Back

McKinsey & Company — ReWoT

Product Lead

Chennai, India20192021

Owned product end-to-end for an internal workflow platform: identified the problem, designed the UX, built the platform, and managed the 2-month change management rollout across 60+ team members. Automated client request intake, replacing a 3 to 4 hour daily manual process with ML-based classification and real-time team availability. Full adoption in 2 weeks, no mandate.

2–3hdaily effort eliminated
60+users fully adopted
2 wksto full adoption

the problem

A Data Analytics team received 10 to 15 client requests daily through a shared inbox. Every morning, one person spent 3 to 4 hours manually opening emails, categorising each request across three service lines, checking an Excel tracker for availability, forwarding to the right team, and updating the file. When they were on leave, requests disappeared. Clients escalated.

my role

I owned the product end to end: identifying the problem, designing the user experience, building the platform, and managing the 2-month change management rollout across 60+ team members including analysts, leads, and directors. A data scientist owned the ML model for service line classification.

before and after

Before
  • 3 to 4 hours of manual intake every morning
  • 3 to 5 requests missed or dropped weekly
  • One Excel file as single point of failure
  • No visibility into team capacity
  • No request traceability
After
  • 5 minutes of intake, fully automated
  • Zero missed requests
  • ML-based classification across all service lines
  • Real-time team availability dashboard
  • Every request searchable by unique ID

key decision: build vs buy

We built instead of buying an existing SaaS tool.

Three reasons: cost was lower to build, we needed full feature control, and most critically McKinsey client data could not touch a third-party system. Privacy was non-negotiable. Building internally gave us data sovereignty, full feature control, and lower total cost.

four features

FEATURE 01

Automated email intake

Pulls new client requests from the shared inbox every morning automatically. No manual checking, no missed emails, no dependency on one person.

FEATURE 02

ML-based classification

Each request is automatically classified into one of three service lines: Business Modelling, Data Engineering, or Predictive Analytics, using a trained ML model.

FEATURE 03

One-click assignment and notification

Assigning a request to a team member triggers an automatic email notification. Every request gets a unique ID. Leadership can search any request and see its status instantly.

FEATURE 04

Real-time availability dashboard

A dedicated page showing team capacity, current workload, and leave dates. Managers can see at a glance who is available to take new work before assigning anything.

how it works

Email arrives shared inbox Auto pull fields to Excel ML classifies service line Manager assigns one click Tracked unique ID

what I learned

The hardest part was not the build. It was the 2-month rollout against internal resistance. Directors who had managed via email for years did not want a new system. What I would do differently: involve the highest-resistance users in testing before launch. Their objections would have shaped the product and converted them into advocates.

stack

JavaScriptVB.NETSQL ServerInternal hosting
GitHub
LinkedIn