A bottleneck in the onboarding pipeline is not a single delayed application — it is a step that is consistently consuming more than its allocated time across multiple applications simultaneously. When the credit assessment step is running behind for 8 of the 12 MSME applications in the pipeline, the problem is not 8 individual delays — it is a capacity or process issue at the credit assessment step that will affect every application that arrives until it is resolved. The Onboarding SLA Agent AI detects bottlenecks by looking at the aggregate step performance across all active applications, not just at individual application delays — identifying which steps are systematically slow, which teams or individuals are the constraint, and what the downstream TAT impact will be if the bottleneck is not resolved.
A bottleneck in the onboarding pipeline is not a single delayed application — it is a step that is consistently consuming more than its allocated time across multiple applications simultaneously. When the credit assessment step is running behind for 8 of the 12 MSME applications in the pipeline, the problem is not 8 individual delays — it is a capacity or process issue at the credit assessment step that will affect every application that arrives until it is resolved. The Onboarding SLA Agent AI detects bottlenecks by looking at the aggregate step performance across all active applications, not just at individual application delays — identifying which steps are systematically slow, which teams or individuals are the constraint, and what the downstream TAT impact will be if the bottleneck is not resolved.
Individual application delay vs systemic bottleneck: why the distinction drives a different response
An individual application delay has an individual cause: a specific document missing, a specific valuer who did not respond, a specific borrower who is travelling. The response is targeted: follow up with the specific document, call the specific valuer, wait for the specific borrower. A systemic bottleneck has an operational cause: a credit analyst who is overloaded because two senior analysts are on leave, a legal team that is working through a backlog from the previous week, a QC reviewer who handles all LAP files but has been assigned to a training programme this week. The response is operational: reallocate workload, defer lower-priority tasks, request backup capacity. Misidentifying a systemic bottleneck as a set of individual delays leads to 8 individual follow-up calls that resolve nothing — because the credit analyst who is behind on 8 files needs relief, not 8 reminder emails.
"When 8 of 12 MSME applications are stuck at the same step, the problem is the step — not the 8 applications. Follow-up calls to each RM will not fix a staffing gap in the credit team."
The bottleneck analysis: November 14, 2025 · 14:30
Bottleneck Analysis — November 14, 2025 · 14:30 · All Active Applications
3 active bottlenecks identified · All linked to staffing or external capacity · Not individual application issues
3Active bottlenecks
12Applications affected
+1.4 daysProjected TAT impact
3Resolutions initiated today
01
CREDIT ASSESSMENT STEP · MSME PRODUCT
Credit analyst queue: 8 MSME applications at or beyond SLA — 2 analysts on leave
Of 12 active MSME applications, 8 are in the credit assessment step. Of those 8, 5 have consumed over 80% of their 4-hour SLA and 3 have exceeded it. Analysis shows: the credit team has 3 MSME analysts; Analyst Priya is on sick leave (Nov 14–17) and Analyst Karthik is at an outstation training programme (Nov 13–15). Analyst Ramesh is handling all 8 files — with a 4-hour SLA per file, this is physically impossible to clear on time without additional capacity.
→ Root cause: staffing gap (2 of 3 analysts unavailable) · Individual follow-up will not resolve this · Capacity solution required
Resolution initiated 14:00: Credit Head notified of capacity gap. Salaried product analyst (Meena) reassigned to MSME assessment for rest of day. Karthik recalled from outstation training Nov 15 morning. All 8 MSME files re-prioritised; Ramesh handles 4, Meena handles 4. Projected clearance: Nov 15 by noon. 3 borrowers proactively updated with revised timelines.
02
PROPERTY VALUATION STEP · HOME LOAN + LAP
Empanelled valuer capacity strained — 3 reports overdue, 2 more at 80% SLA
The institution has 4 empanelled valuers for Bengaluru. Two (Girish and Mahesh) are handling 11 active valuations between them — 7 for Girish, 4 for Mahesh. Girish has been unreachable since Nov 12 (LA-2025-18808 is already breached). Two more of Girish's assigned properties (for LA-2025-18832 and LA-2025-18847) are approaching their valuation SLA deadline. The valuation panel capacity is the structural constraint — not individual application follow-up.
→ Root cause: valuer capacity / single valuer overloaded and unresponsive · 5 applications at risk if not reassigned today
Resolution in progress: Girish's 3 pending files (18808, 18832, 18847) reassigned to valuers on the Mysuru panel (different geography but within RBI guidelines for concurrent assignment). Valuation Head notified to consider expanding Bengaluru panel. LA-2025-18808 borrower updated with revised disbursement estimate of Nov 17.
03
BORROWER DOCUMENT RESPONSE · ALL PRODUCTS
Document pending from 3 borrowers for 2+ working days — onboarding stalled at document collection step
Three applications (LA-2025-18798, LA-2025-18812, LA-2025-18829) have been waiting for borrower document submissions for 2 or more working days. The initial WhatsApp request was sent; no follow-up was sent. This is not a team capacity issue — it is a borrower communication gap. The document collection step's 3-day SLA is at 80–95% for these three applications. Without the documents, the pipeline cannot progress.
→ Root cause: insufficient borrower follow-up cadence · No escalation sequence was triggered after the initial document request
Resolution: Automated follow-up WhatsApp sent to all 3 borrowers at 14:30 — this is the T+2 reminder with urgency framing ("your documents are needed to keep your loan on schedule"). RM of each application notified to call borrower if no document upload within 4 hours. If no response by Nov 15, the 3 applications move to "borrower delay" status and TAT clock pauses pending response.
How bottlenecks are distinguished from individual delays
The SLA Agent AI applies a pattern detection algorithm over the step performance data of all active applications. A step is classified as a bottleneck when three conditions are met: at least 25% of applications currently at that step are at or beyond SLA; the step's average time-consumed across all active files is greater than 60% of the SLA; and the same step showed elevated time consumption in the previous 5 working days (ruling out a one-off spike). When all three conditions are met, the step is flagged as a bottleneck and the resolution workflow is escalated to the operational owner of that step rather than to the individual RM for each affected application.
The distinction between Bottleneck 1 (credit analyst staffing gap — operational) and Bottleneck 3 (borrower document response — communication) is important because they have structurally different resolutions. Bottleneck 1 is resolved by reallocating human capacity. Bottleneck 3 is resolved by escalating the borrower communication. Treating Bottleneck 3 as an operational problem (reallocating staff to "follow up on documents") would produce the wrong intervention — there is nothing an additional staff member can do that an automated follow-up WhatsApp cannot do more effectively.
3Active bottlenecks — credit staffing, valuer capacity, borrower response · Each requires a different resolution type · All detected simultaneously
12Applications affected by systemic bottlenecks — vs 3 individual delayed applications · The distinction changes the response entirely
−1.4 daysProjected TAT improvement if all 3 bottlenecks resolved today — 12 applications move forward · Borrower experience recovered
25%Bottleneck detection threshold — step flagged as systemic when 25%+ of files at that step are at or beyond SLA · Pattern, not exception
Detecting that 8 MSME applications are stuck in credit assessment and calling 8 RMs is the wrong response. Detecting that 2 of 3 credit analysts are unavailable and reassigning a cross-trained analyst is the right response — and only bottleneck detection makes the distinction visible.
Without bottleneck detection, the SLA Agent AI would generate 8 individual credit assessment alerts — one per application. The operations team would attempt 8 individual follow-ups. Analyst Ramesh would receive 8 reminder notifications while trying to complete 8 credit assessments with a 4-hour SLA each. The follow-ups would not help. The bottleneck would not be resolved. The 8 applications would breach their SLAs. The borrowers would be updated with delays. The Credit Head would learn about the problem on Day 10, not Day 4. Bottleneck detection identifies the 2-analyst capacity gap immediately and routes the resolution to the Credit Head with a specific recommendation. The Onboarding SLA Agent AI's bottleneck detection is the difference between 8 ineffective individual interventions and 1 effective operational intervention that resolves the problem before 8 borrowers experience a delay.
