Buying

Ten questions to ask every RMM/PSA vendor (including us)

It is the third vendor call of the week. The service manager has the same slide deck open that she saw on Monday, with the logos swapped. The comparison chart has a green tick in every row for the product being sold and a grey dash for everyone else. She asks whether the device list in the service desk is live or synced overnight, and there is a pause. The pause is the useful part of the call. Comparison charts are written by whoever wins them. The questions below cannot be answered with a tick, which is why they are worth the hour.

Why can a feature matrix not tell you which product to buy?

Because every product in this category has the features. Monitoring, remote access, a service desk, a credential vault, an assistant: each vendor can honestly tick the box. The differences live in the mechanisms underneath the tick. How the device list gets from the monitoring tool to the ticket. Where the AI runs and what it is allowed to do when it is wrong. Whether the audit trail is evidence or a log that application code can rewrite. A matrix flattens all of that into one colour. A question forces the vendor to describe the mechanism, and a mechanism can be checked in a trial.

Bring the list to every call on your shortlist and write the answers down in the vendor’s own words. Ours are here. Three of them are a no, and we would rather you read those before the demo than discover them three months into an implementation.

What are the ten questions?

1. When your AI takes an action, what stops it doing something destructive?

Ours: three approval modes per conversation, a floor that unattended mode cannot cross, the risk and the exact command shown first; skip, reorder, redirect or stop mid-run.

2. Which AI vendor are you sending my data to?

Ours: none. Open-weight models on infrastructure we run, or yours. No OpenAI, Anthropic or Azure OpenAI account anywhere in the product.

3. Can a technician connect to a machine without being shown its password?

Ours: yes. Credentials are injected server-side from the vault; the plaintext never reaches the browser.

4. Is the device list in your service desk live, or synced overnight?

Ours: live. We retired the sync; nothing is duplicated, so nothing drifts.

5. What inbound firewall rules does the agent need?

Ours: none on a managed endpoint. Optional, scoped listeners for a collector and an SNMP trap receiver.

6. What exactly is in your audit trail, and can it be edited?

Ours: sign-ins immutable; service desk history append-only, enforced by the database; every credential reveal recorded. Append-only, not hash-chained; we will not call it tamper-proof.

7. Is my data in a shared instance?

Ours: no. Single-tenant per customer, always.

8. Can I generate my recurring agreement invoices automatically?

Ours: not today. T&M and project billing work; automated agreement billing is on the roadmap.

9. What does it cost to move off you?

Ours: your single-tenant database, the audit trail and documentation export. You cannot take one module elsewhere.

10. Is there a bulk importer for my documentation?

Ours: not yet. If you are migrating a large documentation estate, this shapes the timeline.

What does a good answer sound like?

A good answer names a mechanism you could go and look at. A poor answer names a benefit. The difference is easy to hear once you are listening for it, and it is the same difference in every one of the ten.

A checkbox answer
An answer you can verify
“Our AI has enterprise safety controls”
Named approval modes, a floor unattended mode cannot cross, and the exact command shown before it runs
“Fully integrated with our service desk”
Live or synced; if synced, how often, and what happens when the two disagree
“Enterprise-grade encryption”
Which algorithm, which key for which scope, and an honest statement of what is not zero-knowledge
“Complete audit logging”
Which events are append-only, how that is enforced, and whether it is hash-chained or not

Which question gets the longest silence?

The fourth. Is the device list in your service desk live, or synced overnight? It sounds like a small implementation detail. It is the whole architecture in one sentence. A platform assembled from acquisitions has two databases, so a nightly job copies devices from one to the other, matches on hostname or serial, and drifts. A platform built as one system reads the same table, so there is nothing to copy. Whoever answers “synced” has told you how every other cross-module feature works too, because they all cross the same seam. We wrote about the mechanism in the device list is lying to you.

Where are our three noes?

Questions eight, nine and ten. We do not generate recurring agreement invoices automatically today; time against tickets becomes an invoice at the contract rate and pushes to QuickBooks Online, and automated agreement billing is on the roadmap. You cannot take one module and run it beside another vendor’s, because the modules genuinely run on the shared platform, and that is what makes the linked-records story possible. There is no bulk documentation importer yet, so a large documentation estate shapes your migration timeline. We say these on the first call because the pipeline is shorter when everyone knows them.

What this costs you

Every checkbox answer you accept becomes a seam a technician joins by hand. A device name copied from the monitoring tool into a ticket is minutes per ticket. A nightly sync is a reconciliation per client per month, done by the person who can least afford it. A credential shown on screen is a rotation every time someone leaves. Ask the question now, or pay for the answer in small pieces for the length of the contract.

When we are not your answer

If recurring agreement billing is your month-end, if you are migrating thousands of articles, or if you want to buy one module and keep the rest of your stack, we are probably not your answer yet. Better to know now than three months into an implementation. We wrote the long version in when Guardian Plus is not your answer yet.

And when we probably are

Your network is the hard part. You are tired of the seams. You want automation you will actually enable. If one of those is you, the linked-records argument is not marketing; it is the reason to switch. Put the ten questions to us and to everyone else on the shortlist, and compare the pauses.


The list as a PDF

Each question with why it matters, what a good answer sounds like, and ours. Bring it to every call on your shortlist.