Case study · I run a contact center

The AI takes the routine. Agents keep the emergencies.

Sofratel's 200-seat contact center runs the 24/7 on-call line of 200 social housing operators. A CallShift agent now answers first: it handles routine requests end to end and hands urgent and sensitive calls to Sofratel's agents, with the context. Valophis measured the result on its own line.

Sofratel × ValophisContact center · Social housing

  • AI + agents
  • 24/7 emergency line
  • Mitel
  • French
Agents at their desks on a bright contact center floor
Sofratel × Valophis
01 Results

What changed, in figures.

45%of on-call calls handled end to end by the AI at Valophis, with no agent involved

Calls answered on Valophis's line
Before CallShift89%
With CallShift98%
  • 10,000calls a day at peak
  • 200social housing operators served
  • 24/7emergency line, nights and weekends included

Valophis figures, measured on its on-call line, not a contractual commitment.

02 The need

Where they started.

Sofratel answers the on-call line of 200 social housing operators: evenings, nights and weekends, when the operators' offices are closed. Many of those calls are not emergencies (a dripping tap, a radiator that stays cold, a lost badge, a rent receipt), and every one of them still took an agent.

Then come the peaks: opening time, rent reminders, the heating season, up to 10,000 calls a day. A queue is the one thing an emergency line cannot afford.

03 What we deployed

What runs in production.

Position
In front of the contact center: the AI answers first, in each operator's name
Rules
Each operator's own specifications decide what is urgent
Routine calls
Handled end to end, a callback booked in the agency's opening hours
Urgent and sensitive
Passed to a Sofratel on-call agent, with the context
Peaks
Every call answered in parallel: no queue
Hours
24/7: evenings, nights, weekends
Telephony
Mitel
Language
French
Data
Hosted in France, GDPR compliant
04 How it works

The AI in front, the agents behind.

  1. Every call answered

    The AI picks up at once, in the operator's name, however many call at the same time.

  2. Qualified against the operator's rules

    The same specifications the operator already uses decide what is urgent: nothing to rewrite.

  3. Routine requests: done

    The AI answers, gives the caller the first step (shut off the water, check the breakers) and books a callback in the agency's opening hours.

  4. Emergencies: to an agent, with the context

    Urgent and sensitive calls go up to a Sofratel on-call agent, with the context of the call.

Routine requests the AI handles end to end

From an operator's specifications: the list follows each operator's own, and is not exhaustive.

  • Plumbing and drainsLeak with a working valve · Dripping tap or cistern · Blocked toilet, one home
  • Heating and hot waterTemperature too low · One radiator not heating · Electric water heater, no leak
  • ElectricityOutage with no sparks or smoke · Smoke detector, no smoke · Entry code, intercom
  • Doors and locksDoor slammed shut · Lost or stolen keys or badge · Window, roller shutter
  • PestsInsects, rodents · Bed bugs · Wasps, hornets
  • NeighboursNoise, barking · Obstructive parking
  • Common areasDirt · Bulky waste · Bulb out on a landing
  • Roof and outsideSlight leak, a few tiles · Broken branches, no storm
  • Ventilation, TV aerialVentilation (not gas) · Aerial or cable
  • AdminRent receipt, transfer · Badge, customer account · Appointments
05 In their words

From the people who run it.

Observed in production at Valophis

Fewer nuisance calls, more time per case, and the peaks (opening time, rent reminders, the heating season) absorbed with no extra staff and no outsourced overflow.

06 This case fits you if

Recognise yourself?

  • You answer calls for several clients, sites or brands, each with its own rules.
  • Part of your traffic is routine, and peaks swamp the floor.
  • You want your agents on the calls that need a person, with the context.
Next step

See it on your own calls.

Book a call with a voice AI engineer: we start from your real call flows and show you the agent, the architecture and the numbers that fit them.

1:1with a voice AI engineer, no slides
2architectures we ship in production: speech-to-speech and pipelines
1rollout plan to leave with: volume, pricing, telephony