| Support — the rootevery request, at the desk |
Every ticket is classified and answered by hand, and the conversation with the customer is a phase of every one of the 330 tickets that carry one. 148 fixed tickets sit in Pending without a resolution date, and an approved write backlog sat at 39 of 51 issues for twelve weeks.Support Module § 0.2Data and information § 0.2Stage Classification § 1 |
The agent classifies on request and drafts the reply — about 7 ticket folders per working day with a median of 3 human turns, on one machine, for the ticket a person hands it. Nothing watches the channels, and nothing decides what can wait.Baseline § 2–3Metrics |
No operator on duty at the desk: every request classified, identified and routed the moment it arrives, its clocks running, and what can wait held with its reason and its age. Judged by: the classification correction rate and the re-routing rate; breach warnings raised ahead of the violation; no case waiting without a recorded reason; human turns per case falling to gate decisions.Support Module § 1–2Metrics |
| Configurationone product or tariff change |
9–11 days median to close a configuration or master-data request, against 1–2 days for a correction, over 175 customer-closed tickets. A written specification covers 4 to 6 of the 12 configuration parts, so the rest is discovered per surface.Data and information § 0.2Configuration goals CG-2 |
There is no configuration agent. This is manual labour. A person reads the documents, decides the configuration, writes it and executes it, across four surfaces in two databases with nothing enforcing consistency between them. The one end-to-end machine run — product 9951 — was a single-purpose sibling on one environment, with no sessions, no gates, no audit and no deployment stage.Baseline § 6 |
From files and a short conversation: a product preview within a business day, and customer acceptance within five business days. Judged by: elapsed time from intake to a preview the business can look at (target one business day); elapsed days from intake to customer acceptance against the 9–11 day median (target five business days); unresolved items at the specification gate (at most five) and at close (none); zero unreachable-product incidents.Configuration CG-13Metrics |
| Developmentone change |
106.75 developer-days estimated over 46 dual-estimated rows of one delivery file; the change-request schedule at 179 days. A customer-specific change waits behind vendor communication and the full change procedure.Baseline § 3 |
49.5 developer-days on the same rows — 2.16×; 2.5× personally over one month; the schedule at 136.1 days, −24 %, on technical phases only. A median of 18 human interventions per session.Baseline § 3Metrics |
Any small or customer-specific delivery inside hours or days, safely — instead of vendor correspondence and a long procedure. Judged by: elapsed time from brief to a verified deploy on the named target; human interventions per case; share of changes landing entirely in customer-specific seams (target 80 %, every shared-code landing with a recorded proposition); the port to master opened within a business day of production verification; reverts at most one in twenty deploys, executed within the hour.Development DG-9Metrics |
| Data & informationone correction or answer |
About 40 minutes of a skilled person's own time on the median ticket (0.63 h, adjusted). One person leads 33 of 35 shapes, with half or more of the work on 25 of them; seven shapes have at most two holders with more than one ticket, and the team's own judgement adds print, multi-year, the three cargo families, currency and process registration as knowledge that sits with one or two people.Value and ROI § 2bData and information § 0.2 — knowledge holders |
1.5 h of agent time (2.0 h measured, adjusted −25 %), median over 185 tickets with a reliable timing (02.09.2026), of which the operator spends roughly 15–30 minutes. It produces the written analysis, the ticket notes and the memory entry as by-products, and it runs on one machine at a time. Three retraction cycles came from pattern-matching a past ticket into a confident root cause.Value and ROI § 2bBaseline § 5 |
No reliance on person-specific knowledge: the work can run without any particular human. Judged by: the count of shapes with two or more operators who have closed a case with the skill; the seven thin shapes reduced to none and every shape closed by at least two operators; reuse recorded on at least 60 % of cases within two quarters; verifier and auditor at most a quarter of a case's reasoning; retraction cycles at zero; retrieval on held-out cases at or above 50 % top-5 hit-rate per domain, against the measured 10 % search baseline, before that domain is accepted.Support SG-3, SG-10Agents Memory and Skills § 7.2 |
