guru, Google Docs and Notion.
Tools and actions: your own APIs, webhooks and MCP servers, plus Stripe, Shopify and SendGrid.
8.1 Connect and train
Connect your ticketing system and the systems Lorikeet should act in, most through an admin-approved authorization flow, then train the concierge on your help docs, SOPs and brand guidelines. Every plan includes forward-deployed support.
8.2 Test before customers see it
Replay historical tickets and synthetic scenarios in bulk, review projected resolution quality and knowledge gaps, and deploy topic by topic. The topics the concierge is trained for go live; the rest stay with your team.
8.3 Act and improve
The concierge takes actions through your APIs and MCP servers inside natural-language and deterministic workflows, and Coach reviews the results so you can keep raising resolution quality.
Two public examples of what "inside your helpdesk" looks like in practice. Linktree plugged Lorikeet into Intercom so it could work alongside its human agents, and cut its first AI response time to one minute (Linktree's story). easykind kept its Salesforce Service Cloud stack, with Lorikeet's engineering team building the integration, and email response times dropped 92%, from 24 hours to roughly two (easykind's story).
If you run Zendesk or Intercom specifically, these walkthroughs go step by step: how to add an AI agent to Zendesk and how to add an AI agent to Intercom.
9. A five-question decision checklist
What share of your tickets need an action outside the helpdesk? If it is small, start with native AI. If it is large, test a layer.
Is your current AI resolving, or deflecting? Pull a sample of "resolved" conversations and check whether the customer's problem was actually fixed.
Would getting better AI force a helpdesk migration? If yes, price the migration honestly before you commit.
Can you test the AI on your own historical tickets before launch? Demo scripts prove nothing about your queue.
What do you keep if you leave? Ticket history, SOPs, APIs and the definition of a resolution should all stay yours.
If your helpdesk AI handles the FAQs and your team still does every ticket that needs an action, we should talk. Get a demo and bring a sample of the tickets your current AI hands back, or start a 30-day free trial and try it on your own queue.
