Haleon's Bangalore hub cut packaging tender time from seven weeks to five minutes
Fresh from its GSK demerger, Haleon pointed AI, RPA and NLP at three supply-chain frictions: SKU guessing, slow packaging changes and slow answers.
What was changed
When Haleon emerged from GlaxoSmithKline in July 2022 as an independent consumer-health company — Eno, Iodex, Otrivin, Crocin, Sensodyne, Centrum — it inherited operational frictions that had never been its job to fix. Chief Digital and Technology Officer Amy Landucci's team framed three: sales reps guessing which of many SKU variants to sell into which store, packaging changes taking weeks to reach market, and consumer queries taking up to a ten-day service-level window to resolve.
The first fix targeted the combinatorial mess of field sales. 'A sales rep visiting 300 stores with 10 products in 10 SKUs each would have to make 3 lakh combinations to figure out which product would sell fastest,' Landucci said. Predictive analytics now clusters stores by location, market size, socio-economic status and footfall, so reps get a recommended assortment instead of relying on intuition — protecting margins while avoiding inventory pile-ups at plant or store.
The second and third fixes ran through Haleon's Global Capability Centre in Bangalore. Packaging is regulated country by country — a headache-claim fix can't ship until the pack says so — so the team built a bot using machine learning and OCR to generate packaging proposals across all product variants, cutting new-tender creation from seven weeks to about five minutes on average. For consumer queries, an NLP engine reads incoming questions and proposes responses in multiple languages from the existing database, deployed alongside manual resolution, with a human team as 'a check and balance'.
Why it worked
A newly demerged company can't afford bespoke process everywhere: Haleon chose the frictions with hard cycle-time metrics — tenders, queries, SKU matching.
Simplification came before automation: the team documented and simplified each process, then matched technology to the use case with tech-business interlocks.
Keeping humans in the loop during NLP rollout answered the real blocker — internal confidence — by using the manual team as a check and balance.
The Bangalore GCC gave a 2022-vintage consumer-health startup senior-scale engineering capacity without rebuilding it from scratch.
What can be applied
Post-spinoff companies inherit enterprise problems without enterprise plumbing: pick the three frictions with measurable cycle times and automate those first.
Aftermath
By early 2023 Landucci reported the packaging bot averaging five-minute tender creation and the NLP assistant processing live queries under human review, with faster responses in multiple languages. The company framed the work as raising technology awareness and education across business teams 'to create value' — the groundwork phase of a longer digital rebuild, per ETCIO's case study.