Customer evidenceField study · AI use-case discovery
    Suez Canal BankSCB

    How SCB mapped 470 workflows and found high-value AI use cases

    Seventy-five employees across 16 departments described their daily work in plain language. AUGMTD structured it into 470 workflows and identified 58 detailed workflows as candidates for AI-supported work.

    Workflows mapped
    470
    Use-case candidates
    58
    Employees
    75
    Departments
    16
    01 · WHAT HAPPENED

    People described the work. AUGMTD surfaced the AI use cases.

    During a structured mapping exercise using AUGMTD, 75 employees described their daily work in plain language. They did not build automations or learn process-mapping software. AUGMTD structured those descriptions into 470 workflows covering about 717 hours of recorded process time.

    Seventy-three workflows were refined to individual step level. Fifty-eight contained a majority of work AUGMTD can support, representing 533 recorded hours. Together, they form an evidence-based pipeline of AI use-case candidates: work that could be prepared, coordinated, monitored, or completed through controlled workflows.

    75people described their work
    58AI use-case candidates identified
    01 · DESCRIBE

    Review submissions, gather approvals, update the client, and track the committee response.

    Plain language · no process notation
    AUGMTDidentifies activities, roles, tools, and time
    02 · IDENTIFY USE CASES
    Prepare submissionsDrafting
    Coordinate approvalsHandoffs
    Track responsesFollow-through
    Plain-language descriptions became structured workflows. Detailed workflows could then be assessed as AI use cases and turned into controlled automation candidates.
    02 · ONE REAL PROCESS

    One credit proposal took 26 steps and eight people.

    A Corporate Team Leader mapped how a credit proposal actually came together. The process took 34 hours. Eight hours went to reviews, meetings, follow-ups, debriefs, alignment, notifications, and checking approval status.

    Nobody in that chain was necessarily inefficient. Context had to move between analysts, compliance, the credit committee, and the client, and someone had to remember each handoff.

    34htotal effort
    Core proposal work26 hours76%
    Coordination and chasing8 hours24%
    26 steps8 responsible roles
    87%of documented steps touched two or more tools
    42%of every tool mention was a communication surface

    The coordination work was not hidden in one core system. It was distributed across conversations, meetings, and documents.

    03 · WHEN THE MAPS MET

    Forty-five recurring patterns revealed shared AI opportunities.

    Once all 470 workflows sat side by side, 185 entries grouped into 45 recurring process patterns. Seventeen patterns crossed departmental boundaries. These clusters showed where one well-designed AI workflow could address work repeated across several teams.

    Email management8 departments16
    Risk reporting10
    Control testing9
    Risk assessment9
    Appetite monitoring9
    Stress testing8
    Performance reviews7
    Incident response6
    04 · HOW IT WORKED

    AI use-case discovery started with a paragraph.

    Employees included analysts, relationship managers, HR officers, and compliance staff. They began by writing short descriptions of their working day. AUGMTD proposed a structured set of workflows from those descriptions, then employees reviewed the result and chose what to refine. That structure made manual steps, repetitive coordination, and suitable automation boundaries visible without asking employees to design an AI system themselves.

    100%
    of 445 steps carried a time estimate
    97%
    named the tools involved
    96%
    named the responsible role
    76%
    of employees created five or more workflows
    05 · THE CONTROL BOUNDARY

    The map revealed why handoffs still need accountable people.

    Eighty-one percent of mapped workflows involved two or more responsible roles, averaging 2.8 roles each across 83 distinct roles. A credit committee decision, regulatory review, or final compliance check cannot become an anonymous software action.

    Each use case therefore needs a clear operating boundary. AI can carry permitted context between handoffs, notice what is outstanding, prepare the next move, and complete routine steps where the configured controls allow it. Accountable people remain attached to consequential reviews and decisions.

    1AnalystBuilds proposal evidence
    2Team leadReviews commercial position
    3ComplianceChecks policy and controls
    4Credit committeeMakes the accountable decision
    MEASURED IN THE MAPPING EXERCISE
    • 470 workflows from 75 employees
    • 717 recorded hours of process time
    • 445 steps with structured attributes
    • 45 recurring process patterns
    IDENTIFIED AI USE-CASE PIPELINE
    • 58 of 73 detailed workflows were majority in scope
    • 533 recorded hours represented by those candidates
    • Candidate work includes preparation, coordination, monitoring, and routine execution
    • Candidates were evaluated from documented work at step level
    • Each candidate retained its roles, tools, and recorded effort
    06 · THE AI USE-CASE PIPELINE

    A map of work becomes a practical automation backlog.

    SCB gained a structured view of 470 workflows, 45 recurring patterns, and 58 detailed candidates for AI-supported work. The next step for each candidate is concrete: define the desired outcome, select the routine steps AI can handle, attach the right approvals, and measure what happens when the workflow runs. Roles and processes will continue to change, so the pipeline must evolve with the work it represents.

    December 2025 mapping exercise · Test records excluded · Delivered in partnership with Zero to 100
    WORK MADE VISIBLE470structured workflows
    AUTOMATION PIPELINE58 candidatesready for workflow design and validation
    FIND YOUR AI USE CASES

    Turn real work into an evidence-based automation pipeline.

    Discuss a pilot