Blog · Published · Updated
Local AI for building systems project documents
Short answer
A local AI assistant points a building services engineer to the requirement in the specification and shows it next to the supplier's material. The engineer sees the gap and has a basis for a question, and the project files never leave the office. At Syntalith an app on a local model costs from €3,000 net plus GPU hardware. With 400 questions about specifications a month at 10 minutes each, it pays back in 2.0 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
Design documentation and AI: statistics 2026
2.0 mo.
payback of a €3,000 local app at 400 specification questions a month: €2,327 of work today, €796 with checking and a GPU server. Scenario.
146
According to Bartanowicz and Jassem (LREC 2026), in a RAG system on Polish technical documentation assessed on 146 reference questions, Bielik-11B-v2.3-Instruct consistently beat PLLuM-12B-nc-chat.
49%
According to Anthropic (2024), combining contextual embeddings with BM25 search cut failed retrievals by 49%, and by 67% with reranking.
25.1%
According to a Harvard Business School study (2023) of 758 consultants, tasks within the model's capabilities went 25.1% faster with AI.
40.4%
According to Eurostat (2025), 40.4% of EU companies in professional, scientific and technical activities used AI.
Source: Eurostat, Use of artificial intelligence in enterprises (2025) ↗
Figures checked: 22 October 2026
A points list required, a brochure delivered
The specification requires a list of controls points prepared for this project. The supplier's folder holds a general brochure for a range of devices. The engineer asks where the requirement is written and whether the material received covers it.
The assistant finds the passage in the specification and points out that the brochure describes the range in general, and the material contains no project points list. The engineer opens both documents, checks the scope and sends the supplier a question: „Please send the controls points list for this project, as the specified passage of the specification requires.” Technical assessment, calculations and the choice of solutions stay with the engineer.
- 01Questionthe requirement described in the engineer's own words
- 02Specificationthe passage with its version and a link
- 03Supplier materialthe document that should answer the requirement
- 04Gapwhat is missing in the material
- 05Question to the supplierwith a reference to the passage of the specification
What local AI for building services engineers costs and when it pays back
- Today
- 400 specification questions × 10 min = 66.7 h × €34.90 = €2,327
- With the app, an employee checks the output in 25% of the time
- €582
- Hetzner GEX45 GPU server
- €214
- Saving
- €1,531 a month
Payback of €3,000
2.0 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, 25% of today's work in the scenario. We measure both times in a trial on 20 to 100 cases the team has already handled. Anthropic, in Contextual Retrieval, reports that combining contextual embeddings, BM25 search and reranking cut failed retrievals by 67%. In specifications full of standard numbers and device labels, keyword search next to search by meaning matters a great deal.
What runs locally and what needs agreeing
If the office does not allow project files to be sent to an outside service, processing happens on a named computer or server. LM Studio describes local work with a downloaded model and documents without internet, while searching for and downloading models and updates use the network. A PDF with selectable text differs from a scanned page, where the text first has to be read from the image, so an illegible passage may drop out of the results. In the assessment we compare answers with the document even when the search finds no passage.
- 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
When your current search is enough
If the team knows the document references and uses search in its documentation environment well, start there. Project and version filters and PDF search may already lead to the right place. A model is worth comparing when engineers remember the meaning of a requirement but search for it in other words, or put together passages from several documents. Questions about revision, unit and a missing condition are covered in AI for engineers: how to test a tool.
The team knows the references and searches by number
search in the current environment
no AI
Requirements searched in other words, files may go to the cloud
search with sources in an approved service
from €3,000
Project files stay in the office
a local app on a GPU server
from €3,000 plus hardware
Where to start
The first scope covers one current project: the specification in force and the suppliers' materials. Someone has to make sure the newer version of a document reaches the search and is kept apart from the older one. 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: a local model in a design office
We run the same count for your office in the free process review.
€2,618
a month for the work on 300 questions
EU cloud version payback: 1.4 mo.
Our assumptions are 15 minutes to find an earlier agreement in the documentation, and checking in 15% of the old time. €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 6,000 + 500 tokens per answer in the EU cloud version, at the ECB rate of 1.1205 USD per euro on 22 October 2026. We count the payback for an app with a model in the EU cloud (from €3,000 net). We price the model on your server after a call, and the cost of people's work stays the same.
See the full calculation →
