# Internal applications in the hands of every store and field colleague.

How CVS Health put grounded, AI-assisted decision support in the hands of **200,000+ frontline colleagues** across **10,000+ retail locations** — and turned generic enterprise tools into operations-grade internal applications designed around the way the work actually gets done.

## Client
CVS Health

## Industry
Retail · Healthcare retail

## Scope
Multi-year, multi-app

## Departments
Store Ops · Asset Protection

## Outcomes at scale
### In production, every shift, every store.

- **10,000+** stores running BizzSoftware-built applications
- **200,000+** colleagues using the applications daily
- **100,000+** decisions powered every day
- Thousands of questions answered with AI daily
- **45%** measured productivity gain

## 01 The challenge
CVS Health operates more than 10,000 retail locations and employs more than 200,000 store and field colleagues. Decisions at the edge of that operation — a policy question at the register, a suspected loss prevention case, a planogram exception, a complex customer interaction — happen thousands of times per day, in seconds, with limited training.

Generic enterprise tools couldn't keep up. The off-the-shelf AI products available to retail were trained on the wrong data, integrated badly with operational systems, and made no concession to the realities of frontline work — handheld devices, time pressure, varied training levels.

CVS needed internal applications designed specifically for the realities of retail at this scale: mobile-first, fast, integrated with the systems and workflows colleagues already use, AI features built in where they earned their place, and trusted enough to act on.

## 02 The approach
BizzSoftware partnered with CVS's Store Operations and Asset Protection teams as both strategic advisors and hands-on builders. The engagement spanned the full lifecycle: identifying the highest-leverage workflows, designing application architectures around them, building the production systems, and running them at scale — with AI features designed in where they removed work for colleagues.

The work was anchored by three principles, applied across every application we built.

### Built around the colleague, not the model.
Every interface designed for the actual context of store and field work.

### Integrated, not layered.
Each app connects directly into the operational systems colleagues already use.

### Designed for trust.
Every AI output grounded, citable, auditable. Human-in-the-loop on high-stakes actions.

## 03 What we built
A suite of internal applications — with AI built in — now in daily use across the CVS footprint.

- **01 Frontline decision support for Store Operations**
  Colleagues query policy, procedure, and operational guidance in natural language and get grounded, accurate answers in seconds — backed by retrieval over CVS's authoritative operational documentation, with citations on every response.

- **02 AI-assisted Asset Protection workflows**
  The Asset Protection team uses an internal application, with AI built in, to surface, prioritize, and document loss prevention cases at scale — turning hours of manual case review into focused work on the cases that matter most.

- **03 Mobile-first applications for field colleagues**
  District leaders and field staff make better-informed decisions with applications designed for the mobile context they actually work in — with AI features (search, drafting, decision support) built in.

> Software at the edge of a 200,000-person retail operation isn't a feature. It's a platform decision.
> — BizzSoftware engagement principle

## 04 Outcomes
The applications BizzSoftware built and runs are now part of how the largest pharmacy retailer in the country operates day to day — embedded in the work of every store, every shift, every colleague. CVS measures a **45% productivity gain** on the supported workflows, with usage now spanning **200,000+ colleagues** answering thousands of AI-assisted questions and powering more than **100,000 decisions** every day.

**What the metrics measure:** productivity gain is the time-on-task reduction observed on the operational workflows the applications support (policy lookups, asset-protection case prep, district-leader decision support), benchmarked against pre-deployment baselines on the same tasks. "Decisions" counts each discrete moment a colleague used the application to look up policy, complete a structured workflow, or resolve a case — aggregated across the footprint.

**One workflow, before and after.** Before: a colleague at the register hits a policy ambiguity (return outside the window, prescription transfer edge case) — calls the manager, who pages the district team, who looks it up across three intranet folders. The customer waits. Sometimes the colleague guesses. After: the colleague queries the application in natural language at the register, gets a grounded answer with a citation to the relevant policy, and resolves the moment in seconds. Same colleague, same shift, same training — faster decision, and one the auditor can trace.