Skip to content
For fintech

Careful answers about money, with a person in the loop.

Callario’s AI agent answers from the policies you approve, checks every turn for prompt injection, grounding and toxicity, and hands off when it isn’t sure. In Suggest mode it only drafts — a person on your team sends. Every change to what the AI may say or do lands on a log nobody can quietly edit.

Get early accessBook a demo
  • Safety checks on every turn, in Arabic and English
  • Suggest mode: the AI drafts, your team sends
  • A hash-chained log of every change
WWafraSupport · replies are reviewed

Can I raise my daily transfer limit? Rent is due on Friday.

Omar · Wafra support

Yes — in the app, open Limits → Daily transfers. You can raise it up to the maximum for your verification level straight away; above that, a short ID check comes first.

Perfect, thank you.

What Callario did

  1. Message screened for prompt injection
  2. Draft grounded in “Transfer limits”An approved help-centre article
  3. Suggest mode: drafted, not sentWaiting for a person on the team
  4. Omar reviewed it and sent it

01In Arabic, for a signed-in customer

It explains what happened.
Fraud is for your people to judge.

For a customer your app has verified, the agent reads their account through your own endpoint and explains it from your policy, in Arabic. When the customer says it wasn’t them, your instructions tell it to bring in a person — and it hands the conversation to your fraud team with a briefing.

ووفرةالمساعد الذكي · متصل

بطاقتي لا تعمل منذ الصباح. ما المشكلة؟

بطاقتك مجمّدة تلقائيًا بعد ثلاث محاولات خاطئة لإدخال الرقم السري اليوم. يمكنك إلغاء التجميد من التطبيق: البطاقات ← إلغاء التجميد، بعد تأكيد رمز الدخول.

المصادر:تجميد البطاقة وإلغاؤه

لكني لم أُدخل أي رقم سري اليوم!

شكرًا لإبلاغنا. أحلت محادثتك إلى فريق مكافحة الاحتيال مع التفاصيل، وسيرد عليك أحد الزملاء هنا. أبقِ البطاقة مجمّدة حتى ذلك الحين.

تم تحويل المحادثة إلى فريق مكافحة الاحتيال

Behind the conversation

  1. Customer verified by your appIdentity signed by the Wafra app
  2. Screened for prompt injectionArabic and English checks
  3. Read the account recordCard status, through your endpoint
  4. Cited “Freezing and unfreezing your card”Arabic help-centre article
  5. Possible fraud — handed to the fraud teamWith an AI briefing; the AI stops replying

02Built for regulated teams

Controls you can show
to your risk team.

GuardrailsPer-turn checks, a test gate before publishing, an emergency stop and a tamper-evident change log.AI · Screens every message for prompt injection in Arabic and English, and checks each reply against its sources.AI AgentIn your app and on your site — Autonomous, Suggest or Triage, set per deployment.AI · Answers limit, card and verification questions from approved articles, and reads account status only for verified customers.Knowledge BasePolicy articles with review, versions and a pinned published revision.AI · Gap detection shows the questions your approved content doesn’t answer yet.Contacts & CRMConsent per purpose, GDPR export and erasure, and a legal hold when records must be kept.AI · The AI sees only the fields you allow, filtered for consent and personal data.Workflows & RulesLimit reviews, complaints and account exceptions, each with an approval step.AI · Describe the process and AI drafts it; an approval holds the run until the right person signs off.Reporting & QAA run trace for every AI conversation, with personal details scrubbed before storage.AI · AutoQA scores AI conversations against your rubric, and evals must pass before a publish.

03What changes

Faster answers, without new exposure.

  • A chatbot that improvises about limits, card blocks and verification.

    Answers come from approved articles; a reply its sources don’t support isn’t sent on its own.

  • Anyone in the chat can ask about any account.

    Account details are read only for customers your app has verified, through your endpoint.

  • Nobody can say who changed the bot’s instructions, or when.

    Every change, publish and takeover sits on a hash-chained log you can verify.

  • Arabic-speaking customers get a translated afterthought.

    The same agent answers in Arabic, right to left, and screens Arabic messages for injection too.

04Honest about scope

What your risk team can rely on today, and what isn’t built yet.

Live today

  • Prompt-injection, grounding and toxicity checks on every turn
  • Suggest mode, so a person sends every reply
  • Account lookups only for verified customers, through your endpoint
  • A hash-chained log of configuration changes, publishes and takeovers
  • Consent, GDPR export and erasure — and no training on your data
  • WhatsApp and other messaging-app conversations, under the same AI checks

Not built yet

  • SOC 2 report
  • SSO / SAML
  • A choice of data residency

05Straight answers

The questions your risk
and compliance teams will ask.

Can the AI move money or change an account?
No. It can read a record through an endpoint you provide, for a verified customer, and start one of your workflows — which can wait for a person to approve before it calls your API. It has no way to touch a payment or an account on its own.
What stops someone talking it into another customer’s details?
A lookup runs only when your app has signed who the customer is. Callario sends your endpoint that proven identity — never one typed into the chat — so your system returns only that customer’s records. Every message is screened for prompt injection in Arabic and English first, and a blocked one goes to a person.
Can we keep a person on every reply?
Yes. Set a deployment to Suggest and the AI only drafts — someone on your team reads and sends. Triage mode goes further and leaves all answering to your team.
What exactly is on the audit log?
Changes to what the AI may say and do — instructions, handoff rules, tool permissions, publishes, pauses — and actions such as a takeover, each entry chained to the one before. Individual AI replies live in each conversation’s run trace, with personal details scrubbed, not on the hash chain.
Are you SOC 2 certified? Do you offer SSO or data residency?
Not yet. SOC 2 work has not started, SSO/SAML is not built, and Callario runs in a single hosting region with no residency choice. Today each workspace is isolated in the database with row-level security, with roles, multi-factor sign-in, and GDPR export and erasure. More on our security page.
Design-partner programme · now onboarding

Put it in front of
real customers.

We’re onboarding a small group of teams ready to run their support on Callario’s AI — and to tell us where it breaks. Bring your real conversations; we’ll bring the platform and the people who build it.

Get early accessBook a demo
  • No training on your data
  • English and Arabic
  • Human takeover, any time

What design partners get

  • The whole platform, set up with youAI agent, copilot, workflows, inbox and knowledge — configured around your real conversations.
  • A direct line to the buildersTalk to the people who write the code, not a support queue.
  • A say in what ships nextYour feedback shapes the roadmap, week by week.