Blog · Published · Updated
AI answers from manual photos and tables
Short answer
An AI knowledge base on illustrated manuals has to join each photo to its caption and each table row to its heading, and show the answer next to the passage it comes from. When the document does not let it tell which variant is meant, the app says so. At Syntalith an AI app on company documents costs from €3,000 net. At 400 questions a month with 10 minutes of searching each, it pays back in 1.5 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
- net per application, fixed price before contract
- from €3,000
- from signing the contract to acceptance
- 4–10 weeks
- free prototype on your documents
- €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
AI knowledge retrieval: statistics 2026
1.5 mo.
payback of a €3,000 app at 400 manual questions a month: €2,327 of searching today, €355 with checking and the model. Scenario.
49%
According to Anthropic (2024), combining contextual embeddings with BM25 search cut failed retrievals by 49%, and by 67% with reranking.
11 of 13
According to the NoLiMa study (2025), at 32,000 tokens of context 11 of 13 models fell below half of their short-context performance.
20.0%
According to Eurostat (2025), 20.0% of EU companies with at least 10 employees used AI.
Source: Eurostat, 20% of EU enterprises use AI technologies (2025) ↗
50%
According to the European Commission (2026), half of EU SMEs and mid-caps name high investment costs as a barrier to AI.
Source: European Commission, Flash Eurobarometer 578 (2026) ↗
Figures checked: 22 October 2026
When the caption changes what a photo means
An employee is preparing display material for a shop. The manual shows photos of two similar stands, and a table lists the inserts for each variant. The question is: „Which insert goes with the stand with the wide base?”. The answer needs the photo, the caption and the table row joined together.
- 01NameThe app gives the name of the insert exactly as the document does.
- 02PhotoIt shows the illustration the answer rests on.
- 03TableIt points to the row and the heading that describes it.
- 04PageIt takes you to the right page of a long manual.
- 05GapWhen the caption is illegible, it says so and names the person who owns the material.
A link to the whole manual leaves the work to the user. Whether the app can point to the exact place depends on the format of the material, so look at it on your own files before you widen the scope.
A table needs its headings
Once text is copied out of a table, names and values sit side by side with no visible link. A heading that groups columns, or a note under the table, decides what a row refers to. A photo without a caption may show several items, of which only one is the subject of the question.
Microsoft describes the Layout model in Azure AI Document Intelligence as a tool that extracts text, tables, figures and the structure of a document. That is one part of preparing the sources. Being able to recognise the elements of a page does not prove that the assistant reads the instruction correctly.
Figure descriptions
improve
Someone who knows the product writes what the photo shows.
Legible files
replace
An illegible page image is replaced by a file with a text layer.
Scope of questions
limit
When the material is inconsistent, the app does not answer questions it cannot settle.
What a knowledge base on illustrated manuals costs and when it pays back
- Today
- 400 questions × 10 min = 66.7 h (€2,327)
- With the system, a person checks the output in 15% of the time
- €349
- Model, 1 call per case, Claude Sonnet 5.5
- €5.97
- Cost of one case
- €5.82 today, €0.89 with the system
- Saving
- €1,972 a month
Payback of €3,000
1.5 months
The volume, the time and the checking share are the scenario's assumptions. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat. Model: 6,000 input tokens and 500 output tokens per call at Anthropic's rates, converted at the ECB rate of 1.1205 USD per euro on 22 October 2026.
According to the NoLiMa study (2025), at 32,000 tokens of context 11 of 13 models fell below half of their short-context performance. That is why we base the answer on the passage found, with its illustration, and do not paste the whole manual into the model.
When a better manual is enough
A few short instructions
table of contents, captions, a clear preview
no new app
Lots of material, the employee does not know where to look
an AI app with the answer next to the passage
from €3,000
Illegible images and unexplained markings
improve the documents first
before AI
Questions to an assistant will not fix an illegible page image or explain a marking its author never described. Judge the answers with someone who knows the material and with an employee who will use it, because the second will notice when a correct passage is hard to read at their workstation.
Bring a question that text alone cannot answer
For the call, bring a question that needs a look at a photo or a table, and the name of the tool where you keep the manuals. The scope is described on the private ChatGPT for your company page, and we test work with illustrations separately on your material, in a test on 20 to 100 questions the team has already asked.
Worked example · Scenario
How we count the payback: answers from company documents
We run the same count for your documents in the free analysis.
€3,723
a month for finding answers to 800 questions in documents
Payback: 1 mo.
Our assumptions are 8 minutes to find an answer in documents or ask a colleague, 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 from documents, at the ECB rate of 1.1205 USD per euro on 22 October 2026.
See the full calculation →
