Customer storySwiftreg logo

Number plate tickets answered from live order data, under rules Swiftreg wrote itself

Two eesel agents on Zendesk. One handles support tickets. The other reads private sellers’ replies and tells the buyers waiting for that registration.

Industry
Private number plates
Based in
United Kingdom
With eesel since
June 2026
2,100Zendesk tickets handled in a month8 Aug to 8 Sep 2026
1,500public replies sent to customers1,494 in the same month
1,050internal notes left for the teamwhere a step belongs to staff
630Private Sales runsseller replies and offers, same month
swiftreg.co.uk
Swiftreg website
How it runs

What happens when a customer writes to Swiftreg

  1. 1
    A ticket arrives in Zendesk

    New tickets, customer replies or private sellers in Private Sales.

  2. 2
    eesel checks who is asking

    Order details go only to the verified account holder.

  3. 3
    It runs Swiftreg’s own lookups

    Scripts check orders and tracking, and a validator checks each reply.

  4. 4
    It replies, leaves a note, or stays quiet

    A public reply marked as AI, or a note when a person owns the next step.

Connected integrations

  • Google Drive
  • Slack
  • Zendesk
  • wWebsite

Agents at work

  • eesel1,961 runs in 30 days
  • Private Sales629 runs in 30 days

The company

Swiftreg sells private number plates in the UK. Some registrations come from its own stock or the DVLA. Many are listed by private sellers, with Swiftreg between seller and buyer. Every sale runs through the DVLA transfer, the certificates, the V5C log book and the physical plates posted by Royal Mail.

Visit Swiftreg at swiftreg.co.uk

The problem

A number plate order is not one status. The sale, the ownership, the DVLA assignment, the certificate, the plates and the delivery are separate states, and a customer often asks about two or three in one email. A reply carries the status a lookup actually confirmed, never the next step along. Requested is not processing, and offered is not issued.

On the private sales side, every seller who confirms a registration is still for sale has buyers waiting in another queue. Matching the two by hand is repetitive and easy to get wrong.

How it runs

Swiftreg connected Zendesk in June 2026. eesel reads 889 documents from Zendesk, 74 pages from the website and 48 documents in Google Drive. The instructions are Swiftreg’s own, kept as numbered releases, and the lookups are skills with scripts the team wrote: Order Lookup, Product Lookup, Listing Management Link, Remove Listing from Sale, Cancellation and Royal Mail tracking.

  • Live facts win. A tool result beats the knowledge base, past tickets and anything the customer says. eesel never reports a refund, dispatch, DVLA submission or timescale a lookup has not confirmed.
  • Every question gets answered. Length, difficulty or several questions at once are not reasons to hand a ticket to a person. If one part belongs to staff, eesel leaves that part alone and answers the rest.
  • Some things always go to people, and a step that belongs to staff stays with them. Sellers are only ever “the seller” or “the current owner”, never named.
  • Every public reply ends with a footer saying the AI Support Assistant wrote it and may contain errors.
  • The Private Sales agent works only in the Private Sales group. When a seller confirms a registration is still available, or names a price, eesel checks the listing and the website price, finds every buyer waiting on that registration, replies to each with the approved wording, tags the ticket so an SMS goes out, thanks the seller and solves the ticket. If the seller says it is sold, it confirms the listing is marked sold before telling the buyers.

A month of work

Between 8 August and 8 September 2026 eesel handled about 2,100 Zendesk tickets for Swiftreg. It sent 1,494 public replies, left 1,055 internal notes, tagged 185 tickets and drafted 113 replies for a person to send. The Private Sales agent ran about 630 times. Most tickets land in the working day, and the ones that arrive overnight get the same treatment.

What is next

The setup keeps changing under Swiftreg’s control. The support instructions in use are labelled a candidate build for the next release, rolled out through Swiftreg’s own procedure rather than edited live. Slack is connected, so the team can ask eesel a question there as well as in Zendesk.

A month of work

What eesel did, from its own records

Every count on this page comes from eesel's activity log for Swiftreg, 8 Aug to 8 Sep 2026. The only estimate is the money and time saved, and the receipt shows how it is worked out.
$5,300
of people’s time a month, $64,000 a year
210 hours
about 5 working weeks, every month

$25 is an international benchmark for an hour of support time, not Swiftreg’s own cost.

Conversations handled, 8 Aug to 8 Sep 20262,104
×Minutes a person would spend on one6
=Hours of people’s time210
×Cost of an hour, all in$25
=A month$5,300
A year$64,000
Zendesk tickets and other live work
Live work per week, the last 13 weeks
0339678Week of 22 Jun: 17922 JunWeek of 29 Jun: 512Week of 6 Jul: 5846 JulWeek of 13 Jul: 544Week of 20 Jul: 50020 JulWeek of 27 Jul: 608Week of 3 Aug: 6673 AugWeek of 10 Aug: 676Week of 17 Aug: 67817 AugWeek of 24 Aug: 470Week of 31 Aug: 20131 AugWeek of 7 Sept: 43
Show as a table
Week ofRuns
22 Jun179
29 Jun512
6 Jul584
13 Jul544
20 Jul500
27 Jul608
3 Aug667
10 Aug676
17 Aug678
24 Aug470
31 Aug201
7 Sept43
When the work arrives
Hour of day, UTC. 8% of it lands between 10pm and 7am, when nobody is at a desk.
00:00 UTC: 1100:0001:00 UTC: 1002:00 UTC: 903:00 UTC: 704:00 UTC: 1005:00 UTC: 2006:00 UTC: 5006:0007:00 UTC: 11508:00 UTC: 17209:00 UTC: 18510:00 UTC: 18611:00 UTC: 17312:00 UTC: 15412:0013:00 UTC: 13214:00 UTC: 12815:00 UTC: 14616:00 UTC: 11817:00 UTC: 10418:00 UTC: 10118:0019:00 UTC: 7320:00 UTC: 6421:00 UTC: 5322:00 UTC: 3623:00 UTC: 23
Show as a table
Hour (UTC)Runs
00:0011
01:0010
02:009
03:007
04:0010
05:0020
06:0050
07:00115
08:00172
09:00185
10:00186
11:00173
12:00154
13:00132
14:00128
15:00146
16:00118
17:00104
18:00101
19:0073
20:0064
21:0053
22:0036
23:0023
What it did
Actions taken in the customer's tools, last 30 days
Public replies sentPublic replies sent: 1,4941,494Internal notesInternal notes: 1,0551,055Knowledge searchesKnowledge searches: 245245Tickets taggedTickets tagged: 185185Draft replies for the teamDraft replies for the team: 113113Emails to the teamEmails to the team: 1515
Show as a table
ActionCount
Public replies sent1,494
Internal notes1,055
Knowledge searches245
Tickets tagged185
Draft replies for the team113
Emails to the team15

Hire an eesel teammate for your support

Connect your helpdesk, let it learn from your past tickets, and start with drafts before it replies on its own.

Start free