Blog · Published · Updated
AI for customer follow-up after a release
Short answer
After a feature release, an AI agent links the approved release description with earlier customer requests in the CRM and support, and proposes a contact list to the account manager with a reason and a draft message. The account manager rejects poor matches and sends the messages personally. The agent costs from €3,000 net at Syntalith and, with 200 follow-ups after releases a month at 12 minutes each, pays back in 3.1 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 agent for one process, build paid once
- from €3,000
- from signing the contract to acceptance
- 6–16 weeks
- free prototype on your cases
- €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 success and AI: statistics 2026
3.1 mo.
payback of a €3,000 agent with 200 follow-ups after releases a month: €1,396 of work today, €425 with checking and the model. 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 on AI and algorithmic management (2026) ↗
34.7%
According to Eurostat (2025), 34.7% of EU enterprises that use AI use it for marketing or sales.
Source: Eurostat, Use of artificial intelligence in enterprises (2025) ↗
15%
According to Brynjolfsson, Li and Raymond (2025), customer support agents working with AI resolved on average 15% more issues an hour, in a study of 5,172 people.
Source: Brynjolfsson, Li and Raymond, The Quarterly Journal of Economics (2025) ↗
Figures checked: 22 October 2026
Go back to the need in the customer's own words
A customer asked whether the chosen report columns could be kept, so they would not have to pick them every time they open the report. In the salesperson's note, the request appears as „too much report setup”. The next release adds saving the column layout. Nobody links the two records, because they use different words.
The agent links the conversation with the description of the released change and shows the account manager both sources. It proposes a short message that goes back to the earlier problem and points to the new option described in the release materials. The account manager checks whether the feature is available in the customer's plan. Another customer asked for filters to be remembered: similar words do not mean that saving columns meets their need, so the review makes it easy to reject such a contact.
- Every contact has a source: a record of a conversation or a ticket with its date
- The message stays within the approved release description
- The account manager checks whether the feature is available in the customer's plan
- A customer who has already been contacted does not appear again
The customer's reply goes into the same case. It may explain that the need has changed or calls for a wider conversation with the product team.
What it costs and when it pays back
- Today
- 200 follow-ups after releases × 12 min = 40 h (€1,396)
- With the system, a person checks the output in 30% of the time
- €419
- Model, 2 calls per case, Claude Sonnet 5.5
- €6
- Cost of one case
- €6.98 today, €2.12 with the system
- Saving
- €971 a month
Payback of €3,000
3.1 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 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 time covers finding earlier requests, checking the customer's plan and writing the message.
The count covers the work around a case: searching, reading, writing and passing things on. What remains is the time to check the output, 30% of today's work in the scenario. Before the quote we measure both times on 20 to 100 cases the team has already handled, in a €0 test that takes 5 working days.
A release description is a useful start
Jira creates release notes from the issues assigned to a version, and Rovo can prepare a draft for publishing in Confluence. That answers the question of what was released. A newsletter will inform every recipient, and it will not refer to the problem recorded for a specific customer. If customer requests are already linked to Jira issues, a report of those links is enough. An agent makes sense when requests sit in salespeople's notes and tickets, written in the customers' own words.
Requests linked to Jira issues
a report of the links after the release
configuration
A few releases a year, few customers
a manual review with a list of requests
no build
Requests in CRM notes and tickets, frequent releases
an agent that matches requests with releases
from €3,000
Where to start
The first scope connects one source of release information with a chosen area of the CRM. On a finished release we look with the account managers at which contacts the agent proposes and which ones they reject. We build this as part of AI agents. For the free process review, a description of one case from recent weeks and the tools it ran through is enough. You do not need to send confidential documents, and you get a written plan within 2 working days.
Worked example · Scenario
How we count the payback: CRM in B2B sales
We run the same count for your customers in the free process review.
€698
a month for entering 300 CRM updates by hand
Payback: 5.1 mo.
4 minutes per CRM update 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 + 500 tokens per update, 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 400 updates that get lost today ends in an order, with a margin of €459. In the review we count it on your sales process.
See the full calculation →
