Built on Kultiv.ai
Fill the gap with software, not with a coordinator.
Every operational gap can be closed two ways: build something, or make a person responsible for chasing it. The second one is cheaper this month and ruinous by year three. We build.
Why this is different now
The gaps were never worth a project.
Every MSP has twenty small breaks between systems. Individually none of them justified a six-figure integration, so each one got a human workaround instead: a spreadsheet, a recurring task, someone who just remembers to check.
Twenty workarounds is your margin. It is also why nobody can go on holiday.
Kultiv.ai changes the arithmetic. A build that used to take a quarter and a business case now takes days, which means it is finally rational to fix the small things — and to delete the fix if it turns out not to earn.
No new bureaucracy
Six constraints on everything we ship.
These are what keep AI-built software from becoming the next thing your team has to maintain.
It lives where they already are
Inside the PSA, the RMM, the ticket, the chat channel. If it needs another tab and another login, adoption is a management problem we have created.
No role exists to feed it
If keeping the build useful requires someone to maintain a list, we have not closed the gap. We have moved it and given it a job title.
One metric, one owner
Each build names the number it is supposed to move and the person who answers for it. No metric, no build. That rule kills about a third of requests.
It shows its working
Where AI makes a judgement — a triage call, a priority, a match — it says why, and a human can override it. Silent automation is how trust in data dies.
Small enough to delete
We build the narrowest thing that closes the gap. If it stops earning in six weeks, we rebuild it or bin it. Nothing accumulates by default.
You own it
Source, prompts, configuration and documentation are handed over. Nothing we build is hostage to our retainer continuing.
What we build
The gaps we close most often.
Yours will not be exactly this list. It is close, though — MSPs break in remarkably similar places.
Contract reconciliation
Compares RMM agents against billed PSA contracts nightly and posts the delta to a channel with an owner tagged. Finds unbilled endpoints and phantom billing in the same pass.
Moves: agent-to-contract variance
Inbound triage
Reads the ticket as it arrives, sets category, priority and queue, attaches the relevant documentation, and flags the ones that should have been handled by automation.
Moves: time to first touch, first-time fix rate
QBR assembly
Builds the client pack from your own data on a schedule — ticket trends, project status, risk items, spend. Account managers edit a narrative instead of assembling numbers for a day.
Moves: QBR cadence held, contract margin visibility
Drift detection
Watches endpoints against the known-good configuration and opens a ticket the day one drifts, rather than the day a user notices something broke.
Moves: endpoint standardisation, reactive hours
Onboarding generation
Turns a signed order into the actual task list, asset records and access requests, in the right order, without anyone transcribing the contract.
Moves: onboarding time to first solo close
Margin watch
Recomputes contract margin per client every month from time entries and revenue, and surfaces the ones sliding before renewal, not after.
Moves: contract margin by client, service gross margin
Every build above is sequenced against a dimension of your OML score, so it is obvious what it is for and how we will know it worked.
How a build runs
Gap to live in about two weeks.
Name the gap
Which dimension, which metric, whose name. If any of the three is missing we stop here.
Build thin
Days, not quarters. First version does one thing and does it in the tool they already use.
Run it live
In production with the owner watching, alongside the old way for the first week only.
Keep or kill
Six weeks later, the metric decides. Rebuild, hand over, or delete it and say so.
Guardrails
Where AI does not get a vote.
You sell trust to your clients. We are not going to spend it for you.
- Client data
- Scope, retention and processing are agreed in writing before the first build. Nothing leaves the boundaries your own client contracts allow, and we will tell you where a build would breach them.
- Irreversible actions
- Nothing we build deletes, disables, bills or emails a client on its own judgement. Those steps stay behind a human, permanently, not just during pilot.
- Judgement calls
- Where the model classifies or matches, its reasoning is visible and its output is overridable. Overrides are logged, because a pattern of overrides is telling you the build is wrong.
- Security posture
- Least privilege by default, credentials in your vault not ours, and access that expires with the engagement. Your auditors get a plain answer about what we touched.
- Handover
- Source, prompts, configuration and a plain-English runbook, transferred to you. If we vanished tomorrow, your team could maintain or retire every build we shipped.
Start small
We do not build before we measure.
The diagnostic comes first, always. It tells us which gaps are worth closing and gives you the baseline to hold us to.