01
Answers from your own documents
Questions from customers or from your own team, answered out of approved documents, as a draft a person reads and sends.
AI Integration
We put AI to work inside the software you already run.
An integration does one job from information you already have: a reply prepared from your own documents, an enquiry filed as a record, a long call turned into a short note. We agree what it may read, who may use it and what it does when it doesn’t know, then test it on your own cases before anyone relies on it.
Try it. Pick a question a client might send us. The desk reads only the documents this role is allowed to open, drafts a reply and shows where the answer came from. Untick a document to see what it does without it. Nothing is sent.
Draft, ready for review
3. Draft reply
Your Basic plan covers security updates, uptime monitoring and a monthly health check, with two hours of support a month. The monitoring is what tells us the site is down, and putting it back up comes out of those hours.
Basic care includes security updates, uptime monitoring and a monthly health check, with two hours of support a month. Standard care adds content updates, bug fixes and performance monitoring, with five hours a month.
- Where a person steps in
- Support reads the draft and sends it.
Read 3 of 3 permitted documents, matched section 2.1 of Care plans, what each one covers.
What we connect
A first one is usually one of these four jobs.
Each reads something you already have and prepares one thing. None of them sends anything on its own unless you decide it should.
02
Enquiries filed as records
What arrives by email or through a form is read, sorted and written into your system with the fields filled, so nobody retypes it.
03
Long things made short
A call, a thread or a document reduced to what was decided, what is still open and what happens next, with the original one click away.
04
Search that respects who is asking
One search across your files where each person sees only what their role may read, and the passage that answers is shown with the document it came from.
When one job runs end to end and a person only handles the exceptions, that is an agent rather than an integration. See AI agents
Limits
Most of the risk is in what it reads.
The system needs an agreed answer for each of these, and we settle them with you before the build rather than after it.
-
01
The documents don’t answer the question.
It says so and passes the question to a person, with a note of what it checked.
-
02
The answer sits outside this person’s access.
Documents a role may not read are never searched, so nothing from them can reach a reply, not even a hint that an answer exists.
-
03
A document is out of date.
Each source has an owner and a date. A section its owner has marked for review isn’t quoted until it’s updated.
-
04
The work isn’t written down anywhere.
If the rule only lives in one person’s head, there is nothing to connect to yet. Writing it down is the first job, and now and then it turns out a form or a rule in the software you already have does what you were after.
Responsibility
Each step of a reply has someone responsible for it.
-
Your document owner
The source
Keeps the approved information current and marks a section for review when it changes.
-
Version2
The checks
Tests the agreed cases before the connection is used, and again after every change.
-
The assistant
The draft
Prepares the reply from permitted documents, or passes the question on.
-
Your reviewer
The approval
Reads what needs approval before it reaches a customer.
Questions
What people ask before they connect anything.
Do you need access to everything we have?
No. One source to begin with, read-only wherever that’s possible, and only what the job needs. Access grows when the job grows, and you agree to it each time.
Our documents are a mess. Is that a problem?
It’s the usual starting point, and tidying them is worth doing whether or not you connect anything. We look at what exists in the first week and say plainly what has to be sorted before a connection is worth building.
Will it send anything to a customer on its own?
Not unless you decide it should. The default is a draft that a person reads and sends. Case by case, you tell us which ones can go out on their own and which wait for approval.
Can it work with the software we already use?
Usually. Anything with a documented way in, and quite a lot without one. Where a system has no way in at all, we say so in the first week instead of after the build.
Where does our information go?
The model runs as a service, so a question and the passages it needs are sent to it. We agree with you which provider that is and what it may keep before anything is connected. If the information can’t leave your systems at all, that’s a private AI build rather than an integration. Private AI
What does it cost to run?
Model usage, billed at what it costs, and care if you want it. We work the usage out from your own volumes and put the figure in the written recommendation, before you commit to anything.
Scope and price
We start by looking at one workflow.
We examine one workflow and the data it depends on, then write down the proposed connection, its limits and the next step. That week costs €490 and comes off the build if you go ahead.
The same written recommendation can conclude that the connection isn’t ready to build.
Workflow IntegrationRecommended first step
One data source and one assistant or automation, with permissions set by role. We test it on your real cases before handover.
from €1,900per project, 2 to 3 weeks
Multi-System Integration
Up to three connected systems, roles and an audit trail, monitoring, and a monthly review for the first three months.
from €4,900per project, 4 to 6 weeks
The build is a one-off. Model usage is billed at cost, and care after handover is €190 a month when you want it.