Blog · Published · Updated
Local AI for confidential proposals
Short answer
A local AI model suits confidential proposals when data rules rule out your current tool. The decision covers quality on your documents, user access, performance and responsibility for maintenance, and buying a server is only one of those elements. At Syntalith an app on a local model costs from €3,000 net plus GPU hardware. With 40 proposals a month at 120 minutes each, it pays back in 1.7 months.
How we workDetails ↓
- 01Call30 minutes with an engineer, free
- 02PrototypeA working system on your process and data, free
- 03ImplementationTarget and fixed price in the contract; 50% at signing
- 04Acceptance50% once the agreed target is met
- model on your server or in the cloud in the EU
- quoted after the review
- same work in application with EU cloud model
- from €3,000
- task, data and equipment analysis
- €0
We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓
Author: Syntalith · 3 min read
Statistics
Confidential data and AI: statistics 2026
1.7 mo.
payback of a €3,000 local app at 40 proposals a month: €2,792 of work today, €1,052 with checking and a GPU server. Scenario.
48.8%
According to Eurostat (2025), 48.8% of EU companies that considered AI and did not use it named concerns about data protection and privacy.
Source: Eurostat, Use of artificial intelligence in enterprises (2025) ↗
49%
According to the European Commission (2026), 49% of EU SMEs and mid-caps name legal uncertainty as a barrier to using AI.
Source: European Commission, Flash Eurobarometer 578 (2026) ↗
52.5%
According to Eurostat (2025), 52.5% of EU companies that considered AI and did not use it named a lack of clarity about the legal consequences.
Source: Eurostat, Use of artificial intelligence in enterprises (2025) ↗
83%
According to Smits and Walkowiak (EACL 2026), in RAG on the Polish Wikipedia assessed on 1,000 PolQA questions, PLLuM-12B matched the quality of 70B models with 83% less energy.
Figures checked: 22 October 2026
Establish what your data rules forbid
An industrial proposal contains the client's drawings, supplier terms, your own costing and a description of an unannounced project. The team wants to compare requirements faster or prepare a draft reply, and has no permission to put the material into the chat tool it uses.
„Data must not leave the company” can mean a ban on public tools, a required region or full isolation from the network. These are different conditions. The person responsible for security and for the client contract confirms which one applies. The location of the model does not describe the whole flow: material also ends up in logs, backups, the document-reading tool and the support system. A label saying „on-premise” does not confirm compliance with the contract.
| Option | What to settle |
|---|---|
| An approved API | processing terms, features and data restrictions |
| A private cloud environment | location, access control, maintenance and dependencies |
| Your own infrastructure | quality on the task, capacity, updates and support |
- 01Resulta summary of the proposal's requirements with the source shown
- 02Controlwhat the salesperson checks and what the engineer checks
- 03Trialtables, scanned pages, unclear requirements and long material
- 04No basisthe system reveals a gap instead of adding a commitment
- 05Loadone person working, and several people at once
What a local model for confidential proposals costs and when it pays back
- Today
- 40 proposals × 120 min = 80 h × €34.90 = €2,792
- With the app, an employee checks the output in 30% of the time
- €838
- Hetzner GEX45 GPU server
- €214
- Saving
- €1,740 a month
Payback of €3,000
1.7 months
The volume, the time and the checking share are the scenario's assumptions, which we measure in the trial. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat. Price of the GEX45 with 24 GB of GPU memory from Hetzner's website on 22 October 2026, without the one-off setup fee of €209.
The count covers the time spent searching, reading and compiling the material. What stays is the time to check the output against the source, 30% of today's work in the scenario. We measure both times in a trial on 20 to 100 cases the team has already handled. We leave 30% of today's work for checking because a salesperson or an engineer reads every commitment in a proposal.
What runs locally and what needs agreeing
In a commercial proposal, an unconfirmed commitment weighs more than a few seconds of response time. So the trial covers situations with no correct answer: the system has to reveal that there is no basis. A trial on one laptop assesses a single task, and readiness for a department needs a load check with several people at once. If the main difficulty is finding the current version of a proposal, you also need a knowledge base with versions, because the model alone does not know which document applies.
- Where the app takes documents from and where it saves results
- What stays in logs, backups and the history of questions
- Who has access to questions, answers and summaries
- Who updates the model, the app and the set of documents
Who is responsible for running it
Name the owner of the infrastructure, updates, access and user tickets. Ask who will check quality after a model change and what happens during an outage. A small local model may be enough for narrow work, and you should not assume it will be better or cheaper than every outside service. A study discussed by IBM shows that companies choose models for cost, security and fit. It is a vendor's study and does not point to a solution for your company's documents.
The contract allows processing by a provider
an approved API with data terms
licence and tokens
A required region and access control
a private cloud environment
priced after the review
The material may not leave the company
a local model with an app for proposals
from €3,000 plus hardware
Where to start
For the first call, prepare a description of the confidentiality requirements, the type of documents and the number of people who would use the solution. We compare the whole period of use: preparation, hardware, maintenance, changes and checking work. We roll out local models as part of private LLMs, and we build the view for working with sources as an AI app. For the free process review, a description of one case and the places where the material sits is enough. Confidential files stay in the company, and you get a written plan within 2 working days.
Worked example · Scenario
How we count the payback: preparing proposals
We run the same count for your proposals in the free process review.
€2,094
a month for preparing 120 quotes from customer enquiries by hand
Payback: 1.7 mo.
30 minutes per quote from an enquiry and checking in 15% of the old time are our assumptions. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat, for companies with 10 or more employees. Model cost: Claude Sonnet 5.5, about 4,000 + 1,500 tokens per draft quote, at the ECB rate of 1.1205 USD per euro on 22 October 2026. The extra margin is a scenario: we assume that 1 in 40 enquiries quoted the same day ends in an order that a slow reply loses today, with a margin of €344. In the review we count it on your sales process.
See the full calculation →
