NEWMindMap Digital has acquired Bluetide.co— deepening our data & agentic-AI stack.Read more →
Home · Customer Stories · African Tier-1 Bank
BFSI · Africa

Branch Automation at an African Tier-1 Bank — 56% Teller-Task Reduction Across 800 Branches

Helpdesk Automator + DocuMage + Multi-Agent Orchestrator collapsing in-branch transaction processing into a tablet-led, agent-assisted workflow.

56%
Teller-task time reduction
30w
Delivery duration
Hybrid
Deployment
5
Accelerators used
HybridAfrican Tier-1 Bank — 56% Teller-task time reduction
56%
Teller-task time reduction
800
Branches live
78%
Forms eliminated
4.3 → 4.8
Branch CSAT
In this storyBranch BankingBFSIHybridTablet-ledAfrica
01
The challenge

The challenge

The bank — a Tier-1 universal bank in a major African market with approximately 800 branches across four countries — was operating a branch network whose customer-experience metrics were trending downward despite ongoing investment. Average in-branch transaction time was 17 minutes, average wait time at peak was 38 minutes, and the bank's customer-experience research consistently identified the branch experience as the lowest-rated touchpoint in the customer journey.

The structural issue was the teller workflow. A typical teller's day was dominated by repetitive transaction processing — cash deposits, cash withdrawals, account-balance enquiries, cheque encashment, foreign-exchange transactions, money-transfer processing — each requiring the teller to navigate the bank's core-banking-system terminal, complete a paper form (later filed and scanned by a back-office team), validate the customer's identity, and process the transaction. The teller had little time for the relationship-banking and advisory work that the bank's leadership wanted the branch to actually do.

The bank's CIO had set a target of redirecting 50% of teller capacity from transaction processing to advisory work, without reducing the branch network's customer-handling capacity. Previous transformation attempts — including a self-service ATM expansion and a contact-centre re-platform — had partially addressed the problem but had not transformed the in-branch experience.

02
The approach

The approach

MindMap deployed an in-branch platform composed of Helpdesk Automator (Hd) as the transaction-orchestration layer, DocuMage as the document-intelligence layer (eliminating most paper forms), Multi-Agent Orchestrator (Mo) for the multi-system transaction processing, FaceMatch (Fm) for customer-identity validation and Policy Q&A Agent (Pq) as the teller's policy-and-procedure assistant. The platform was packaged into a tablet-led teller workstation that replaced the previous core-banking-terminal-and-paper-form workflow.

Phase one was the transaction-flow rebuild. We worked with the bank's branch-operations team to redesign the top 20 transaction flows around the new platform. Each flow was designed to be tablet-led — the teller and the customer interact across the tablet, with the customer providing inputs (signature, biometric confirmation, transaction confirmation) directly through the tablet rather than through paper.

Phase two was the document elimination. Approximately 78% of the paper forms that the previous workflow generated were eliminated by the new platform — replaced by tablet-captured inputs with customer biometric confirmation. The remaining 22% — primarily forms required by regulation in physical form — are completed as templated tablet outputs that are printed once for customer retention rather than completed manually.

Phase three was the teller redeployment. With the transaction time per teller reduced, the bank's branch operations team was able to redeploy teller capacity into the new branch role design: relationship bankers (focused on customer relationship management), product advisors (focused on cross-sell and product education), and service ambassadors (focused on branch-floor customer guidance). The transaction-processing role itself was retained but reduced in headcount per branch.

Accelerators in this engagement

The pre-built building blocks

Rather than commission a ground-up build, the engagement leaned on MindMap's pre-built accelerator library — production-tested components that compress what would otherwise be a six-to-nine-month build into weeks.

Hd

Helpdesk Automator

Transaction orchestration and teller-workstation logic

Dm

DocuMage

Customer-ID and form-data extraction at the tablet

Mo

Multi-Agent Orchestrator

Cross-system transaction parallelisation

Fm

FaceMatch

Customer-identity validation at the tablet

Pq

Policy Q&A Agent

Teller policy-and-procedure conversational assistant

03
The architecture

The architecture

The platform runs as a hybrid: the central orchestration and audit layer lives in the bank's primary data centre, while branch-specific processing (transaction queuing, document capture, customer-tablet interaction) runs on a branch-local appliance that maintains a sync to the central platform.

The teller tablet runs a custom Android application that wraps the platform's APIs and provides the unified transaction workflow. The branch-local appliance handles the local-network performance optimisation — most transaction processing completes inside the branch's local network rather than requiring a round-trip to the central data centre — and provides a degraded-mode capability when the branch's connection to the central network is impaired.

DocuMage's document-intelligence runs partly on-tablet (for the document-capture validation that needs to be instant — checking that the customer's ID document is in frame, in focus and not occluded) and partly on the branch-local appliance (for the full extraction that drives the transaction processing). The on-tablet component is a Phi-3.5-vision-based small model; the appliance-based component is a fine-tuned LLM optimised for the bank's specific document types.

Multi-Agent Orchestrator coordinates the cross-system transaction processing — a typical money-transfer transaction involves the core banking platform, the SWIFT or regional payment-rail gateway, the fraud-and-screening engine, the audit-and-compliance system and the customer-notification system. The orchestrator parallelises where possible and serialises where dependencies require, with built-in retry-and-rollback handling for partial failures.

Integration with the core banking platform (Flexcube in the home market, Finacle in two of the other markets) is via the existing core-banking APIs, with the platform adding caching and parallelisation but not replacing the underlying core. The bank did not need to replace the core banking platform.

The outcomes

The numbers behind the story

56%
Teller-task time reduction
800
Branches live
78%
Forms eliminated
4.3 → 4.8
Branch CSAT

Average teller-task time has dropped by 56%. The reduction is driven by three factors: the elimination of paper-form completion (which previously absorbed 30-40% of transaction time), the parallelisation of cross-system transaction processing (which eliminated sequential waiting), and the on-tablet customer self-input (which moved input time from the teller to the customer in parallel with other steps).

Average customer wait time at peak has dropped from 38 minutes to 11 minutes, despite no increase in teller headcount. The bank's branch capacity to handle transaction volume has roughly doubled at the same staffing level.

Approximately 78% of the paper forms the previous workflow generated have been eliminated, with corresponding reductions in back-office form-scanning and form-storage costs. The bank's branch document-storage spend has dropped meaningfully.

The teller-to-relationship-banker redeployment has happened at scale. The bank's branch network now operates with roughly half the previous teller headcount and a corresponding increase in relationship-banker and product-advisor capacity. The bank's per-branch revenue contribution has increased materially, driven by the cross-sell uplift the relationship-banker role has produced.

Branch CSAT has risen from 4.3 to 4.8 on the bank's five-point scale. The improvement is roughly evenly attributable to the reduced wait times and to the improved interaction quality in the branch (more time per customer, more advisor presence on the branch floor).

The transformation in our branch experience has been the largest single customer-experience improvement in the bank's recent history. Our tellers are now doing relationship banking and advisory work, not paper forms. Customer wait times have collapsed. And we did not have to replace the core banking platform to get there.
Group Head of Retail Banking· African Tier-1 Bank
04
Why MindMap was chosen

Why MindMap was chosen

The bank had previously evaluated a major global banking-platform vendor's branch-automation product and concluded that it required wholesale replacement of the bank's core banking platform — a multi-year programme the CIO was not prepared to undertake. Two regional system integrators had proposed branch-automation builds in the $20-30m range with 18-24 month timelines.

MindMap's accelerator-composition approach — bringing pre-built accelerators together into a unified branch platform without replacing the core banking platform — was a structurally different proposition. We could demonstrate the pattern in production at another African bank, walk the bank's branch-operations leadership through the redesigned workflows, and commit to a 22-week first-cohort rollout with measurable acceptance criteria.

The hybrid on-premises-plus-branch-local architecture, with degraded-mode capability for branches with intermittent connectivity, was a unique commercial position. The bank's branch network includes a meaningful number of branches in rural and peri-urban locations with sub-optimal connectivity, and the architecture's tolerance of that reality was a critical bid commitment.

Want an outcome like this?

Start with a 2-week AI Readiness Sprint. We deliver a prioritised use-case backlog and business case grounded in what's actually buildable with our accelerator library.

Book a walkthrough →Explore BFSI
Talk to the product team