Work
Five platform modernisations in twenty-six years, the pattern I keep noticing, and what I'm doing now in outline. The detail of the current work belongs to my employer, which is how it should be.
The pattern
Every one of these started the same way. The reports were slow, the numbers got argued about in meetings, and somebody had built their own version in Excel, which was the one people actually used. The plan on the table assumed the replacement would take years. My job was to make it take months. Then build the team that runs it. Then make myself unnecessary. Six, twelve, thirty-six.
Here's what changed since I started doing this: the technology is no longer the hard part. A stack that needed a large programme in 2021 costs a fraction of that today, and AI-assisted engineering has shrunk the build time again. What hasn't changed is people. A sponsor who owns a P&L, a catalyst with a date on it, and a leadership team willing to be measured. If those three exist, I move fast. If they don't, no tool will save the project, and I'll say so early (not always popular).
The record
2024 – present · Patties Food Group · Data & AI Lead
Azure-based stack → Snowflake, dbt, Fivetran and ThoughtSpot. Migration complete in under twelve months.
CI/CD for data engineering in dbt and GitHub replaced hand-run ETL. Time-to-market for new data products went from years to months. Self-service analytics in ThoughtSpot, with business teams upskilled to answer their own questions. AI and LLMs embedded in analytics operations for code generation, documentation and repetitive work. Executive buy-in for an AI-driven analytics strategy.
2021 – 2024 · Carlisle Homes · BI Analytics Manager
Legacy BI → cloud analytics stack, with reliability and access as the measures.
Reporting latency down, data pipelines hardened. Marketing moved to self-service and stopped queueing for the technical team. Data strategy aligned with executive objectives rather than with the tool roadmap.
2018 – 2021 · Patties Foods · BI Analytics Lead
SSIS and Cognos 10 → ThoughtSpot, Cognos Analytics 11, Planning Analytics (TM1) and an Azure data platform with TimeXtender.
Governance standards established so the numbers could be trusted. Built and ran an offshore BI team paired with onshore consultants for an end-to-end service. Self-service training that cut IT dependency. The Azure implementation was published as a Microsoft customer success story.
2017 – 2018 · GPC Asia Pacific · BI Analytics Solutions Lead
Oracle Discoverer → modern analytics platform, a multi-million-dollar programme.
Wrote the business case and presented it to the board. Secured funding. Ran an RFQ across four vendors, selected and onboarded the implementation partner, and managed the transition and stakeholder adoption.
2013 – 2017 · Simplot Australia · BI Technical Lead
Enterprise Cognos estate made faster and cheaper to run.
Data-culture initiatives that raised analytics awareness across teams, on a platform the business already owned.
2005 – 2012 · BlueScope Steel · Business Intelligence Manager, Asia
Built and ran BI as a shared service across BlueScope's Asian businesses with a multinational team.
Data warehousing, Cognos, dashboarding and scorecarding, outsourced development. Country IT Manager for Vietnam along the way. Internal awards. Before that: ERP functional analyst for manufacturing.
2000 – 2005 · Where it started
Business analyst on a core life-insurance system at Bao Minh CMG (now Dai-ichi Life), building reporting in SQL Server and Crystal Reports. Before that, a year in coffee export documentation. Twenty-six years, one thread: making the numbers usable by the people who have to act on them.
Current, in outline
There is serious AI work in my day job. It belongs to Patties, so what follows is the shape, not the numbers. Four exhibits. One theme, which regular readers will recognise: raise the floor before you chase the ceiling.
- AdoptionInternal apps that business people actually open in the morning. Adoption is the metric. A demo nobody uses is a slide.
- Developer productivityAI coding agents (Snowflake Cortex Code) in the data team's daily workflow. I measure the change in the team's own numbers, not the vendor's.
- Knowledge-work automationChecking a brief written in Word against the artwork that came back as a PDF. Tedious, error-prone, and it turns out a machine is good at it. Delivered with a partner. Not a data problem. Most of the AI value in a company like ours isn't.
- PlanningAI in integrated business planning, which in most FMCG companies still runs on spreadsheets and tools from another decade. Early days. If you ask me where the biggest untapped opportunity in this industry is, it's here.
How I work
- Player-coach. I run the team and the budget, and I still read the SQL. Usually before the meeting, so I know when the slide is wrong.
- Measured or it didn't happen. Baseline, unit, date. If I can't put a number on it, I won't claim it.
- Months, not years. The stack stopped being the constraint a few years ago. Readiness is the constraint now. I'd rather find the catalyst than fight the culture. I've tried it the other way. It took years and I don't recommend it.
- Diagnosis before prescription. I'll tell you what isn't worth doing. AI for the sake of AI is at the top of that list.
Credentials
Executive MBA, RMIT · ThoughtSpot Certified Data Expert · IBM Certified Solutions Expert, Cognos BI · FLMI, Life Management Institute · Bachelor of Business Administration, University of Foreign Trade, Ho Chi Minh City
Want the story behind any line above? linkedin@congtam.net. I reply faster than LinkedIn does. What I'd say yes to is on Now.