Blog · Published · Updated
Ticket triage: how the workflow works
Short answer
The model that assigns residents' tickets in our demonstrator scored macro-F1 0.828, below the 0.850 quality threshold, so the routing gate failed. We publish the result because it shows why a dispatcher approves each assignment and why urgent and ambiguous cases stay in a manual queue. At Syntalith an automation costs from €3,000 net. At 1,000 resident tickets a month of 4 minutes each, it pays back in 2.2 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
- analysis with an engineer and a free prototype
- €0
- team in ChatGPT or Claude, up to 10 people
- from €1,150
- automation, agent or application for one process
- from €3,000
We belong to the OpenAI and Anthropic partner programmes and learn through their official courses. What this means ↓
Author: Syntalith · 3 min read
Statistics
Tickets and AI: statistics 2026
2.2 mo.
payback of a €3,000 automation at 1,000 resident tickets a month: €2,327 of work today, €936 with checking and the model. Scenario.
0.828
According to the Syntalith demonstrator (2026), the model that assigns residents' tickets scored macro-F1 0.828 against a quality threshold of 0.850, so the routing gate failed.
Source: Syntalith, resident ticket queue demonstrator (2026) ↗
400
According to the Syntalith demonstrator (2026), the public run of the ticket queue covered 400 tickets from 40 buildings and 24 months of history.
Source: Syntalith, resident ticket queue demonstrator (2026) ↗
90%
According to the European Commission (2026), 90% of EU companies that use AI say people take the decisions.
Source: European Commission, Flash Eurobarometer 578 (2026) ↗
24.8%
According to Eurostat (2025), 24.8% of EU enterprises in real estate activities used AI.
Figures checked: 22 October 2026
When similar tickets describe one fault
Residents' tickets have to be linked to the right case and the right team. Several residents report the same lift failure in different words and through different channels. Assigning an urgent problem wrongly can be worse than leaving it for manual review. The demonstrator merges repeat tickets, sets a priority by safety and deadlines, and hands the decision to the dispatcher. It does not contact contractors and does not close cases.
The public run covered 400 tickets from 40 buildings and 24 months of history. The assignment model scored macro-F1 0.828 against a threshold of 0.850. After the run we added a taxonomy correction and high-precision boundary rules, but they have not yet passed a new live evaluation, so the failure remains the valid result.
| Acceptance scenario on 100 tickets | Result |
|---|---|
| Assignment proposals | 80 of 100, coverage 80% |
| Accepted by the reviewer | 73 of 80, accuracy 91.25% |
| Urgent cases detected | 18 of 20, so 90% |
| Urgent cases missed | 2, to fix before launch |
- A quality gate only means something when its failure is visible
- A missed urgent case matters more than average accuracy
- Fixes without a new evaluation do not change the result
- The client sees what to expect from a pilot
What it costs and when it pays back
- Today
- 1,000 resident tickets × 4 min = 66.7 h (€2,327)
- With the system, a person checks the output in 40% of the time
- €931
- Model, 1 call per case, Claude Sonnet 5.5
- €5
- Cost of one case
- €2.33 today, €0.94 with the system
- Saving
- €1,391 a month
Payback of €3,000
2.2 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: 2,000 input tokens and 200 output tokens per call at Anthropic's rates, at the ECB rate of 1.1205 USD per euro on 22 October 2026.
The count covers the work around a case: searching, reading, writing and handing over. What remains is the time to check the result, 40% 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 free test that takes 5 working days.
The count keeps 40% of today's work because the dispatcher approves every assignment and reviews the cases with no proposal. In the demonstrator, the full recorded run of 24 tickets cost $0.010717.
What this means for a pilot
A model that scores below the threshold does not assign tickets on its own. The dispatcher approves each assignment and reviews the cases with no proposal, and urgent cases and ambiguous addresses stay in a manual queue. The figures in the acceptance scenario are the exercise's assumptions: 2 missed urgent cases out of 20 need fixing before launch. Anthropic, in its guide to evaluating agents, advises judging the outcome in the environment and repeating the trials, because a single run easily misleads.
Few tickets, the dispatcher keeps up
a shared queue with no model
no build
Many tickets, repeat reports of the same fault
assignment proposals approved by the dispatcher
from €3,000
The model has passed the gate on the property manager's tickets
automatic routing of non-urgent cases
priced after the analysis
Where to start
A pilot starts with the dispatcher judging the recommendations on tickets whose labels an employee has approved. The scope for property managers is described in AI for property managers. We build this kind of rollout as part of AI automations. For the free analysis, a description of one case from recent weeks and of the tools it passed 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: resident tickets
We run the same count for your buildings in the free analysis.
€2,792
a month for handling 800 resident requests by hand
Payback: 1.3 mo.
6 minutes to receive, assign and answer a request 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: reading a document in our case file system (about €0.0002) plus a description of what is missing (Claude Sonnet 5.5, about 2,000 + 300 tokens), at the ECB rate of 1.1205 USD per euro on 22 October 2026.
See the full calculation →
