Trust

Trust is the point, not a feature.

We help teams run on AI. That only works if two things are true: you can trust what the AI is doing, and you can trust how we operate. This page covers both.

How we think about trust

As your team runs on AI, trust becomes the hard part.

Getting a model to do the work is no longer the challenge. The challenge is knowing that what it produced lines up with your policies, your obligations, and your risk tolerance. That kind of trust isn't a setting you switch on. It comes from calibration to your specific risk posture, work that stays visible and reviewable, and accountability for the judgment calls involved.

That belief shapes how we build. The Skills we deliver are tuned to your playbook, run inside your own Claude environment, and stay editable by your team. Nothing is a black box. It's also where our work is heading: as more of your day-to-day runs on AI, we're building toward a trust layer that sits over all of it.

It starts with the most basic question of all: how the Skills we give you handle your data. The short version is that they don't touch our systems at all.

How we handle your data

Your data stays in your tenant. We never see it.

Built for the security and compliance teams that will review this. Share this URL with them, and we're happy to walk through any of it on a call.

Four principles. Zero exceptions.

01

Skills run entirely inside your Claude environment

Once installed, the skill lives in your authorized Claude account (Cowork, Claude.ai, or Claude Code). There is no AbundanceAI server in the loop.

02

We don't see, store, log, or process any of your data

The skills are static markdown files. We never receive or have access to the documents you analyze or the outputs the skills produce.

03

Data handling follows your existing Claude tenant terms

All data processed by a skill is governed by your existing agreement with Anthropic. The same controls, residency, and retention that apply to your other Claude usage.

04

No telemetry, no analytics, no callbacks

Skill files are plain text with no usage-tracking mechanism. We do not know which skills you use, how often, or against what inputs.

Scope

What our Skills are. And what they aren't.

What our Skills are

  • Structured prompts and frameworks encoded as Claude instructions
  • Decision-support tools calibrated to market-standard positions
  • Drafts and analyses designed for review and editing by qualified counsel
  • Tools that compress time spent on routine analytical work
  • Customizable. Edit any SKILL.md to reflect your firm's playbook

What they are not

  • Not legal advice
  • Not a replacement for qualified counsel
  • Not autonomous agents. They require user invocation and review
  • Not connected to AbundanceAI infrastructure after install
  • Not certified or warranted to produce specific outcomes

Want to walk through this with your security team?

One call. We answer the questions. You decide.

Talk to us