Continuous reconciliation
Every claim a tool makes is compared against independently gathered truth, continuously, across every client — and the difference is reported in dollars where it has a dollar value.
Capability & limits · for the technical lead
One system, one data model, and an AI that works across all of it.
Monitoring, security, patching, service desk, agreements, documentation and compliance in one system with one data model — and an AI that works across all of it rather than sitting in a corner of one screen.
The four things that follow from that
Provenance on every billable minute
Time is drafted from work actually performed, with the evidence attached. A client clicks an invoice line and sees exactly what was done.
Autonomous client onboarding
Discovers the network, builds the asset model, maps topology, profiles security posture, and drafts the agreement from what it actually found.
Self-maintaining documentation
Written from discovered state, and flagged the moment reality diverges. The stale-documentation problem, structurally solved.
One agent, forty-two things — including the ones you buy separately
One installer, one signed channel, no second daemon on the endpoint. Every action is a declared verb on the autonomy ladder with its risk tier — there is no back door and no unlisted capability.
Remote control is included, not an add-on. Sessions, remote shell, file transfer and chat with the person at the desk. Most MSPs pay a separate vendor per technician per month for exactly this.
And what it deliberately cannot do: no keystroke logging, no browsing history, no per-user application profiling — the health data structure has nowhere to put a user, a window title or a path. Every verb, and the privacy boundary in full.
An analyst on every workstation you manage
A user notices something, tolerates it for two days, calls, describes it badly, and a technician spends twenty minutes reproducing it. The condition was visible in telemetry the whole time.
Your RMM already alerts on thresholds. That is not this. A threshold says disk 92% full. This says:
“This disk will be full in nine days on its current trend. 7.2 GB of the last 30 came from one application's log directory, which has no rotation configured. Here is the growth series.”
And it is judged against its own history, not a number somebody set once. “Boot time went from 40 seconds to three minutes over eleven days” is a finding no threshold catches, because nobody sets a threshold on a machine's own trend.
A failing disk from SMART trend and rising latency. A leaking application. A battery at end of life. Wi-Fi problems from roaming patterns rather than the internet is slow. Each becomes a ticket that arrives with the diagnosis already attached — ranked causes where the evidence is ambiguous, every claim citing the telemetry behind it.
And it proposes turning itself down.
A finding class that keeps raising tickets nobody actions becomes a decision waiting for you, and Genie proposes stopping it. Every MSP has alert fatigue. No other tool in this market offers to be quieter.
It watches system health and nothing else — no keystrokes, no screen content, no browsing history, no per-user application profiling. That is enforced in the database rather than promised in a policy: a signal about a person cannot be stored. And the resource ceiling is a hard limit, because an agent that makes a workstation slow is one your client will make you remove.
Triage happens before the ticket exists
Every PSA in this market creates the ticket first and triages afterwards. That is why service desks fill with autoreplies, vendor newsletters, delivery receipts, calendar invitations, out-of-office replies and the eleven bounces from one bad send — all of it now a ticket somebody has to read and close.
This decides before a ticket is created, so the queue only contains work.
And nothing is silently discarded. What Genie holds sits in a view with his reason and one click to convert. When he is unsure, he creates the ticket — a wrong ticket costs thirty seconds, a wrong dismissal costs a client.
By the time a technician opens it the client is identified, the asset resolved, the agreement known and the duplicate linked. A sender nobody recognises is unresolved, never guessed — matching on a domain is how one client’s request lands in another’s queue.
The four that save the most time
Everything below this is covered by area. These four are the ones an MSP notices in the first fortnight.
Describe a job, get everything to deliver it
Plain English in. The quotation your client signs, the plan you schedule from, and the runbook your technician works from at 2am — all three from one source, so no figure in one can contradict another. Half a day, and the runbook usually never got written at all.
Onboard a client while you do something else
Discovery, asset model, topology, posture, coverage and a drafted agreement. Then he stops and shows you what he observed, what he could not reach and what he is unsure of — three lists that are never merged. Two to five days becomes an afternoon.
Ask anything in plain English
"Every backup never restore-tested." "What changed on this server, and who changed it." Every figure clicks through to its rows, and where he generated a query, the query is shown. The report nobody had time to build.
Settle "network or server" from the trace
A capture on the technician's own machine, on the client's VLAN, analysed in seconds. The most escalated question in IT, answered with evidence instead of argued about for a week.
By area
Everything an MSP does in a day has an owner here. Where something is deliberately absent, it says so and why.
Monitoring & RMM
Agent and agentless monitoring, deep SNMP, topology, baselines, flap detection, dependency-aware suppression. Remote access with a full interactive toolset — terminal in both SYSTEM and user context, registry, services, event log, file transfer, drive health, Wake-on-LAN across subnets. Plus three nobody ships: an on-demand speed test, ad-hoc ping and traceroute from the agent, and a desktop message so Genie can warn a user before he touches their machine.
Patching
Ring-based deployment with canary, early wave and broad. Third-party packages. CVE correlation. Supersedence as a per-product chain, so a machine carrying the newer patch isn't reported as missing the older one — the false positive that makes an MSP stop trusting a patch report. Client maintenance windows with timezones and blackouts, enforced rather than advised: a job outside its window does not start.
Service desk
Tickets from email, portal, phone and Genie himself — one object, one SLA. Triaged before the ticket exists, so autoreplies and newsletters never become work. Nothing is silently discarded: what he holds sits in a view with his reason and one click to convert. And when he's unsure, he creates the ticket — a wrong ticket costs thirty seconds, a wrong dismissal costs a client.
Microsoft 365
Multi-tenant management, provisioning, joiners and leavers, licence reconciliation, security posture. Built into the same data model rather than bolted on — which is why Genie can say "this policy claims MFA is enforced, I verified it 40 minutes ago, and Entra was unreachable for eleven minutes at 09:20 so I am not reporting that window as compliant."
Backup & continuity
Normalised across whatever backup products you actually run, in three states, never one green tick — verified means a restore was tested, reported means the vendor said so and nobody checked, silent means no report arrived and that is not success. Network configs are backed up, diffed and restorable, with the restore proven by re-reading the device.
Security & SIEM
Log ingestion with deep parser coverage, detections, correlation and retention policy with compliance holds. Log forensics that gives you ranked candidates each citing the lines it read — and "I can't tell from this, collect X" as a real answer. Email security operations, quarantine, phishing triage. Caller identity verification at the helpdesk, which no PSA ships.
Commercial
Agreements, provenanced time capture, billing, contract renewals, warranty and asset lifecycle, procurement with live distributor pricing carrying its source and its age. Distributor-to-contract-to-invoice reconciliation — six divergence classes, and over-billing a client ranks above under-billing, because one is a money problem and the other is a relationship problem.
Projects & proposals
Describe the job in plain English. Genie reads the estate, builds the plan from your own history, prices it from your rate card, and produces three documents from one source — the quotation your client signs, the plan you schedule from, and the implementation runbook your technician works from at 2am, with the ticket on every step and the point of no return marked.
The client relationship
Who you talk to, what authority they have, and what you promised them — tracked as objects with owners and dates, because "you said you'd sort that" is the most damaging sentence in an MSP relationship and nothing anywhere tracks it. Client portal, QBRs, satisfaction. Not a CRM, and we say why.
Documentation & compliance
Documentation and credential vault, with continuous drift detection against observed state — so a documented fact that stopped being true becomes an exception the day it drifts, not at audit time. Compliance frameworks with evidence collected rather than asserted. And a control with no automatable evidence is marked manual, never auto-passed.
Agreements, time & billing
Contracts with their inclusions, entitlements and rates. Time tracked with provenance on every minute — Genie's entries are marked automated, never disguised as a technician's, and parallel work is never billed as wall-clock. Invoices reconcile against the agreement, and margin drift on a contract is a finding before it is a renewal conversation.
Projects & delivery
A living plan from the moment a piece of work is accepted — timeline, dependencies, critical path, and the slippage drawn rather than described.
Progress is computed from observable checks, never typed by a technician. Every PSA on the market has somebody enter a percentage; here a task is done when the thing it produces exists. Verified, claimed and unverifiable are three figures and never summed.
Can you take it on
Capacity is measured, not contracted hours minus bookings — from what your practice actually delivered, with reactive load predicted from your own ticket history. Three confidence tiers, and the weakest input governs the answer. No history means no figure, said plainly rather than estimated. And it is never a performance tool: nothing here ranks a technician.
Procurement & hardware lifecycle
Quotes from live distributor pricing, orders tracked to delivery, and the device that arrives becomes the asset that was quoted — no re-keying. Warranty and end-of-life read from the vendor rather than from a spreadsheet somebody stopped updating, so a machine out of support is a finding rather than a surprise during an outage.
Scheduling & on-call
Dispatch that knows travel, skills and what is already booked. An on-call rota with escalation that confirms somebody actually received the page rather than recording that one was sent — an unacknowledged escalation is the failure this exists to prevent, and it escalates again.
New business
Prospects, opportunities and stages — and stages that advance because something happened, not because somebody dragged a card.
Where you have run an evaluation on a prospect's estate, the qualification is arithmetic rather than judgement: what they run, what is wrong, and what it would cost at your rate card — every figure traced to a finding. And a won deal becomes the project without re-keying, because the assessment, the quote and the plan are one set of records. No lead scores — at the volume an MSP closes, a confidence score is a guess with a decimal point.
An AI that earns its permissions
Every capability starts supervised. It is promoted on a measured accuracy record and demoted automatically the moment that record slips. Permission is scoped to what an action would touch — its change class and blast radius — never to which tool it happens to call.
Hard limits, written in code — not policy
Genie never approves billable time · never commits to your client on scope, price or dates · never attests compliance · never wipes a device · never acts unsupervised on a domain controller or on identity · never declares a security incident.
AI does all the work. A human makes every commitment.
What we don't do
We don't sell email security or phishing protection. Those are threat-intelligence businesses and we'd do them badly.
We don't sell security awareness training content.
We don't match the twenty-year device catalogues of the incumbent monitoring vendors. We publish exactly which devices and log sources we support, and keep that list honest.
We're not a hosted service. Dream-Genie runs on your hardware — a deliberate choice, and not everyone wants it.
Every module, on your hardware
One price, no tiers, no AI metering. Try it against your real clients first.
Genie is the AI at the centre of it — with tools, boundaries and an evidence standard, not a chat box over a knowledge base. What he actually does.
It reads from Microsoft 365, Entra ID, Active Directory, SNMP devices, Hyper-V and VMware, and writes to your accounting software. What it talks to, and what it refuses to.