Blog · Published · Updated
AI has the answer. Does it sound like your company?
Short answer
When the team keeps shortening and rearranging AI replies, first name the edits that recur and add them to style guidance with examples. Fine-tuning helps when the model returns to the same habits despite that guidance. In a scenario of 2,000 customer replies a month, style guidance with examples cuts errors from 30% to 12%, and fine-tuning to 5%, which together save 16.7 hours of corrections, or €582 a 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
- 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
Customer service and AI: statistics 2026
16.7 h
fewer corrections a month at 2,000 customer replies: style guidance with examples cuts errors from 30% to 12%, fine-tuning to 5%, at 2 min per correction. Scenario.
40%
According to a study published in Science (2023), ChatGPT cut the time 453 professionals spent on writing tasks by 40%.
Source: Noy and Zhang, Science (2023) ↗
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) ↗
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) ↗
14–56%
According to Anthropic (2025), studies with a control group found smaller time savings, from 14% to 56%.
Source: Anthropic, Estimating AI productivity gains (2025) ↗
Figures checked: 22 October 2026
The same information, less rewriting
A service company wants to answer customers directly and politely. AI drafts an appointment confirmation: „Dear Sir, we are pleased to inform you that the visit of our technician has been confirmed for Tuesday in the afternoon hours”. The editor changes it to: „Hello, we confirm the technician's visit on Tuesday afternoon”. The customer gets the same information, and the message sounds like the company.
A single edit is small. Across hundreds of similar replies, the team keeps making the same edit. „Write in our style” says little when everyone reads style differently. An example shows the concrete choices: a normal greeting, the information first, no officialese.
- Ceremonial openings the editor deletes
- The key information moved to the start
- Reply length by type of case
- Phrases the company does not use
A short appointment confirmation and a reply to a complex complaint need different lengths. When the model always has to write briefly, a complaint gets a reply that is too curt, so the style rules depend on the type of case.
What a simpler change gives, and what training adds
- Today: 30% errors × 2,000 × 2 min
- 20 h (€698)
- After a change without training (style guidance with examples): 12%
- 8 h (€279)
- After fine-tuning: 5%
- 3.3 h (€116)
- The simpler change saves 12 h, fine-tuning adds 4.7 h (€163) a month
Compare fine-tuning with the cost of examples, training and a new evaluation
The error shares and the correction time are the scenario's assumptions, which we measure on the client's cases not used for training. €34.90 an hour is the EU average hourly labour cost for 2025 according to Eurostat.
In this count, style guidance with examples gives most of the effect for a fraction of the cost. Training makes sense at a larger scale or when the remaining errors are expensive.
- Evaluation cases kept apart from training examples
- The same set for the current model, the simpler change and the adapted model
- Correct answers prepared by the people who correct the results today
- Errors counted separately for each kind of mistake
- A plan for a new evaluation after a change of base model or process
The set of training examples has to be maintained: when the process, the naming or the base model changes, the evaluation is repeated on the same test set. We put that cost in the quote together with the training itself.
Simpler options before training
A recurring date confirmation can use a template with no model. For varied replies, style guidance with a few before-and-after examples helps. Google describes tuning the style of answers, for example conciseness, when instructions alone give inconsistent results. Current information, such as the appointment date, still has to be given with every message.
Recurring confirmations
a template with no model
no training
Varied replies, edits you can name
style guidance with examples
no training
The model returns to its habits despite the guidance
fine-tuning on the team's edits
after a trial
Draft replies in the team's working tool
We assess the trial on messages not used for adaptation: staff prepare replies with the current tool and with the adapted version and measure the editing time. We adapt models as part of our private LLMs, and we build the app where an employee sees the result next to its source as part of our AI apps. For the free analysis, a few examples of an error that keeps coming back and a note of who corrects it today are enough.
Worked example · Scenario
How we count the payback: AI in customer service
We run the same count for your replies in the free analysis.
€5,817
a month for answering 2,000 customer emails
Payback: 0.6 mo.
5 minutes to read an email and write a reply 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 3,000 + 600 tokens per message with a draft reply, at the ECB rate of 1.1205 USD per euro on 22 October 2026.
See the full calculation →
