From First Call to AI Running in Your Building
Five phases, one on-site visit, and a team that stays after the install. Here is exactly what happens, and who is responsible for each part of it.
Typical time from the first call to staff using it every day
On site, with a NextWrite engineer in your building for the install
Per-question fees, no matter how much your staff use it, now or later
- 01Week 1
We find out whether this is right for you
A working session with whoever knows your data and whoever signs off on spending. We map what you hold, the rules attached to it, and the work your staff actually want help with.
NextWrite handles- Inventory the record types you handle and the obligations that come with each
- Size the hardware against your headcount and document volume
- Price the whole deployment as one fixed number, not a range
Your team handles- Bring the person who knows the files and the person who approves the spend
- Tell us the constraints you cannot move on
You end up with: A written recommendation and a fixed quote, including "not yet" if that is the honest answer.
- 02Weeks 2 to 3
Your server gets built, tested, and shipped
We spec and source a machine sized to your team, favoring refurbished hardware wherever it meets the requirement. It belongs to your organization from the day it arrives.
NextWrite handles- Source, assemble, and test the server hard before it ever reaches you
- Install the open-source model and test it against sample documents you approve
- Produce the purchase paperwork your finance team and auditors will ask for
Your team handles- Confirm a lockable space, power, and a network drop
- Add the machine to your asset register
You end up with: A tested AI server owned outright by your organization. No lease, no vendor lock, no contract that takes it away.
- 03One day, on site
We come to you and stand the whole thing up
An engineer is physically in your building for the install. The server goes in, the private connection comes up, and accounts are set up to match the logins your staff already use.
NextWrite handles- Install and connect the server, then bring up the private, encrypted connection
- Configure accounts, roles, and permissions to match how your organization is actually structured
- Run your real documents through it while your team watches
Your team handles- Give us access to the space and a network contact for the day
- Choose the first documents you want it tested against
You end up with: A working system your staff can sign into before we leave the building.
- 04Weeks 4 to 5
Your staff learn it on their own work
Training built around the jobs your team actually does, not a generic product tour. Case workers, program staff, and development staff each get a session aimed at their own workload.
NextWrite handles- Run role-specific training using your documents and your workflows
- Draft the internal guidance and the AI use policy your board can adopt
- Sit with the first real tasks until people are confident working unsupervised
Your team handles- Free up ninety minutes per team
- Name one internal champion per department
You end up with: Staff who use it without being chased, and a written AI policy you can hand to a funder or a board.
- 05Ongoing
It keeps getting better without new spend
Open-source models improve constantly. When a stronger one fits your hardware, we install it. The same machine you already bought becomes more capable without another purchase order.
NextWrite handles- Push model upgrades as better open-source releases ship
- Monitor system health remotely, with no access to your prompts or your answers
- Review capacity with you as your team grows
Your team handles- Call a named person when something is wrong
- Tell us when headcount changes materially
You end up with: No specialist hire, no per-query bill, and hardware that gains capability instead of losing it.
At no point in any of this does a client record leave your building.
Not during the assessment, not on install day, and not in any support session afterward. Our engineers work on the system. They do not work on your data, and they cannot read what your staff ask it.