Trust & AI Governance | Eminence VSP

What touches what

Tools, and their boundaries.

Four categories of tooling appear in our delivery. Each has an explicit boundary.

BUILD

Claude Code + Salesforce DX

Anthropic's Claude Code generates metadata, configuration and tests against development orgs through governed tooling. It works on metadata and synthetic data only — it is never connected to an org containing your customer data.

PLATFORM

Salesforce-native AI

Agentforce, Einstein and Data Cloud run inside your Salesforce environment, under your org's permissions, encryption and audit. We configure them; your data never leaves your tenancy to reach them.

REVIEW

Quality gates

Everything AI produces passes the same gates as hand-built work: static analysis, automated tests, security scanning and UAT — then a named architect signs it. No unreviewed output reaches a client org.

THIS SITE

Website & communications

This website is hosted on Vercel with privacy-respecting analytics and no advertising trackers. Contact-form details are used only to reply to you — the privacy policy has the specifics.

Inside a live engagement

What access actually looks like.

Take a digital lending build — our flagship engagement — step by step. The pattern is the same for every vertical.

01
Your org, your rules. The architect works as a named user in your Salesforce environment — permissions you assign, MFA you enforce, every action in your own audit trail. Access is revocable by you, instantly, at any time.
02
The build lives in sandboxes. Configuration, flows and decision matrices are built in sandboxes and scratch orgs. Test data is synthetic; where a full sandbox carries production data, your masking policy governs it.
03
Nothing lands on our systems. We are not a data processor in your stack — Salesforce is, under your existing contract and region. Documents stay in your systems; our laptops hold metadata and notes, never your customers.
04
One named person. You always know exactly who has access, because it is one identifiable architect — not a rotating offshore pool. The same name signs the deliverables.
05
Clean offboarding. At engagement end, access is removed and we retain only what the contract requires — deliverable documentation, nothing operational, nothing about your customers.

Data handling

Five commitments.

01
Your data stays in your org. Customer data is accessed only inside your governed Salesforce environment, under your access controls, by named individuals you have approved.
02
Nothing of yours trains a model. No client data, metadata or documentation is used to train any AI system — public or private. Ever.
03
AI works on metadata and synthetic data. Where AI accelerates the build, it sees structure and test fixtures — not customer records.
04
A named person answers for every deliverable. AI suggests; an accountable architect decides, signs and answers for it — the same principle SMCR applies to your own people.
05
Transparent on request. In any engagement, ask which parts were AI-accelerated and how they were verified. We will show you, artefact by artefact.

Before you ask

The diligence questions, answered.

Where does our customer data live during an engagement?

Exactly where it lives today: in your Salesforce tenancy, in your contracted region, under your agreement with Salesforce. An engagement with us adds no new data residence — we work inside your environment, not alongside it.

Does any of our data ever leave the platform?

No. Documents, records and reports stay in your systems. Our own devices and tools hold engagement metadata — designs, configuration, project notes — never your customer data.

What does your AI tooling actually see?

Structure, not customers: metadata, configuration and synthetic test fixtures in development orgs. Claude Code is never connected to an org containing your customer data, and nothing from your engagement trains any model.

What happens when the engagement ends?

Your admin removes our user — one click, effective immediately, logged in your audit trail. We retain deliverable documentation as the contract requires, and nothing else.

Do you hold security certifications?

We don't list badges we don't hold. What we offer instead: your security questionnaire answered personally by the architect within days, every answer standing up to a follow-up call, and contractual commitments on confidentiality and data handling in the engagement agreement.

Who do we call if something goes wrong?

The architect — directly, not a ticket queue. One named, accountable person is the whole point of working with a boutique.

Straight answers

What we don't claim.

We are a boutique, and we won't decorate this page with badges we don't hold. What we offer instead: direct answers from the person who does the work. Send us your security questionnaire — it comes back completed in days, not quarters, and you can put every answer to the architect in a call.

How we build with AI — the full method →

Built for your risk team