miniflo

What miniflo is

miniflo designs, builds and runs AI voice agents and agentic AI automation for operations teams in the UK and India. Each agent is engineered like a piece of infrastructure: it’s documented, tested and monitored, and it goes live in four weeks.

As co-founder, I split my time between two jobs. I build the agents and the automations behind them, and I own the brand, the website and how we explain all of it to clients.

12 voice agents, built to spec

Every agent is a standard component that we configure to the client’s stack:

• Calls and support: inbound receptionist, outbound campaigns, support and FAQ agent, internal service desk
• Finance: insurance renewals, EMI reminders, banking and card servicing
• Retail: order tracking and returns, shopping and cart recovery
• Telecom and healthcare: service alerts and routing, patient appointments, clinical documentation

Agents answer around the clock, handle up to 20 calls at once and speak 40+ languages.

Anatomy of an agent

Every miniflo agent is built from the same five layers, so it stays predictable, testable and easy to audit:

1. Interface: chat, email, voice or API, wherever the work arrives
2. Reasoning: the right model for each task
3. Memory: the client’s SOPs, CRM history and past decisions
4. Tools: secure actions in HubSpot, Stripe, Slack and 150+ apps
5. Guardrails: approvals, limits and a full audit log

We plug into the tools a team already runs, including Salesforce, HubSpot, Zendesk, Freshdesk, Twilio, Exotel, WhatsApp, Slack and n8n, so nothing has to be ripped out.

The website

I designed the site around a single idea: automation, drawn to spec. It uses a blueprint grid, technical callouts (OP-01, OP-02…), spec tables and a deep brand blue. The job is to make AI feel engineered rather than magical.

• A “manual work is a hidden tax” table puts a number on the problem before we pitch anything.
• There’s a use-cases library covering 20 use cases across 4 industries, with FAQ and JSON-LD for search and AI crawlers.
• An llms.txt file and structured data help AI assistants describe miniflo correctly.