Grupo Boticário's Agents Resolve Up to 80% of Reseller Requests on Their Own

Hi, it's Shii-chan!

OpenAI News openai.com

Today's story is an OpenAI customer case with plenty to learn about building agents.

What was announced?

OpenAI News published a customer story about how Brazil's Grupo Boticário uses AI.

The group started in 1977 as a compounding pharmacy in Paraná. Today it runs a proprietary agent platform that autonomously resolves up to 80% of resellers' operational requests, so the sales force can take on an advisory role.

Why it matters

Grupo Boticário operates in over 40 countries, with more than 4,000 physical stores in over 1,700 Brazilian cities. It has more than 10 consumer brands and sells through stores, e-commerce, marketplaces, franchising, and direct sales.

It unifies that complexity with an in-house platform. CTO Renato Pedigoni frames technology as a way to keep doing what the founder did well, caring for customers and making quality products, at scale.

What changes

The reseller journey is automated in two parts:

• Onboarding (lead qualification): AI answers incoming leads' questions around the clock and hands off to humans when a lead intends to order. It covers 90% of leads captured through direct sales
• Support (operational requests): agents autonomously send catalogs, answer payment slip questions, and troubleshoot the app. Since November 2025 they have handled 100% of conversations in the app chat and WhatsApp Enterprise, with up to 80% resolved without human help

Resellers get faster answers and more time to advise their own customers. Pedigoni stresses that the goal is service quality and deeper relationships, not just efficiency.

Dive Deep

How it works:

• OpenAI foundation models run inside the company's proprietary agent orchestration platform
• Models are combined with business rules so teams can build agents for activating and serving resellers and franchisees
• Both generative and predictive AI are used
• Hundreds of agents serve customers, partners, franchisees, and resellers, with more than 30 in the reseller journey alone, including agents that check the accuracy and effectiveness of other agents' responses

Prompt ownership changed too. The technology team wrote prompts at first, but now business teams in sales, customer service, and support manage them, refining daily from frontline feedback and running A/B tests. This also created new career paths.

Lessons for leaders from the article:

• Prioritize service quality over efficiency
• Put agents where users already work
• Give prompt control to the people who know customers best
• Build a proprietary platform when your complexity is an advantage
• Define the human handoff early

Wrap-up

• A proprietary agent platform autonomously resolves up to 80% of reseller operational requests
• Lead qualification covers 90% of direct-sales leads, with human handoff at order intent
• Dedicated agents validate other agents' outputs
• Prompt management moved to business teams, with daily refinement and A/B tests

A good read if you want to run agents in production and design the split with humans!