Applied AI / North East England
AI that runs where you decide, measured before you rely on it.
We build AI into the work businesses already do: reading documents, drafting, finding things in your own files. We run open models on our own GPU hardware, so sensitive material does not have to go to a third-party service, and every feature ships with a pass rate on your real examples.
Start a conversationStart here
AI Fit Assessment
Two to three days on one process in your business, to find out whether AI is the right tool for it before you spend anything bigger.
- How it works today, and where the time actually goes.
- What data it touches, and which systems are involved.
- What would break if an AI got it wrong.
You get a written recommendation with one of four conclusions: build it with AI, buy an off-the-shelf tool, fix it a simpler way (a macro, a form, a process change), or leave it alone. It includes a cost-benefit check, and a fixed price for a pilot if we recommend one.
Book one by email:
info@cienvita.com
if you go ahead
01 / What we do
Four kinds of AI work.
Documents and admin
Invoices, forms, contracts, inbox triage, reports someone rekeys every Monday. The model reads and drafts, a person approves, your systems get the result.
Search over your own files
Ask questions of your policies, manuals and project history and get answers that point to the source. For your team first; nothing talks to your customers unsupervised.
Private models
Open models on hardware you own, on ours, or on GPUs rented for the length of a job. We size the hardware against your workload and tell you whether a hosted API would be cheaper.
Using AI providers without losing track
If you already pay for Anthropic, OpenAI or Azure, we put the keys in one place and record usage per team and per model, so the monthly bill stops being a surprise.
02 / Lab
What we have run on our own hardware.
Our GPU box is an NVIDIA DGX Spark: one desk-sized machine with 128 GB of memory shared between CPU and GPU. These are things we ran on it before anyone paid us to.
gpt-oss-120b on the AIME 2025 maths problems, run locally and averaged over repeated runs. A 117-billion-parameter model, no data leaving the room.
The same 120B model generating text for a single user, and about 80 tok/s for the smaller gpt-oss-20b. Fast enough for interactive use by a small team.
Of audio transcribed on our own GPU rather than sent to someone else's service.
An open model turning a single photograph into a 3D mesh, including the part that did not work. See the models.
03 / Principles
How we keep it honest.
Measured before you trust it
Every AI feature ships with real examples and a pass rate we can show you. If we cannot say how often it is right, it does not go live.
A person stays in the loop
We do not wire a model into anything that spends money or speaks to your customers unsupervised.
We will tell you not to
Sometimes the answer is a spreadsheet, an off-the-shelf tool, or waiting. We would rather say so than sell you a project.
04 / How we work
A small first step, at a fixed price.
- A discovery call, free. Thirty minutes on the problem you are trying to solve. Within a week we send a short note saying whether we think AI is the right tool, which service fits, and a rough idea of cost. If we do not think we can help, we say so.
- An AI Fit Assessment, at a fixed fee. One process reviewed in detail, ending in a written recommendation and, if we recommend one, a fixed price for a pilot.
- A pilot on your real data. Scope agreed in writing before we start, at a fixed price that does not move unless the scope does, so you see it working in your business before committing to anything larger.
- Two weeks of support, then your choice. Every pilot includes two weeks of support after go-live. After that, keeping it running is a monthly plan, rolling, with 30 days' notice.
05 / Prices
What it costs.
What moves the price: the number of systems to connect, how clean the data is, and whether it must run on your own infrastructure. We tell you which applies on the first call.
| Getting started | |
|---|---|
| Discovery callA 30-minute conversation about the problem you are trying to solve. No pitch, no obligation. Within a week we send a short written note saying whether we think AI is the right tool, which of our services fits, and a rough idea of cost. | Free |
| AI Fit AssessmentTwo to three days reviewing one process in detail, ending in a written recommendation: build it with AI, buy an off-the-shelf tool, fix it a simpler way, or leave it alone. Includes a cost-benefit check and a fixed price for a pilot if we recommend one. Credited against the pilot if you go ahead, so for clients who proceed the assessment is effectively free. | £1,500fixed fee |
| Building it | |
| Pilot, smallOne process, one system to connect to, reasonably clean data. Typically around four weeks. Scope agreed in writing before we start; the price does not move unless the scope does. Runs on your real data. Includes two weeks of support after go-live. | from £6,000fixed price |
| Pilot, standardOne process, but several systems to connect to, messy or inconsistent data, or human review steps built in. Typically around six weeks. Same terms as the small pilot. | from £12,000fixed price |
| Self-hosted deploymentAdded to a pilot when your compliance or data policy requires it to run on servers you control rather than our UK/EU hosting. Covers setup, hardening and handover documentation. | from £3,000 |
| Running it | |
| EssentialMonitoring, security patches, and fixes when something breaks. Nothing more. For a stable solution that does not need to change. | £400a month |
| StandardEverything in Essential, plus model and provider updates, small improvements, and one support day a month for questions, tweaks and training. | £900a month |
| PriorityEverything in Standard, plus three improvement days a month and same-working-day response. For solutions that are business-critical or still evolving. | £1,800a month |
| Hosting and AI provider costsCloud hosting and AI provider usage are passed through on your invoice with no mark-up, so you see exactly what the solution costs to run each month. | At cost |
| Anything else | |
| Work outside scopeAnything not covered by a pilot or a monthly plan, quoted and agreed before we start. We do not bill by the hour on projects. | £850a day |
- All prices exclude VAT.
- Fixed-price work is invoiced 50% on signing and 50% on delivery.
- Monthly plans are invoiced monthly in advance, rolling, and cancelled with 30 days' notice.
06 / Who
A family-run software company.
CIENVITA AI is the AI practice of CIENVITA LTD, based in County Durham. Ziya Suzen does the engineering, after three decades across internet infrastructure, public transport, investment banking and telecoms. Rachel Suzen runs the business side and keeps clients supported from the first call. More about us at cienvita.com.
07 / Contact
Tell us about the problem.
A few lines is plenty. One of us will reply from this address.
info@cienvita.com