Home · Solutions · Customer service
Solution · Customer serviceThe site director sees the complaint before the importer's survey does
Every complaint in one register, whatever the channel
Complaints from every channel and the importer's survey comments land in one register with the site, brand, vehicle and deadline attached; a person confirms the category and owns the reply.
Executive summary
Your sites log complaints their own way, and the importer's score moves before the group knows what happened.
The register comes first, and it is deliberately dull: one list, one case per complaint, carrying the site, the brand, the vehicle, the channel.
Every complaint sits in one place with an owner and the date it arrived, so a case cannot fall between a site mailbox.
the complaint register in Microsoft Lists; case files on SharePoint; the Power BI group model
Business problem
Customer experience
A complaint in a dealer group is not one thing. It reaches the service reception as a raised voice, the site mailbox as an email about a repair that did not hold, the marketing inbox as a public review, and the manufacturer weeks later as a sentence typed into a survey box. Every channel has a different owner, a different tool and a different idea of what counts as a complaint. Some sites keep a spreadsheet, some a mailbox folder, some rely on the adviser remembering.
The consequences suit nobody. A site handles what it can see and closes it, so its own numbers look reasonable. The group learns the pattern at the quarterly review, from a report built by counting whatever the sites wrote down. The manufacturer measures satisfaction directly with the group's customers and reports a score that sits in the bonus conversation; by then the customers who moved it have been served, invoiced and in some cases lost. What changes what a customer says next time is somebody calling back quickly and putting it right, which depends on a person noticing in time.
How it works today
This is how a complaint travels through a multi-brand group today, whichever dealer management system each site runs.
- PersonAn adviser or receptionist takes the complaint at the counter, by phone or from the site mailbox, and writes it into a site spreadsheet if that site keeps one
- WaitingThe case waits for the adviser to come off the workshop floor; any reply deadline is counted in somebody's head, if at all
- PersonA marketing coordinator watches the social and review inboxes and forwards what looks serious to a site manager
- SystemSatisfaction results arrive as a monthly importer export; the comments are read once, for the review slide, and never joined to a case
- Risk of errorOne customer and fault can sit in three records, so a repeat looks like three first contacts and the group's figures say the site is fine
Why the current process costs more than it appears
The most expensive part of this process has no cost line.
- Counting what each site chose to record is not measurement. A site with a diligent adviser records more complaints than a site with none and looks worse for it, which is the wrong incentive to put in front of a site director.
- Deadlines live in heads. Without one register the group cannot say afterwards when a complaint arrived, who saw it and when the answer went out, which is the evidence it needs the day a case stops being a conversation.
- Survey comments are the cheapest customer research a dealer group is given, and one person reads them once for a slide. A sentence naming a recurring fault gets the same minute as one praising the coffee.
Cost of inaction
The score the importer publishes moves on its own schedule, and the rows above price only the desk work behind it. What they cannot price is the part that decides the score: the customer who wrote once, was answered late or not at all, and said so in a survey nobody read for a month. Nor do they price the case that escalates with no register behind it, where the group's position rests on an adviser's memory of last spring. Nothing here fails on any particular day, which is why it survives.
A plausible organisation with realistic proportions. The figures are there to be recalculated on your data; they are not a client result.
An illustrative Polish dealer group: nine sites, four brands, showrooms and authorised workshops, about 700 staff, a dealer management system per brand plus the importer portals, Microsoft 365 across the group.
640 complaints a month across the counter, the site mailboxes, the web form and the social and review pages, plus 2,300 responses a month to the manufacturers' satisfaction surveys, a fifth with a written comment.
Each site logs what it receives, in its own way. Marketing forwards public posts. The survey export is opened once a month for a slide. No single list holds the case, the owner and the date the clock started.
About 16 minutes per case of registering, finding the vehicle and owner, deciding which site owns it and keeping the record; and no reliable way to see a repeat, a pattern or a case running out of time.
Every channel, survey comments included, feeds one register. UiPath Communications Mining proposes the category, the site, the brand and a sentiment reading; an adviser confirms on a card in Microsoft Teams. The clock starts at receipt and is visible to the owner and the site director. A detractor comment opens a same-day call-back, once the consent and objection record for that channel has been checked.
In the modelled case the group reads complaints by site, brand and category with the scores beside them from the first month, cases stop disappearing between channels, and detractor call-backs happen in hours rather than after the next export. None of that was measured at a client.
Proposed solution
The register comes first, and it is deliberately dull: one list, one case per complaint, carrying the site, the brand, the vehicle, the channel, the owner and the dates it arrived and was answered. Everything else exists to fill that list without anyone retyping. UiPath Integration Service collects the mailboxes and the web form; the social and review inboxes come in through the tools marketing already uses; satisfaction results are read from the importer's export as a file, which is how they reach a dealer group.
UiPath Communications Mining reads the text, whatever channel it came from, and proposes a category, a site, a brand and a sentiment reading, with survey comments treated as one more stream rather than as reporting. A proposal stays a proposal: the case card arrives in Microsoft Teams and a person confirms or corrects it in seconds, which is also how the model improves. From there the deadline clock runs against the rule your legal adviser wrote for that complaint type, the card returns before it expires, and the group view in Power BI shows cases by site, brand and category with the survey scores beside them. A detractor comment opens a call-back the same day, once the flow has checked the consent and objection record for the channel it would use.
UiPath Communications Mining in UiPath IXP (taxonomy trained by annotation, sentiment, extraction fields, multilingual streams); UiPath Integration Service connectors for Microsoft Outlook 365, Microsoft Teams and Microsoft OneDrive & SharePoint; UiPath Action Center tasks completed inside Microsoft Teams; UiPath Orchestrator queues, triggers and audit
The register; the category set, agreed with your service and legal people and trained on your own messages; the deadline rule table and its escalations; the case card, reminder and call-back task in Teams; the consent check before every outbound contact; the case file on SharePoint; the Power BI group model
Mapping the mailboxes, the web form, the social and review inboxes and the importer survey exports into Communications Mining streams; reading vehicle, owner, site and consent state from the group's customer records
How the automated process works
- AutomationNew mail in the mailboxes, web-form entries, social and review messages and each survey export are collected and queued in UiPath Orchestrator
- AutomationCommunications Mining labels every item with a category and a sentiment reading and extracts the register fields: site, brand, vehicle, customer, channel
- PersonThe case card reaches the owning site in Microsoft Teams; an adviser confirms or corrects category and owner, and the case opens with its receipt date
- AutomationThe clock runs against the rule for that complaint type; a reminder reaches the owner and the site director before it expires, and an unanswered case escalates
- AutomationA detractor comment is checked against the consent and objection record for its channel, then opened as a same-day call-back or left for the next visit
- AutomationThe answer, the decision and the documents behind them go to the case file on SharePoint; the Power BI model updates overnight
Human-in-the-loop model
Automation handles
- Collecting every channel, survey comments included, and proposing a category, a site, a brand and a sentiment reading
- Running the clock from the receipt date, reminding the owner, escalating what is unanswered, keeping the case file complete
- Checking consent and objection before any outbound contact, and building the group view and the monthly pack
People decide
- Whether the proposed category and the owning site are right, confirmed on the card in seconds
- What the customer is told, what is offered, and whether a case is settled, refused or handed to the legal adviser
- The rules themselves: the categories, the deadline table, the escalation thresholds and what counts as a complaint
Before and after
Systems and integrations
Every entry can be checked in vendor documentation. The evidence class is stated next to each one.
Inputs
- service and sales mailboxes in Microsoft 365
- the group's web complaint form
- social and review inboxes
- the importer's satisfaction export
- complaints taken at the counter
Automation layer
- UiPath Communications Mining (IXP)
- UiPath Orchestrator
- UiPath Robots
- UiPath Integration Service
- UiPath Action Center
Target systems
- the complaint register in Microsoft Lists
- case files on SharePoint
- the Power BI group model
- the group's customer record
Human touchpoints: case cards and deadline reminders in Microsoft Teams; category confirmation in Action Center; the same-day call-back; the monthly group review
Technologies used
reads complaint mail, public reviews and survey comments; proposes category, sentiment and the register fields
Aqueue every item, run the deadline clocks and escalations, retry, log and audit
Acollects the mailboxes, posts the cards, files the evidence
Acategory and owner confirmed without leaving Teams
Athe case card, the deadline reminder, the call-back task, the monthly review
Acases by site, brand and category with the scores beside them
Avehicle, owner, site, brand and consent state on each case
CIllustrative economic model
The arithmetic is open, so it can be argued with.
Sixteen minutes is the administrative life of a case in this model, not the repair behind it: registering it, finding the vehicle and the owner, deciding which site owns it, keeping the record and closing it. €23 an hour is a fully loaded cost for a service adviser in a Polish dealer group. Both are assumptions of this scenario, and the survey responses are not priced here; only the registered complaints are.
Run the numbers on your data
An illustrative estimate from your own inputs. It models released capacity; it is not a promise of savings.
Business benefits
- Every complaint sits in one place with an owner and the date it arrived, so a case cannot fall between a site mailbox, a review page and a survey export
- Deadlines stop depending on memory: the clock starts at receipt, the reminder arrives before it expires, and the group can show afterwards what happened and when
- A detractor comment produces a call-back in hours instead of after the next export, while what the customer says next can still be changed
- The group reads complaints by site, brand and category with the scores beside them, so a recurring fault or a struggling reception desk shows up in weeks rather than at the quarterly review
The management view
- Sites are compared on the same definitions, because the category set, the clock and the closing rule are the same at all nine rather than nine local habits
- The satisfaction score stops being a number that arrives from outside: each comment sits beside the case it belongs to, so the argument is about causes, not the score
- Escalations reach the aftersales director while the case is open, and goodwill given to settle one becomes a visible cost with a reason attached
Board-level KPIs
Security and governance
Security is designed with the process, not after it.
- Customer correspondence stays inside your Microsoft 365 tenant and is read from the EU region of UiPath Automation Cloud; no copy of a complaint goes to a third-party analytics service
- Ingestion runs under a dedicated application identity, with mail permissions scoped to the named mailboxes and read-only access to the customer records; every robot step is in the Orchestrator log
- Consent and objection state is read before every outbound contact, and an objection recorded at one site applies at all of them because the suppression record is held once for the group; the deadline table, the category set and the escalation thresholds are versioned and change only with sign-off from the complaint-policy owner
Why now
Satisfaction is one of the few things a manufacturer measures directly with your customers, and the survey runs on the importer's calendar whether or not the group has an answer ready; the modelled €3,925 a month of administrative time is the smaller half of the cost
Consent and objection have to work across the group, not showroom by showroom: a customer may object to direct marketing at any time under Article 21(2) of the GDPR, and where brands sit in separate legal entities each is its own controller. A register carrying the consent state per channel is the cheapest place to make that true
Communications Mining is a supervised model over your own text rather than a general-purpose language model, so a category means the same thing every month, which is what makes a complaint count defensible in front of an importer
Relevant executive roles
Sites are compared on one definition of a complaint, not on how diligently each writes things down
Open cases, ageing and repeat faults are visible per site and brand while something can still be done
Goodwill given, cases settled and administrative time become a line with a reason attached, not a paragraph of site commentary
Common questions and objections
The register is filled by the flow, not the adviser. Case, vehicle, channel and receipt date are proposed and confirmed in seconds on a card in Teams. What disappears is the spreadsheet, the forwarded email and the mailbox search when the case returns three months later.
The report is what this uses. Results usually reach a group as an export, and the comments in it name the site, the visit and often the vehicle. Reading them as one more channel costs nothing extra, and it is the only version of the customer's view the manufacturer also reads.
It is not trusted to. It proposes, a person confirms, and those confirmations train the model. Where the proposal is weak the case still opens, without a category, and still has an owner and a running clock.
When this is not the right solution
- A single site with one adviser and a handful of complaints a week, where a shared list and a rule about who answers costs less than a register and a trained model
- No agreement about what counts as a complaint: if each site keeps its own definition after the project, the register produces a better-documented disagreement and little else
A question for the next management meeting
If a customer complains at one of our sites tomorrow and tells the importer's survey about it a month later, what in this group connects those two records, and by when were we supposed to have answered the first one?
Implementation approach
Delivery runs in stages, so it can be stopped at any point.
We deliver
- A read of one quarter of your complaint traffic: the channels, what each site records, how often one customer appears twice
- The category set and the annotation programme, trained by your own service and customer-care people on your own messages
- The register, the deadline rule table and the escalation path, agreed with whoever owns your complaint policy
- The case card, reminder and call-back task in Microsoft Teams, with the consent check before every outbound contact
- The Power BI group model, the monthly review format and a runbook for the reception desks
We need from you
- Twelve months of complaint correspondence and the importer satisfaction exports for the brands in scope
- A process owner in aftersales, a customer-care lead and a counterpart for the deadline and consent rules
- Access for a dedicated service identity to the mailboxes, the web form and the customer records
Stages
Discovery
Channels, volumes, what the sites record, where one case appears twice
Design
The category set, the register, the deadline table, the escalations, the consent check
Build
Ingestion and labelling, the Teams card and call-back, the case file, the group model
Go-live
A parallel run at one site and one brand, then rollout with hypercare and handover
Departmental. Effort is driven by the number of channels and brands, by how much argument there is about what counts as a complaint, and by how reliably a vehicle and owner can be found from the message.
The importer knows what your customers think before your site directors do.
Send us one month of complaint traffic: the channels it arrives through, what each site records, and the last satisfaction export for one brand. We come back with a draft category set and a written read-out of where cases go missing between the channels.
Trace one complaint across your channelsThe neighbouring process usually has the same problem
Your CSAT score gives you a number. Your customers' emails give you the reason, and nobody reads them.
View solution Customer serviceRepair status by message, additional work approved by a tapStop parking cars on the ramp while an adviser dials voicemail for a yes the customer would have tapped in a minute.
View solution Customer serviceService quality measured on every call, not on five a monthYour agents are rated on five calls out of five hundred, and the other 495 are never opened.
View solutionIndustries we deliver this in most oftenAutomotive retail