Where is my payment, what is my limit, how much does a transfer cost: the answers are already in your core system. After an identity check, the assistant answers according to your rules and passes disputes and suspicious transactions straight to an employee as a ticket in Dynamics 365 or Salesforce with a conversation summary.
Built for ISO 27001NIS2 DORA
A customer service team at a payment institution receives hundreds of emails in an Outlook inbox and self-service messages a day, and most are similar: has my payment arrived, why is the transfer still processing, what is my daily limit, how much will a transfer to a third country cost, which documents do I submit to raise my limit. An employee opens the core system, Mambu or another platform, finds the customer and checks the status, then opens Dynamics 365 or Salesforce to log the contact and writes a reply. Meanwhile a customer with a real problem, a disputed transaction or a blocked account, is waiting. In the evenings and on weekends, replies slip to Monday.
The assistant works in email, the self-service portal and a chat window. In the portal the customer is already logged in; by email the assistant first confirms identity with a one-time code or a link to the portal, and without verification it answers only in general terms about fees and procedures. For a verified customer, the assistant pulls the payment status, limits or fees from the core system with read-only access and answers from an approved rule set. A question about a dispute, a suspicious transaction, a complaint or credit terms immediately becomes a ticket in Dynamics 365 Customer Service or Salesforce Service Cloud with a conversation summary, customer ID and category, and the reply to the customer goes out from the employee's Outlook. We connect to the core system and the CRM through an API, a file export or an email flow, depending on your version. The Claude model runs in an Amazon Bedrock EU region, conversations are not used for training, and every answer and handoff is logged in line with GDPR and DORA.
Four steps from input to a record in your system, with human approval wherever there is risk.
The enquiry arrives in the service Outlook inbox, the self-service chat window or the website. In the portal, identity is already confirmed by login; by email the assistant sends a one-time code or a link to the portal. Until identity is confirmed, it answers only about public fees and procedures, with no account data from the core system.
With read-only access through an API, the assistant pulls the payment status, limits, fees or the list of required documents from the core system and from the customer record in Dynamics 365 or Salesforce. It answers in the customer's language from an approved answer set, with no improvisation. Topics not in the set go automatically to handoff.
A dispute, a suspicious transaction, a complaint, a block or a question about credit terms creates a ticket in Dynamics 365 Customer Service or Salesforce Service Cloud: customer identifier, conversation summary, category and initial urgency. The customer gets a message that the enquiry has been passed on, with an expected response time. From there a person communicates from their own Outlook.
Every answer, core system query and handoff is logged with time and customer ID. The service manager reviews sample answers in a CRM report every week and flags inaccurate ones. We update the answer set and the list of blocked topics when products or fees change.
An assistant for the Outlook inbox, the self-service portal and the website chat window, with an identity check and one answer set drawn from core system data.
Creating tickets in Dynamics 365 Customer Service or Salesforce Service Cloud with a conversation summary, category and customer data for the employee.
Records of all answers and handoffs with time, topic and outcome, and a monthly CRM report on topics and handoff reasons.
A documented set of permitted topics and answers, a list of read-only queries to the core system, their updates and quality monitoring after launch.
Claude agent via API in an Amazon Bedrock EU region + read-only access to the core system and ticket creation in Dynamics 365 or Salesforce. We connect to your systems through an API, a file export or an email flow, depending on the version; we choose the method after a check. Learn more about this service and its limits on the service page.
No. The assistant has read-only access to the core system and cannot initiate a payment or change a limit or account details. It passes such requests to an employee through a Dynamics 365 or Salesforce ticket, and the employee acts according to your procedure.
We connect to Dynamics 365 and Salesforce through an API, or, if the licence does not cover the API, through an email flow into a ticket inbox. With the core system, whether Mambu or another platform, we use only a read-only API or a periodic export, depending on your version. We choose the integration method after a check in the first week.
Claude runs in an Amazon Bedrock EU region, conversations are stored in your environment according to your retention periods and are not used to train the model. We describe the flow in your GDPR record and in your DORA register of third-party providers.
The customer passes the Onfido or Sumsub check, but someone still retypes the business register extract, proof of income and beneficial owner…
A quarterly report to the financial regulator means days of collecting data from Sage, the core system and registers into Excel and checking it by…
Complaints arrive in a shared Outlook inbox, ICT incident reports come from monitoring and Teams, and deadlines run from receipt.
What is the contract termination procedure, when is enhanced customer due diligence required, who approves an exception: staff ask colleagues in…
This service is delivered by specialists with valid international certifications. Their expertise is confirmed by independent organisations, not by us.
AI systems are deployed with protection against data leakage and model abuse.
The AI security programme and its controls are managed following the ISACA methodology.
AI risks are identified and managed like any other IT risk.
We reply within 1 business day
