CalQuity LabsBook a discovery call

Financial services

More judgement, less document chasing.

AI agents that collect and check customer documents on WhatsApp, run suitable models on your infrastructure and answer routine service questions, alongside your CRM, shared drives and Excel. Decisions stay with your people.

Isometric pen-and-pencil drawing of a credit officer reviewing a draft brief linked to source documents, beside open loan files, a colleague with a summary card and an unused approval stamp.

A loan file waits three days for a bank statement that arrived as a blurred photo. Sensitive case notes need processing, but the team has to establish where that data will go. A customer asks the same question for the third time because nobody can see what was said before. Most of the delay is in collecting, checking and finding. Very little of it is in the decision.

Connects to the tools you already use

  • WhatsApp
  • Gmail
  • Outlook
  • Excel
  • Google Sheets
  • Google Drive
  • SharePoint
  • Salesforce
  • HubSpot
  • Microsoft Teams

Anything else connects through its API or exports.

Where the hours go

Four desks where an agent earns its place.

Each is a routine your team repeats every day. The agent does the checking and the chasing. Your people make the call.

Isometric pen-and-pencil drawing of a field agent photographing customer documents, sent as a chat photo to a CASE FILE beside a CHECKLIST with a MISSING tag.

Onboarding and operations

Customer documents, collected on WhatsApp.

Today
Customers and field staff send PAN cards, bank statements and signed forms as photos, PDFs and emails. Operations sorts them by hand and chases what is missing, often more than once.
With an agent
Each photo or file is read, named and filed against the right customer and case. The agent checks the file against the document list for that product and asks the customer on WhatsApp for anything missing, unreadable or out of date.
Who decides
Your team verifies the documents and approves the case. The agent only collects, sorts and asks.

Works inWhatsAppGmailGoogle DriveSalesforce

Isometric drawing of engineers testing a model on a laptop connected to a local server, beside deployment and evaluation documents.

Local models

Process sensitive case notes inside your environment.

Today
A team needs to classify case notes or prepare short internal summaries, while controlling where customer information is processed.
With an agent
CalQuity evaluates suitable models on permitted sample notes, adapts the chosen model where needed and deploys it on company-controlled hardware. Quality, response time and operating cost are tested at expected load.
Who decides
IT approves access, logging and outbound connections. Staff review outputs; credit and investment decisions stay with authorised people.

Works inExcelGoogle DriveSharePointOutlook

Isometric drawing of two support agents in headsets beside an automated answering unit that files simple requests and passes one complex request to a tray labelled QUEUE.

Customer service

Answer routine questions from the customer’s own account.

Today
Customers ask about statements, EMI dates and document status, and each question means someone opening the account.
With an agent
Routine questions are answered on WhatsApp or by phone from approved answers and the customer’s account. Advice, complaints, account changes and anything sensitive go to staff with the conversation attached.
Who decides
Your team approves every answer the agent may give and handles anything that needs judgement.

Works inWhatsAppSalesforceHubSpotGmail

Isometric drawing of a man asking a question at a laptop that answers with source markers, in front of shelves labelled MANUALS and POLICIES and a locked server rack.

Compliance and policy

Find the circular that applies, with its date.

Today
Staff search shared drives and email for the current policy, product note or regulator circular, and an older version is easy to pick up by mistake.
With an agent
Ask in plain words and get the answer from the current approved document, with its date and reference shown. Superseded or conflicting guidance is flagged.
Who decides
Compliance decides which documents are current and who can see each one.

Works inSharePointGoogle DriveMicrosoft TeamsOutlook

A sample run

One loan file, checked as the documents arrive.

Case 7731 · Sample applicantagainst Personal loan checklist
DocumentRequiredReceivedCheck
PAN cardClear copyPhoto, clearReceived
Bank statementsLast 6 months4 months2 months missing
Address proofDated within 3 monthsPhoto, date unreadableUnreadable
Application formSigned on every pagePage 3 unsignedSignature missing
  1. The applicant sends four photos and a PDF on WhatsApp.
  2. The agent names each file, files it against case 7731 and checks it against the personal loan checklist.
  3. Three gaps are found: two months of statements, an address proof with an unreadable date and an unsigned page 3.
  4. The applicant gets one WhatsApp message listing exactly what to resend.

When the last document arrives, the file goes to the credit officer complete, with the original photos attached for verification.

Sample data. Names and figures are illustrative, not from a client.

How you will know

Four numbers that show whether it worked.

We record these before building anything. Source accuracy and access controls are tested too, because speed alone is not the goal.

  • Time to a complete file.

    From the first document arriving to the case being ready for review.

  • Documents asked for twice.

    How often a customer is asked again for something they already sent.

  • Local model quality and operating cost.

    Task accuracy, response time under load and total cost per accepted task, including staff review.

  • Repeat service enquiries.

    How often the same customer asks the same question, alongside handling time.

Before we begin

What financial services teams ask first.

Does customer data leave our control?

It does not have to. Private and on-premise deployment are options, and access follows the roles your team already uses.

Will the agent make credit or investment decisions?

No. It collects, checks, drafts and answers routine questions. Credit, investment and significant account decisions stay with authorised people.

Does this make our process compliant?

Not on its own. Security, retention and deployment are reviewed against your own requirements, and your compliance team decides what is acceptable.

Customers send documents in every format.

Photos, PDFs, scans and email attachments are all read. A file that cannot be read reliably is asked for again, never guessed at.

Will a local model be faster and cheaper?

It depends on the task, hardware and usage. We compare output quality, response time under load and full operating cost before recommending deployment. Local processing also needs access, logging and network controls.

Start with the context

Bring one case that took too long.

A loan file that waited on documents, a recurring task that needs local processing or an enquiry that bounced between teams. Bring it with the person who signs off, and we will work out together what an agent could take off their desk.

or email us directly athello@calquity.com