How I handle your data, in plain terms. Xcelerent builds AI into report- and document-heavy workflows for advisory, valuation, and professional-services teams — often the kind where confidentiality and data residency are the first questions asked. This page gathers, in one place, how I handle your information across everything I build. It describes how I work; it is not a legal privacy policy — data residency, hosting, and approvals are scoped for each engagement.
Your data stays where it is#
The systems I build are deployable in your own environment, or a private, isolated instance — and where the data genuinely cannot leave the building, I use local inference so it does not have to. Connectors read where your data already lives; on-prem systems connect through a governed connector built for your environment, and your data stays on your systems. Data residency and hosting are constraints we agree up front, not afterthoughts.
Retrieve and cite — never train on your data#
The knowledge layer retrieves and cites real answers rather than generating plausible-sounding ones — every answer traces back to a real source you can check. Your documents are never used to train third-party models. Confidential material is kept out of reach by design: only reviewed, published knowledge is ever answerable, and restricted or sensitive material is excluded by design, not by policy alone.
Scoped, permissioned, logged — with a person in control#
Assistants run read-only by default, inside the permissions you set, with every query logged and an audit trail. Anything that writes back needs a person in the loop. The system flags risk — stale, unreviewed, or contradictory material — but never auto-certifies anything as true; promotion to “trusted” is a human step. The goal is an AI that answers from your reviewed knowledge, with sources, under your control — not a general model guessing.
Communications handled with restraint#
I do not vacuum up inboxes or chat history. Sources such as Microsoft Teams messages or email are included only with your explicit authority, scoped access, and human-reviewed governance — and only where you choose to.
You own what I build#
Ownership is the default, not an upsell. Documentation and handover are part of the scope, not an extra invoice — you should own what we build, with no lock-in. If it cannot be handed over cleanly, it is not finished.
A note on this site#
The contact form is handled by Formspree, and I use your message only to reply and decide whether there is a sensible next step. Please do not send sensitive client data, credentials, or confidential documents through the form — bring those to a scoped engagement instead.
Product names are referenced for compatibility and description only, and do not imply partnership, endorsement, or certification with Microsoft or any third party.
Data residency, hosting, approvals, and risk are worked through for your context on the discovery call. Request a discovery call · Email lingtao@xcelerent.com