Tallie AI

AI you actually own.

Tallie is an AI control layer for finance, operations, and sales teams. Your data stays in your systems, you choose the model you run, and the cost of the stack stays in your hands. Real engineers embed with your team to encode how you actually work.

OUR MISSION

Make governed AI execution a default, not a privilege.

Most enterprise AI today asks you to copy your data into a vendor's environment, accept their model choice, accept their per-seat pricing, and trust that none of those will change in ways that hurt you. That's a bad trade for finance, ops, and sales teams that take their data perimeter, their model strategy, and their cost base seriously.

Tallie inverts that. The agent runs on your infrastructure, reads from your systems in place, uses the model you choose per task, and executes only the skills your team has approved. A forward-deployed engineer embeds with you to make sure the work the agent does is the work you actually do — not a generic copilot's best guess.

We think the result is the only credible AI posture for regulated, mid-market, and enterprise teams: AI productivity without giving up control of your data, your model, or your cost.

OUR PRINCIPLES

What we believe

Own your data

AI execution should happen against your data, in your perimeter — not by copying it into a vendor's central AI lake. Read-only by default. Tight audit on every run.

LLM-agnostic by contract

OpenAI, Anthropic, open-weights, your own self-hosted model — and route per task. Switching providers when pricing or capability shifts is a configuration change, not a rebuild.

Forward-deployed engineers

Real engineers embed with your finance, ops, and sales teams to wire up data and author the skills that encode how you actually work. The opposite of an empty platform with a docs link.

Cost-efficient stack

The marginal cost of running a workflow is set by your routing strategy — not a per-seat SaaS curve. Open-weights and self-hosted models are first-class so you can drive cost toward zero on high-volume work.

THE TEAM

Built by Mercury Labs

Tallie is built by Mercury Labs, a London-based AI engineering studio that designs and deploys practical AI for businesses.

Our team of data scientists and full-stack engineers has spent years building AI products across fintech, healthcare, manufacturing, and market intelligence — for organisations including PrimaryBid, Augmentive, MTC, HVM Catapult, and Forestreet.

As an approved supplier through the Crown Commercial Service, we've also worked directly with the UK Government on AI projects, partnering with Innovate UK and other public sector bodies to deliver AI at scale in regulated contexts.

Tallie is the platform we kept wishing existed every time a customer asked for AI productivity without handing over their data, their model choice, or their cost base. So we built it — and we deploy it the way we wish all enterprise software was deployed: with a real engineer in the room.

Based in Eccleston Yards, London.

Talk to us

Your data. Your model. Your infrastructure.

Bring AI productivity to your finance, operations, and sales teams without handing over your data estate, your deployment posture, your model strategy, or the cost of your stack. Your processes encoded as skills, authored with you by our engineers — on-prem or VPC, LLM-agnostic, governed by default.