Operations, Data & AI Transformation Portfolio

Operational transformation

Redesigning how the operation actually runs

Case study 01

Price Change Request Workflow Transformation

Context: the merchant price-change process was inefficient, error-prone and time-consuming. Account managers downloaded live pricing into Excel and reformatted files, merchants often returned data in the wrong structure, and merchant operations and pricing teams repeatedly corrected it. The external development agency reported that 7 of 8 billed hours were spent reformatting before scripts could run.

Approach: mapped stakeholder pain across merchants, account managers, merchant operations, director and product stakeholders and the external dev agency; redesigned the workflow in phases; standardised the required format; proposed a HubSpot form that outputs the correct structure; planned later pre-population from live pricing data; added validation at entry; and defined the future-state flow and rollout steps.

Projected impact

Up to 90% reduction in external invoiced reformatting hours (projected)
Less internal rework across operations and pricing
Fewer errors and faster processing
Better merchant experience
Staff time redirected to value-added work

Current state

01Account manager downloads live pricing into Excel
02File is manually reformatted before it can be sent
03Merchant returns data in the wrong structure
04Merchant operations and pricing teams correct it repeatedly
05External dev agency reformats before scripts can run — 7 of 8 billed hours
06Errors, rework and slow turnaround for the merchant

Future state

01Merchant submits through a HubSpot form in the required structure
02Validation applied at the point of entry
03Later phase: fields pre-populated from live pricing data
04Standardised file passes straight to processing
05Exceptions routed to a named owner, not the whole team
06Faster turnaround, fewer errors, time redirected to value-added work

Case study 02

Customer Service CSAT Operating Model

Context: customer service only sent external review requests selectively, so there was no reliable internal view of satisfaction, or of team and ticket quality.

Approach: proposed automatic CSAT collection on closed Zoho Desk tickets and designed the reporting dimensions — agent, team, customer, gym or location, membership type and channel — specifying fields, filters, grouping, dashboards and manager notifications for poor ratings.

Operational outcome

A measurable service-quality framework supporting coaching, trend analysis, customer segmentation, team performance and continuous service improvement.

Case study 03

End-to-End Operating Model & AI-Governed Process Design

End-to-end operating models covering lead intake, fulfilment, escalation, retention and reporting

Governance frameworks for record ownership, statuses, lifecycle rules, escalation paths and automation triggers

AI-enabled workflows with adoption controls

Documentation, training and runbooks

A clinic operating model standardising consultation → recommendation → follow-up → reorder → rebooking

Human sign-off and audit trails at every material decision point

Reliable data underneath the reporting

Restore trust in revenue reporting

Audited Databricks and CRM data, identified the root cause of a cross-system pricing and workflow inconsistency, remapped affected fields and designed a validation layer.

Projected impact: Up to 90% reduction in external data-reformatting cost.

AI-Ready Data Platform

Moving fragmented CRM and operational data into a governed Lakehouse pattern, transforming it and exposing clean data for analytics and AI workflows.

Stack: SQL, Microsoft Fabric, PySpark, Databricks, pipelines, Power BI, Git/GitHub. Planned build.

Automation that stays governed as it scales

Automation governance architecture

Central automation register, business-event conflict review, data-quality problem log, controlled remediation and historical KPI snapshots across the automation estate.

Outcome: Visibility across automations, explicit ownership and risk, auditable data-quality evidence and safer change control.

AI Integration Service

A Python API service that accepts structured business data, calls an LLM, validates the output and returns reliable JSON.

Stack: Python, FastAPI, REST APIs, environment variables, error handling, testing, Git/GitHub. Build in progress.

Automation Governance Assistant

Turning existing automation-governance architecture into a searchable AI assistant for workflow ownership, conflicts, risk and change control.

Stack: Python, RAG, vector search, LLM APIs, structured outputs, governance data. Planned build.

Create more qualified pipeline without adding SDR headcount

I started with the commercial constraint — an 85% new-business growth target, 2.4x pipeline coverage and 7% SDR-to-SQL conversion — then designed a prioritisation workflow to make existing SDR capacity more productive.

What I designed: Salesforce firmographics + HubSpot intent + Gong similarity + external triggers + Clay scoring. Deterministic logic controls priority and routing; AI produces the rep-facing why-this-account, why-now, outreach angle and next action.

KPIs

SDR-to-SQL conversion
SQLs and qualified pipeline per SDR
Research time per account
Meeting and conversion rate by priority tier

A 7% → 10% SDR-to-SQL improvement models to approximately 43% more SQLs from the same SDR input — 30 additional SQLs per 1,000 prospects worked, and an illustrative ~£288k additional new ARR.

Modelled and illustrative — not a realised result.

Give AEs more selling time

Designed a Gong-to-Salesforce workflow that extracts next steps, stakeholders, objections and MEDDPICC evidence, with material CRM updates kept human-in-the-loop.

Measure: CRM admin time, call-to-update time, MEDDPICC completeness, stale opportunities, selling time and forecast accuracy.

Detect deal risk earlier

Designed a risk layer across opportunity, engagement and activity signals to surface declining engagement, overdue next steps and stalled buying processes before forecast calls.

Measure: Late-stage slippage, win rate on flagged deals, stage conversion, forecast accuracy and win rate after manager intervention.

Protect recurring revenue earlier

Designed a customer-health model combining adoption, engagement, support and renewal data so CSMs can act on deterioration before renewal pressure peaks.

Measure: Churn ARR, downsell ARR, GRR, NRR, risk-detection lead time, recovery rate and renewal rate of flagged accounts.

Selected Work & Outcomes