Talk to Us

MERCHANTPAISA PRIVATE AI COPILOT

From payment records
to clearer answers.

A private assistant designed to help your team investigate payment exceptions, understand transaction records and prepare reports using internal data they are authorized to access.

Explore the sample workspaceDiscuss your existing platform

Product Preview. Assistant foundation implemented. Live private-model inference and independent deployment remain unvalidated and disabled by default.

MerchantPaisa AI CopilotPRODUCT PREVIEW
PAYMENT OPERATIONS / SAMPLE DATA

Which failed payments
need investigation?

DEMO-MERCHANT-001 · 1–7 Sep 2026 · USD

Recorded evidence01 / REVIEW

Two failed Visa records.
USD 127.00 to investigate.

DEMO-TXN-1012 Sep 2026USD 45.00Failed
DEMO-TXN-1046 Sep 2026USD 82.00Failed

Recorded states only. No issuer decline reason is established by this sample.

Source: DEMO-FIXTURE-TXN-01
Snapshot: 8 Sep 2026, 09:00 UTC
Synthetic data · No live AI connection

01 / SAMPLE QUESTIONS

See what your team
could ask.

Choose a team to read a prewritten example. Every result uses synthetic data, not a live AI response. No customer records or AI services are connected.

MerchantPaisa / Copilot previewPRODUCT PREVIEW · SAMPLE DATA

CHOOSE YOUR TEAM

Limits & secondary examples

Showing a synthetic transaction example.

Evidence-backed sampleSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Payment operations

Which failed payments need investigation?

Illustrative scopeDEMO-MERCHANT-001 · USD
Interpreted filters1–7 Sep 2026 · Visa · Failed

Two synthetic failed Visa records match this sample merchant and period, totaling USD 127.00. The recorded outcomes need review; this fixture does not establish an issuer decline reason.

Two synthetic transaction results
ReferenceDateAmountState
DEMO-TXN-1012 Sep 2026USD 45.00Failed
DEMO-TXN-1046 Sep 2026USD 82.00Failed

Demo evidence: DEMO-FIXTURE-TXN-01

Difference for reviewSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Finance and reconciliation

Which settlement records do not match?

Illustrative scopeDEMO-MERCHANT-001 · USD
Interpreted filters7 Sep 2026 · Two settlement records

DEMO-SET-201 has a USD 2.00 difference: the internal record is USD 100.00 and the sample statement is USD 98.00. DEMO-SET-202 matches at USD 240.00. The cause is not recorded; no adjustment has been posted.

Synthetic record comparison
ReferenceInternalStatementDifference
DEMO-SET-201USD 100.00USD 98.00USD 2.00
DEMO-SET-202USD 240.00USD 240.00USD 0.00

Demo evidence: DEMO-FIXTURE-RECON-01 · internal records + sample statement

Unresolved exceptionsSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Settlement operations

Which batches require attention?

Illustrative scopeDEMO-PORTFOLIO-001 · USD
Interpreted filtersSnapshot of two settlement batches

Two sample batches need review. BATCH-31 is awaiting confirmation with an expected date of 8 Sep; BATCH-32 has a recorded reconciliation exception and no expected date. Neither record establishes receipt of funds.

Synthetic batch review
BatchRecorded stateExpected dateOpen item
BATCH-31Awaiting confirmation8 Sep 2026Receipt unconfirmed
BATCH-32ExceptionNot suppliedReconciliation review

Demo evidence: DEMO-FIXTURE-BATCH-01

Support draft for reviewSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Merchant support

Help me explain this payment status.

Illustrative scopeDEMO-MERCHANT-001 · DEMO-TXN-109
Interpreted filtersRecorded state: Pending · USD 60.00

Draft: “Our current record for DEMO-TXN-109 is Pending. A final outcome has not been recorded in this snapshot. We will follow the approved inquiry process before advising you further.” This does not confirm payment success, a decline reason or a settlement date.

Demo evidence: DEMO-FIXTURE-SUPPORT-01 · recorded Pending state + DEMO-GUIDANCE-02

Authorized sample portfolioSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Relationship management

Summarize the merchants assigned to me.

Illustrative scopeDEMO-RELATIONSHIP-001 · three assigned merchants
Interpreted filtersAssignment list and onboarding records only

Three merchants are assigned in this fixture: two have a completed onboarding record and one is awaiting document review. These are onboarding states, not confirmation of live payment enablement. No commission or payment-volume source is included.

Synthetic assigned portfolio
MerchantOnboarding state
DEMO-MERCHANT-001Completed
DEMO-MERCHANT-002Completed
DEMO-MERCHANT-003Document review

Demo evidence: DEMO-FIXTURE-PORTFOLIO-01

Sourced period comparisonSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Management

How did payment performance change this week?

Illustrative scopeDEMO-MERCHANT-001 · recorded attempts only
Interpreted filters1–7 Sep vs 25–31 Aug 2026 · UTC

The sample shows 8 approved of 10 attempts this week (80%), versus 6 of 8 last week (75%): a 5 percentage-point difference. Attempt count increased by 2. This small synthetic fixture establishes no cause, trend forecast or real product performance.

Synthetic weekly comparison
PeriodApprovedAttemptsShare
1–7 Sep 202681080%
25–31 Aug 20266875%

Demo evidence: DEMO-FIXTURE-WEEK-01

Source unavailableSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Unavailable source

Which settlements need review in this sample portfolio?

Illustrative scopeDEMO-MERCHANT-001 · USD
Interpreted filters1–7 Sep 2026 · Settlements

The settlement source is unavailable in this synthetic scenario. I cannot calculate a review count or confirm a payout. An unavailable source is not zero transactions or zero settlements.

Demo evidence: DEMO-FIXTURE-SET-UNAVAILABLE

Insufficient informationSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Missing information

Summarize this partner’s permitted merchant activity.

Illustrative scopeDEMO-PARTNER-001 · assignment unverified
Interpreted filtersPartner portfolio · Selected period not supplied

This fixture has no verified partner assignment, selected period or commission ledger. I cannot summarize a portfolio or forecast earnings from the available information.

Demo evidence: DEMO-FIXTURE-PARTNER-INCOMPLETE

Draft for confirmationSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / Payment-link draft

Prepare a USD 250 payment-link request.

Illustrative scopeDEMO-MERCHANT-001 · USD
Interpreted filtersAmount USD 250.00 · Description and expiry unconfirmed

Proposed amount: USD 250.00. Please confirm the description, expiry, invoice or customer reference and permitted payment methods. A separately authorized workflow would handle creation and delivery. Draft only—nothing created or sent.

Demo evidence: DEMO-GUIDANCE-LINK-01

No matching sample recordsSample as of 8 Sep 2026, 09:00 UTC

SAMPLE QUESTION / No matching records

Find approved Mastercard transactions on 7 September 2026.

Illustrative scopeDEMO-MERCHANT-001 · USD
Interpreted filters7 Sep 2026 · Mastercard · Approved

No matching records appear in this limited synthetic transaction fixture. This says nothing about real merchant activity or an unavailable source.

Demo evidence: DEMO-FIXTURE-TXN-01

Illustrative product preview · Synthetic data · No live AI connection

02 / HOW IT IS INTENDED TO WORK

Connect your data.
Set the access rules.

These are the steps a Copilot integration would follow. Each data source, access rule and processing component must pass implementation and validation checks before use.

  1. 01

    Connect the right sources

    Agree which internal sources Copilot may use, then configure and test each connection. It does not connect to any database automatically.

  2. 02

    Apply your access rules

    Enforce customer identity, tenant, merchant, store and field permissions before retrieval.

  3. 03

    Ask a supported question

    Ask about the records and tasks the integration supports. The assistant must say when it does not have enough evidence.

  4. 04

    Review the evidence

    Check the sources and how current the data is before using an answer or report. Your team makes the decisions.

03 / FOR YOUR EXISTING PLATFORM

Copilot under
your own brand.

Copilot is designed to work under your brand, with MerchantPaisa or your existing payment platform. Connecting your systems requires agreed integrations with approved internal data sources.

The assistant foundation is implemented, but standalone deployment, live model responses and complete customer isolation have not yet been validated. These need to be verified for your environment.

AI Copilot for our existing platform

Your identity

Plan for your own branding, interface and administration. The proposed standalone model does not require MerchantPaisa as the host.

Your access model

Customer identity and access integration, with server-side role and field restrictions.

Your data boundary

Define where customer data is stored, how it is kept separate and which private connections are allowed.

NO OPENAI AND NO EXTERNAL APIs
Only approved internal services and private/local processing across model serving, retrieval, embeddings, reranking, tracing and supporting tools.

04 / PLANNED USE CASES

Where Copilot
could help.

These use cases describe the proposed assistance, not features ready for every deployment. Confirm availability with our team. Financial actions still require separate permissions and approval controls.

03

AI reconciliation

Review unmatched items against the available records and suggest possible matches.

Matching suggestions do not post adjustments or resolve financial exceptions automatically.

04

AI payment orchestration

Explain recorded routing factors and compare supported provider performance or cost in context.

Recommendations are advisory. The Acquirer retains production routing control and execution.

05

AI partner management

Summarize assigned partner or ISO portfolios, merchant activity and recorded commissions.

Earned values remain distinct from estimates. Partner scope is enforced by the management service.

06

AI merchant support

Help answer merchant questions using approved guidance and transaction records the user can access.

Escalate to a person where supported; do not invent bank decline reasons or promise an outcome.

07

Settlement tracking

Explain recorded settlement, payout and exception states with relevant source references.

Scheduled settlement is not confirmed receipt of funds.

08

Predictive reporting

Explore future forecasting for activity, cash flow or settlement trends when the required data and method are available.

Future scope. Forecasts need sufficient authorized history, a validated method, a horizon and visible uncertainty—not guaranteed outcomes.

05 / THE PRIVATE DEPLOYMENT REQUIREMENTS

Private by design.
Permissioned by default.

These are requirements for a private deployment, not controls demonstrated by this preview. The model, licence, hardware, hosting and security checks need to be agreed for your environment.

01

Grounded in your records

Answers about your business should come from records the user is allowed to access and approved reference material—not public web searches or guesses from a model.

02

A private processing path

NO OPENAI AND NO EXTERNAL APIs across the intended Copilot runtime, processing and tools. Only approved internal services and private/local model serving, retrieval, embeddings, reranking, tracing and supporting tools may be used. No external fallback.

03

No external AI training or tracing

Business prompts, retrieved records, documents and answers must not be exported to external AI training, tracing or model-improvement services.

04

Server-side permission gates

Identity, tenant ownership, partner relationships, merchant and store scope, and field restrictions must be checked before data reaches a model. Prompt wording is not access control.

05

Evidence before confidence

Answers must show their sources and how current the data is. If no records match, information is missing or a source is unavailable, the assistant must say so.

06

People retain control

Separately enabled actions require human confirmation and existing authorization and approval controls. A model is never the authority for payments, balances or settlement.

MERCHANTPAISA INTEGRATION EXAMPLE

Alongside the platform.
Outside the payment path.

MerchantPaisa is one integration example, not a mandatory host. In this flow the Management API owns identity and authorization; the Acquirer retains payment processing, routing execution and authoritative financial outcomes.

See how the platform connects
  1. 01
    Authenticated user

    Signed-in identity and defined role.

  2. 02
    Management API authorization and scope

    Server-side ownership, permission and field checks.

  3. 03
    Separate private Copilot

    Approved private/local runtime; not a payment engine.

  4. 04
    Approved, permission-filtered data tools

    Scoped evidence from approved internal services.

  5. 05
    Grounded answer with evidence

    Sources, freshness, limitations and human review.

A FEW IMPORTANT DETAILS

Common questions
about Copilot.

Is this preview connected to my business?

No. All ten examples use fixed, synthetic data. Choosing a team displays a prewritten answer already included in the page. It does not call an AI model, private service, business database or payment system.

Can we use Copilot with our existing platform?

That is the proposed standalone, white-label model. We would first need to agree your data connections, identity, administration, storage and customer isolation requirements. Independent deployment and live private-model inference are not yet validated.

Will prompts, records or tool calls leave the private boundary?

The product requirement is NO OPENAI AND NO EXTERNAL APIs across runtime, processing and tools. Model serving, retrieval, embeddings, reranking, tracing and supporting tools must use only approved internal services and private/local processing, without external fallback. This website does not implement or validate that runtime.

Can the Copilot change a payment or routing decision?

Not from this public preview. Payment-link output is only a draft; creation and delivery are separate authorized workflows. Routing and provider analysis is advisory. Separately enabled actions need the business application’s permissions, human confirmation and approval controls.

Are forecasts available or guaranteed?

Predictive reporting is future scope. It needs sufficient authorized history, a validated method, a stated horizon and visible uncertainty. A sample comparison is not a forecast or a promise of an outcome.

DISCUSS YOUR COPILOT REQUIREMENTS

What does your team
need to investigate?

Share the questions your team asks, the systems you use and the data you can make available. We’ll discuss what needs to be connected, configured and tested.

Discuss Private AI