---
title: A women's health AI agent, live in 12 weeks
---
[](https://pages.melio.ai/?hsLang=en)
- [Home](https://pages.melio.ai/)
- [Services](https://pages.melio.ai/services)
- [Consulting](https://pages.melio.ai/consulting)
- [Retainer](https://pages.melio.ai/retainer)
- [AI As A Service](https://pages.melio.ai/ai-as-service)
- [Fluid Assessment](https://pages.melio.ai/fluid-assessment)
- [Workshops](https://pages.melio.ai/workshops)
- [Products](https://pages.melio.ai/products)
- [Kaidex](https://pages.melio.ai/kaidex)
- [RubrlQ](https://pages.melio.ai/rubrliq)
- [About](https://pages.melio.ai/about)
- [Case Studies](https://pages.melio.ai/case-studies)
- [Insights](https://pages.melio.ai/insights)
- [Contact](https://pages.melio.ai/contact)
[Home](https://pages.melio.ai/?hsLang=en) › [Case Studies](https://pages.melio.ai/case-studies?hsLang=en) › Case Studies
Healthcare
# A women's health AI agent, live in 12 weeks
A chat-based agent trained on hormone and mental-health data connects in-lab test results to personalised, real-time guidance for FemmFlo's users.
### Case Study
**Industry** Banking
**Company Size** 1,000+ employees
**Project Length** 6 months
**Team Size** 4.5 FTE
### Problem Space
Personalisation
Generative AI
Campaign Automation
Customer Data
### Technology Used
Amazon Bedrock AWS Python React FastAPI Kubernetes PostgreSQL SageMaker
## Business Context
Stanbic Uganda wanted to move beyond broad customer segments and generic campaign content. Marketing teams were spending weeks manually building segments, creative, and copy — while customer data sat fragmented across systems, making true personalisation impossible at scale.
 Problem
### Fragmented data and manual campaign production
- Customer data was fragmented across multiple systems, making a single, reliable view of each customer difficult to build and forcing broad, generic segmentation rather than anything tailored to the individual.
- Campaign creation relied on manual effort at every stage — 15 analysts to hand-build segments, two resources spending three weeks on creative imagery, and four analysts spending two weeks on content — which meant a full campaign took roughly 8 weeks from creation to launch.
- Without personalisation, campaigns were built around just 3 broad segments and a single generic email, and performance reflected it: weaker open rates, weaker reactivation of dormant customers, and lower conversion than a targeted approach could achieve.
 Solution
### AI-powered personalisation platform on AWS
- Melio AI built an AI-powered personalisation platform using Amazon Bedrock for generative AI within Stanbic's own AWS environment, alongside a prototype so executives could see the end-state experience before committing to a full build.
- The platform ingests campaign briefs and raw customer data, automatically generates insights and hyper-segments, and bulk-generates on-brand images and text for each segment — with collaboration, governance, and a human-in-the-loop review step throughout.
- This was proven on a live customer reactivation campaign: 46 unique hyper-segments, each with personalised imagery, text and varied calls to action delivered by email, replacing the 3 broad segments and single generic email used previously.
 Benefits
### Outcomes delivered
- Campaign turnaround dropped from 8 weeks to 1 week, freeing the personalisation team to focus on strategy rather than manual production.
- Open rates rose 45% (14.10% to 20.40%) and conversion rates rose 62% (4.74% to 7.7%), showing that hyper-personalised content resonates far more than generic, broad-segment messaging.
- Dormant customer reactivations rose 62% (1,701 to 2,765), and the campaign delivered a UGX 872 million uplift in balance sheet impact — a 556% increase over the equivalent traditional campaign.
## Measurable benefits
**12 weeks**
from kickoff to live users
**+62%**
conversion rate uplift
**+45%**
open rate improvement
**556%**
balance sheet impact vs traditional
## Prototype-first, then production at scale
We started with a working prototype inside Stanbic's AWS account so stakeholders could experience the end-to-end workflow — from brief ingestion to segment generation to creative review — before committing to a full build.
The production platform was designed around governance from day one: human-in-the-loop approval for all AI-generated content, audit trails for every campaign decision, and role-based access aligned to Stanbic's marketing operations.
1
### Discovery & data mapping
Mapped customer data sources, campaign workflows, and brand guidelines to define integration points and personalisation boundaries.
2
### Proof of concept
Built a clickable prototype on Amazon Bedrock with live data samples, validated with marketing leadership on a reactivation use case.
3
### Production rollout
Deployed the full platform with segment engine, creative generation, review workflows, and email delivery integration.
## Platform in action
Screenshots from the deployed solution — dashboards, workflows, and interfaces built for this client.
### Campaign overview dashboard
- Single view of problem, solution, and measured outcomes for stakeholder reporting.
- Tracks segment count, creative variants, and campaign status in one place.
- Used by marketing leadership to sign off before each send.
### Hyper-segment builder
- Generates 46 unique segments from a single campaign brief and customer data.
- Replaces manual analyst work that previously took weeks per campaign.
- Each segment gets tailored imagery, copy, and call-to-action variants.
### Creative review workflow
- Human-in-the-loop approval for all AI-generated imagery and copy.
- Brand guidelines enforced automatically before content reaches reviewers.
- Full audit trail of every edit and approval before email send.
## In their words
“
"The prototype let our executives see exactly what hyper-personalisation would look like before we committed budget. That changed the conversation entirely."
**Marketing Director** Stanbic Uganda
“
"We went from three broad segments to 46 tailored ones in a single campaign. The team finally has time to think about strategy instead of manual production."
**Head of Customer Engagement** Stanbic Uganda
[All case studies](https://pages.melio.ai/case-studies?hsLang=en) [Discuss a similar project](https://pages.melio.ai/contact?hsLang=en)
[](https://pages.melio.ai/?hsLang=en)
We build and run production AI systems for enterprise – agentic platforms, MLOps infrastructure, and document intelligence – for financial services, healthcare, manufacturing and technology leaders. An AWS Advanced Tier Partner with AI Competency, we take AI from pilot to production, and productize what we learn through Highwind, our AI-as-a-Service platform.
Company
[Home](https://pages.melio.ai/?hsLang=en) [Services](https://pages.melio.ai/services?hsLang=en) [Products](https://pages.melio.ai/products?hsLang=en) [About](https://pages.melio.ai/about?hsLang=en) [Case Studies](https://pages.melio.ai/case-studies?hsLang=en) [Insights](https://pages.melio.ai/insights?hsLang=en) [Contact](https://pages.melio.ai/contact?hsLang=en)
Get in touch
Workshop17 The Bank
24 Cradock Avenue, Johannesburg
[poke@melio.ai](mailto:poke@melio.ai)
© 2026 melio.ai. All rights reserved.
[Privacy Policy](https://pages.melio.ai/privacy-policy?hsLang=en) [PAIA Manual](https://pages.melio.ai/paia-manual?hsLang=en)
Cookie Settings
Built on AWS · Powered by humans who actually give a damn.