The queue needed fixing this quarter. The rollout plan came back in four.
That is the usual shape of a ServiceNow evaluation at mid-market size. The platform can do what was asked. Getting there runs through scoping workshops, a certified administrator, and a licence agreement priced for a company several times larger.
Latch Workflow is a self-hosted ticketing system for teams that need governed, audited ticket workflows running inside their own boundary in weeks. A ticket here, or a case if you work in a regulated function, carries its own approvals and its own evidence.
install - route - approve - prove
On-prem, private cloud, or air-gapped. Identity comes from your existing OIDC provider. No ticket data leaves the network.
Shared inbox, forms, and webhooks land in one queue. Ticket routing, SLA management, and the escalation policy are configured by the ops lead, not by a certified administrator.
One plugin calls the downstream system. Above a threshold, a second person in a different role has to approve before it runs, and the denied attempts stay on the record.
Every triage call and every correction your operators make is training signal. Suggestions improve because your team corrected them, inside your deployment.
Four things ServiceNow has that Latch does not
A comparison page that pretends the incumbent is weak is not useful to anyone running an evaluation. ServiceNow is the reference implementation of enterprise ITSM, and these four areas are why.
One platform for far more than the service desk
ITSM sits next to IT operations management, HR service delivery, customer service management, security operations, and risk on a shared data model. Latch does ticketing, routing, approvals, plugin actions, and the audit trail. That is the whole list.
Configuration data and discovery that actually works
If your incident process depends on knowing which business service a failing host supports, and which change touched it last week, the CMDB and discovery tooling are the reason to buy ServiceNow. Latch has no CMDB and no discovery.
Partners, certified apps, and a hiring market
Thousands of implementation partners, a store of certified applications, and a recruitment market where you can hire someone who already knows the platform. Latch has a plugin SDK in TypeScript and Go, and a small team behind it.
Change advisory, ITIL process content, and scale evidence
Change advisory boards, ITIL-aligned process content, impact analysis backed by configuration data, and two decades of deployments at organisations with tens of thousands of employees. For a 10,000-person enterprise with a dedicated platform team, ServiceNow is a defensible choice.
If you run a large IT estate, need configuration data underneath your incident process, and have a platform team to keep the thing healthy, buy ServiceNow. This page is not written for you.
The honest objection is fit, not capability
Nothing below says the platform cannot do the work. It says the work costs three things a mid-market operations or finance team usually does not have spare: time, a platform practice, and a licence budget sized for an estate you are not running.
Discovery, process design, the data model, integrations, user acceptance testing, then go-live. That sequence is appropriate for a platform that will carry six departments. It is a long wait when what broke is one queue.
Certified administrators, two platform upgrades a year, and regression testing of every customisation. Mid-market teams rarely have a platform practice, so they rent one from a partner and keep renting it.
ServiceNow prices per fulfiller under a negotiated enterprise agreement, costed as though the platform will also carry HR, facilities, and customer service. If it carries one governed queue, the cost per resolved ticket is hard to defend at review.
None of that is a defect. It is what a platform of that breadth costs to run. The question in the evaluation is whether you are buying the breadth or paying for it.
Side by side, with the losses marked
Marks below are written the way an evaluator would write them, including the rows where Latch is the weaker product.
| Capability | Latch Workflow | ServiceNow |
|---|---|---|
| Deployment and data | ||
| Deployment model | Self-hosted, private cloud, or air-gapped | Vendor-operated cloud instance, including regulated government regions |
| Where ticket data sits | Your infrastructure, under your data residency rules | Infrastructure the vendor operates, in the region you contract for |
| Where model inference runs | Inside your boundary, on models you choose | Vendor-hosted models |
| Operator corrections improve the queue without leaving | ✓ | Learning happens on vendor infrastructure |
| Rollout | ||
| Time to a routed queue with an escalation policy | Days | Weeks, after scoping and process design |
| Time to an audited approval workflow | Weeks | Commonly a quarter or more |
| Who administers it | An ops lead or one platform engineer | Certified administrators, often a partner or an in-house practice |
| Scope | ||
| Ticketing, ticket routing, SLA management, escalation | ✓ | ✓ |
| Two-person review, with self-approval blocked | Enforced in the core | Built in the workflow engine as a configuration |
| Audit trail that records denied attempts | ✓ | Available, depending on how the workflow is built |
| CMDB and discovery | — | ✓ |
| IT operations, HR service delivery, customer service modules | — | ✓ |
| Partner ecosystem and certified app store | Plugin SDK in TypeScript and Go, small ecosystem | Thousands of partners and certified applications |
| Commercial | ||
| Licence model | Flat plan tiers, with on-prem and air-gapped deployment on the enterprise tier | Per fulfiller, under a negotiated enterprise agreement |
ServiceNow capabilities vary by module and contract. Check the rows above against your own quote before using them in an evaluation.
A help desk that learns, on hardware you control
Latch is a lightweight ServiceNow alternative in scope, not in control. Most systems that learn from your tickets require you to hand over the tickets first. Latch does not. The model improves on your queue, inside your deployment, and the improvement stays there.
On-prem, private cloud, or air-gapped
Tickets, attachments, model inference, and identity stay inside the network you already audit. That is what makes the system usable where data residency rules or an air-gapped environment take a vendor-operated instance off the table. The deployment and security model sets out the tenancy and identity details.
The queue sharpens with the decisions your team already makes
When operators keep reassigning a class of ticket, the system learns the correction and stops proposing the wrong owner. When they never escalate a certain reversal, it stops surfacing it. See how triage and ticket routing use that signal.
Approvals are how a learning system earns the right to touch real work
A suggestion is not an execution. Actions above a threshold pass through approval workflow and two-person review, and the request, the reviewer, the denied attempts, and the downstream result all land in the audit trail.
Most teams that adopt Latch keep ServiceNow
The replacement framing is usually wrong. ServiceNow keeps the service desk, the configuration data, and the change process. Latch takes the queue that cannot go to a vendor-operated instance, where the work is a case rather than an incident.
Tickets raised in either system can be linked through the API, so the enterprise service desk keeps its single view without the regulated work leaving the boundary.
Where this comparison stops working
Latch is a narrow product on purpose. If the list below describes your requirement, the shorter implementation is not worth what you give up.
Your incident and change process depends on configuration data, discovery, or service mapping. Latch has none of that and will not have it soon.
You are consolidating IT, HR, facilities, and customer service onto one platform. Buying a ticketing product for one of those queues makes the consolidation worse.
You need governed, audited ticket workflows on your own infrastructure, in weeks, run by the team that already runs the queue. Start with self-hosted ticketing or pricing.
Questions evaluators ask when the quote arrives
These come up when a mid-market team has a ServiceNow proposal on the table and is checking whether a smaller, self-hosted system covers the requirement.
Is Latch a full ITSM replacement for ServiceNow?
No. Latch covers ticketing, intake, ticket routing, SLA management, escalation policy, approvals, plugin actions, and the audit trail. It has no CMDB, no discovery, no IT operations management, and no HR service delivery module. Teams that depend on those keep ServiceNow and should not read this page as a migration argument.
How long does a Latch implementation actually take?
The software installs in a day on your own infrastructure, and a routed queue with an escalation policy usually runs within the first week. The first approval workflow with two-person review takes a few weeks. The long pole is not the product. It is the integration with the downstream system and the escalation rules your team has never written down.
Who administers Latch if we have no platform team?
An operations lead or a single platform engineer. Queues, routing rules, approval thresholds, and roles are configured in the interface. Plugins that call an ERP, a payment processor, or an internal API are written with the TypeScript or Go SDK. There is no certification programme and no upgrade project twice a year.
Can Latch run on-premise or air-gapped?
Yes. Latch runs self-hosted, in your private cloud, or fully air-gapped. Ticket data, model inference, and identity stay inside your boundary, which is what makes it workable under data residency rules that rule out a vendor-operated cloud instance.
Can we run ServiceNow and Latch at the same time?
That is the common shape. ServiceNow keeps enterprise ITSM and the configuration data. Latch takes the regulated queue that cannot leave the boundary, such as payment exceptions, sanctions hits, or KYC reviews. Tickets can be raised in either system and linked through the API.
How does Latch cost compare with a ServiceNow quote?
Latch is priced in flat plan tiers rather than per fulfiller under a negotiated agreement. On-prem, air-gapped, and open-source access sit on the enterprise tier, which is quoted rather than listed. A fair comparison still needs your actual ServiceNow numbers, because their pricing depends on which modules are in the deal. Published Latch tiers are on the pricing page.
Does Latch learn from our tickets, and does that data leave?
It learns and the data does not leave. Every triage decision, approval, rejection, and correction an operator makes is training signal for routing and classification suggestions. The model runs inside your deployment, so the improvement stays on your side of the boundary rather than in a vendor model.
Walk one governed ticket workflow end to end
Take the queue that triggered the ServiceNow evaluation, and follow it from intake through routing, approval, and the audit record. It takes about ten minutes to see whether the fit is there.