Case study · Customer care

Lumea: from email analysis to a live AI Email Agent in nine weeks.

An e-shop in 10 countries with 200 agents and 60,000 tickets a month. Lumea is a model company; every number is modelled and shown with its maths.

The team

200 agents on email and phone, 10 countries, 60,000 tickets a month.

Before

Every peak season meant hiring. QA heard a few calls a week. Nobody knew how many tickets repeat or what they cost.

What we did

Email Insight on 12 months, rules and approved knowledge, a test on 1,000 old tickets, live on one market, then all markets in week 9.

3 months after launch28%▲ 16,800 a monthtickets closed without a person29 FTE▲ ≈ €1.1Mof work freed a year−15%no peak hiringteam cost
~3%of calls QA actually reviews. The rest is a guess.
~23%of an average contact centre's budget goes to repeat calls.SQM Group
−1%every extra percentage point of first-call resolution (say 70% → 71%) cuts operating cost by about 1%.SQM Group
The call loop

One call on your line, start to finish.

Click any step: what your best agent does at that moment, what TheAICalls does there, and what research says it is worth.

Solved once.
Not twice.
Every call scored, every week one topic per agent.
ShiftBefore
Greet0:00
Diagnose0:30
FrustrationMid-call
SolveEnd
FeedbackSeconds after
Check7 days later
CoachEvery week
Step 1 · Before

Enough people at the peak

Your best agent
Shifts follow when customers actually call.
TheAICalls
Calls and missed calls by hour and market. Unusual topics, such as “Where is my parcel”, flagged as soon as they jump.
Worth
Fewer missed callsand a spike caught a week earlier
In the app, step by step

From the app · model company Domira · no sound

The gap, in numbers

Where the money is in customer care.

The first column is the effect research measured. The second is what we expect from coaching, before we measure your own team.

BehaviourEffectRealistic with coachingSource
Solves it the first timeFCR one point up ≈ −1% costup to +5 pts, up to −5% costSQM Group
No repeat call within 7 days~23% of budget on repeat callsmeasured on your callsSQM Group
Takes the customer's side−74% customer effortmeasured on your callsCEB
Productivity with AI support+14%, new agents +34%+14%NBER
Model case · Lumea

From email analysis to a live AI Email Agent in nine weeks.

Lumea is a model company: an e-shop in 10 countries with 200 agents and 60,000 tickets a month. Its numbers are modelled, not results of a specific client.

Week by week
  1. Weeks 1–2 · Email Insight12 months of email sorted by request type: how much the AI can answer, how much needs a person. 42% of tickets
  2. Weeks 3–4 · Rules and knowledgeThe AI answers only from approved sources. Complaints, damaged goods and legal questions always go to a person.
  3. Weeks 5–6 · Tested on 1,000 old ticketsA category goes live only when 95% or more of the AI drafts are approved.
  4. Weeks 7–8 · Live on one marketThe AI Email Agent handles tickets; anything unsure goes to a person with a draft.
  5. Week 9 · All marketsRolled out everywhere. The team handles the peak season without hiring.
What the AI can take · by request type
Where is my order
70%
Returns and refunds
45%
Complaints, damage
10%
Change or cancel
35%
Product questions
30%
Payment and invoice
40%
Other
0%
tickets a yearthe AI can answer
All tickets720,000
The AI can answer (Insight)302,400 · 42%
The AI answered252,000 · 35%
Closed without a person201,600 · 28%
−15%team cost
29 FTE≈ 56,000 hours a year, 15% of the team
≈ €1.1Mof work freed a year, at €3,000 per person a month
Model company, 3 months after launch. Deliberately cautious: Klarna reports its AI assistant took two thirds of chats in its first month.
Your numbers

How much would it save your team?

Enter your team. The example is Lumea. You see the maths line by line.

Your team
What the AI takes over
Share of tickets the AI closes
28%
0%TYPICAL 20–45% · LUMEA 28%70%
Share of calls the AI Voice Agent takes
0%
0%TYPICAL 20–40%70%
Result for your teamUpdates as you type
29FTE of work freed
€1.1Ma year
−15%of team cost
AI takes over · 29.4 FTEpeople on email · 75.6people on the phone · 45other work · 50
People on emailagents × time on contact × email share200 × 75 % × 70 % = 105
People on the phoneagents × time on contact × the rest45
Email work the AI takes overpeople on email × ticket share29.4 FTE
Value of freed workFTE × cost × 1229.4 × 3,000 € × 12 = €1.1M
A year, before our fee€1.1M

Gross saving before our automation fee, which depends on the work the AI does. The first step measures each line on your own data.

How it works

Calls and email in one place. AI Agents on top.

0

Analyse

Every call and email scored against your process. Starts on data you already have.

1

Connect

Inbox and ticketing; every agent sees their own calls and weekly topic.

2

Through us

Calls and tickets in one platform, written back to your CRM or helpdesk.

3

AI Agents

Closes or prepares the tickets that repeat. The AI Voice Agent does the same on the phone.

23+countries
1,000+agents on the platform
Millionsof minutes analysed
NotinoLoxoneand others
Next

Thirty minutes that save you a lot of guessing.

Thirty minutes, live, on our model data or straight on your own calls: we show you what is in them and what it would find across your team. If it is not useful, we will not chase you.