Blog · Published · Updated
AI knowledge base with customer-specific access
Short answer
An AI knowledge base for signed-in customers joins the shared product instructions with that customer's agreements and answers from sources available only to their organisation, titles and links included. At Syntalith an AI app on company documents costs from €3,000 net. At 600 customer questions a month with 10 minutes of searching each, it pays back in 1.0 month.
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 bases: statistics 2026
1.0 mo.
payback of a €3,000 app at 600 customer questions a month: €3,490 of searching today, €533 with the app. Scenario.
15%
According to Brynjolfsson, Li and Raymond (2025), customer support agents working with AI resolved 15% more issues an hour on average, in a study of 5,172 people.
Source: Brynjolfsson, Li and Raymond, The Quarterly Journal of Economics (2025) ↗
47%
According to the European Commission (2026), customer service is the most common use of AI among EU SMEs and mid-caps that already use it: 47%.
Source: European Commission, Flash Eurobarometer 578 (2026) ↗
48.8%
According to Eurostat (2025), 48.8% of EU enterprises that considered AI but did not use it named concerns about data protection and privacy.
49%
According to Anthropic (2024), combining contextual embeddings with BM25 search cut failed retrievals by 49%, and by 67% with reranking.
Figures checked: 22 October 2026
Two companies ask how to prepare
The product instruction describes how to prepare a product list file for import. Customer A's contract names a contact in the implementation team to discuss that file with. Customer B agreed a different way of working in a separate document. An employee of customer A signs in and asks what to do before the rollout starts. Today the account manager gathers this information by hand at every question.
- The app selects sources by permission before it prepares an answer
- Titles and links of other customers do not appear in the result
- The account manager's internal notes stay out of the customer's sources
- When a customer's employee leaves, their access to sources and saved conversations ends
- A question beyond the documents goes to the account manager, without new terms being added
Asking the model to keep things confidential does not replace restricting the sources. In the saved screens of a reading room example you can see the question, the source and the user's access scope on synthetic data, without running a model.
What a customer knowledge base costs and when it pays back
- Today
- 600 customer questions × 10 min = 100 h (€3,490)
- With the system, a person checks the output in 15% of the time
- €524
- Model, 1 call per case, Claude Sonnet 5.5
- €9.18
- Cost of one case
- €5.82 today, €0.89 with the system
- Saving
- €2,957 a month
Payback of €3,000
1.0 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.
The model is a small item here. The result depends on the time that searching and asking around take today, and on how much of it is left for checking the answer.
Before we quote a build, we run a test on 20 to 100 questions the team has already asked: we compare the app's answer with what an employee found by hand and measure the time from the question to the source. The test takes 5 working days and costs €0.
Could your existing help centre do the job?
If a customer looks for a few well-written articles, start with your current help centre. Zendesk describes user segments that limit signed-in users' access to articles by organisation, tags or individual users. Your own app makes sense when customers ask in their own words and need an instruction joined with an individual agreement.
A few articles, access by organisation
segments in the help centre
no new app
Instruction plus the customer's agreements, questions in their own words
an AI app with access per customer
from €3,000
Answers for site visitors who are not signed in
an AI chatbot
a separate scope
Where to start
Recurring customer questions and a description of where customer service finds the answers are enough to begin. For the first scope, choose content already meant for external readers, and the account manager will say what needs completing before it is shared. The team checks the answers from the accounts of different customers, including similar questions. An assistant for site visitors who are not signed in is described on the AI chatbot for your company page.
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 →
