skip to main contentSkip to main content
Passer au contenu principal

Voir cette page en English ?

— BEAMDESK / HELPDESK ENTERPRISEMMXXVI · BUILD 4.2.1

Un helpdesk qui
lit les tickets, lance les outils,
et boucle le dossier.

Beam oriente chaque conversation vers le bon modèle parmi treize, recherche dans l’index vectoriel de votre connaissance, appelle vos propres API quand une preuve manque, puis rédige une réponse. Les agents gèrent les exceptions. Le reste est enregistré, audité et facturé en crédits — pas au résultat.

Commencer gratuitement →Demander un briefing

Adopté par des partenaires de lancement dans :

SaaS Série BStartup marketplaceFintechDevToolsSantéE-commerce

« Nous avons remplacé 60 % de nos coûts de support niveau 1 par les résolutions automatiques de Beamdesk en 3 semaines. »

VP Support · SaaS Série B · 50 agents
FIG. 01 — ESPACE AGENT / CAPTURE LIVE3 sept. 2026, 20:00 UTC
ROUTE CURRENCY
Prices follow the active locale: EUR.
EUR
§ 06 — PRICING

Seat + AI work.
Honest margins.

Flat seat fee for features and access. Metered AI resolution overage on top — same pattern as Intercom, unchanged across regions. WhatsApp costs pass through at 0% markup. Audit our invoice against Meta Business Manager anytime.

HOW MANY SEATS?
Price updates live. More than 50 seats goes to sales.

Free

NO CARD

For side projects and prototypes.

€0
/ mo
AIFlash 2.5 + Haiku 4.5

1 seat

25 AI resolutions (workspace total)

No overage · hard cap

Start free →

Starter

Small support teams testing the water.

€18
/ seat / mo
AIFlash · Haiku · Sonnet 4.6
1 seats × 18 € = 18 €/mois

Up to 10 seats

50 AI resolutions / seat / mo

0,25 € / overage

1 AI persona · standard (50 articles)

  • ·all 11 channels
  • ·routing rules + macros
  • ·CSAT + NPS collection

Pro

RECOMMENDED

Growing teams running AI in production.

€45
/ seat / mo
AISonnet 4.6 · Opus 4.7 on complex
1 seats × 45 € = 45 €/mois

Up to 25 seats

200 AI resolutions / seat / mo

0,20 € / overage

5 AI personas · extended (500 articles)

  • ·all 11 channels
  • ·escalation chains + SLA timers
  • ·hallucination guard

Scale

High-volume teams optimising quality.

€119
/ seat / mo
AIOpus 4.7 default · BYOK
1 seats × 119 € = 119 €/mois

Up to 100 seats

500 AI resolutions / seat / mo

0,15 € / overage

Unlimited AI agents · unlimited articles

  • ·all 11 channels
  • ·all integrations + API access
  • ·multi-model consensus

Enterprise

CUSTOM

Custom — talk to us.

Custom pricing
AICustom models · BYOK · VPC
50+ seats minimum
  • ·SOC 2 Type II report
  • ·BAA on request
  • ·dedicated CSM
Talk to sales →
HOW IT BILLS

Seat fee × seats, plus resolution overage. Worked example for a 5-seat Pro workspace that handled 1,150 AI resolutions this month:

Pro seat fee5 × 45 €225,00 €
Included resolutions5 × 200 = 1,000—
Overage150 × 0,20 €30,00 €
Monthly total255,00 €
Scale volume reference13 × 119 €1 547 €

Resolution pools are workspace-wide — a seat under quota subsidises one over. Set a hard monthly cap and Beam escalates to human review instead of blowing the budget.

ADD-ONS & OVERAGES
Extra human seatsame seat price · up to tier cap
Extra AI resolutions0,25 € / 0,20 € / 0,15 € (Starter / Pro / Scale)
Extra AI agentsupgrade tier · no per-agent fee
WhatsApp messagesMeta rate · 0% markup
Annual billing−15% on seat fee

Regional PPP. Available in 100+ regions — tiered seat discounts at checkout.

WHATSAPP COST · 0% MARKUP
Meta's rate, line-item visible on every invoice.

We don't mark up WhatsApp. Marketing / Utility / Authentication templates pass through at the per-country rate Meta publishes — US Marketing ~$0.025/msg, Brazil Marketing ~€0.0685/msg, India Auth-International ~$2.30/msg. Customer-initiated service messages (72h window) remain free. Your invoice shows every line. Audit it against your Meta Business Manager anytime.

PROPULSÉ PAR /
OpenAI GPT-5 ·Anthropic Claude 4.7 ·Google Gemini 3.1 Pro ·Stripe ·Cloudflare ·Supabase
§ 07 — QUESTIONS

Frequently asked.

How does the routing actually decide?+

Each inbound message is scored for intent and complexity against a small local classifier. That determines a shortlist of eligible models. Beam picks the cheapest model whose historical quality on similar tickets clears your configured bar. You can override per queue or per customer.

What counts as a "resolution"?+

A resolution is a ticket where AI produced a reply that was sent to the customer (auto-send or agent-approved from the draft queue). Drafts discarded by agents, internal notes, and re-runs on the same ticket don't count. Tickets you close manually without AI involvement don't count either.

When does overage kick in, and what does it cost?+

Your seat count × included resolutions is the monthly pool. At Pro that's 200/seat — 5 seats = 1,000 resolutions. The 1,001st resolution in that month bills at your tier's overage rate ($0.25 Starter, $0.20 Pro, $0.15 Scale). You set a hard cap per workspace; AI escalates to human review instead of blowing the budget.

How does the regional discount work?+

Seat fee (only) is discounted by region: Tier 1 (US/UK/DE/AU/Nordics) pays 100%, Tier 2 (FR/JP/KR/UAE) pays 75%, Tier 3 (PL/BR/IN/ZA) pays 50%. Detection uses billing address + IP + card BIN country — all three must agree or you default to Tier 1. Overage and WhatsApp costs are uniform globally because token prices don't have regions.

What about WhatsApp costs?+

Pass-through at 0% markup, broken out as separate invoice lines at Meta's published per-country rate. US Marketing ~$0.025/msg, Brazil Marketing ~€0.0685/msg, India Auth-International ~$2.30/msg. Customer-initiated service messages (inside the 72h window) stay free. Audit against your Meta Business Manager anytime.

What runs on-prem vs. in the cloud?+

Core app and Qdrant run in your region (EU by default). Model calls are forwarded to provider APIs with zero-retention headers. SSO, audit log and KB embeddings never leave your tenant.

Can we import from Zendesk or Intercom?+

Yes. CSV + API importers cover tickets, contacts, macros and KB. Email routing continues during migration via a forwarding phase.

Is my data used to train models?+

No. All provider calls carry zero-retention flags. Your KB embeddings are yours; the gap log is not shared cross-tenant.

GET STARTED
Free tier. 25 resolutions. No card.
Upgrade when you hit the cap.
Start freeBook demo