Ask the hard questions. We’ll show you the proof.
Bring your security team. Before a digital worker goes live, they can read its blueprint (a step-by-step plan anyone can follow) and watch a trial run on test data. An optional test even attacks it the way a hacker would. This page shows how we back each promise, and what we don’t claim yet.
Europe's data protection law
California's consumer privacy law
Encrypted on the move and in storage
A person approves the key steps
An outside audit of our security
First on our roadmap — target date: being finalized.
Managing information security
Responsible AI management
Usable by people with disabilities
“In place” means it’s true now. “Roadmap” means a certification from an outside auditor that we’re working toward and don’t hold yet. We won’t claim a certification until we have it. For the current status and documents, email security@nexgile.com.
You’ll see Roadmap on some labels, and that’s deliberate. We separate what’s true now from what’s still coming, with the same labels on every page of this site.
SOC 2 Type II — target date: being finalized.
Until then, here’s what you can count on:
- Your workers run on your own systems, under the security policies your team already uses.
- You can read our audit-trail specification, the full list of what a worker records.
- Your security team can request sample files from a download, with sensitive details removed.
Security work, done before handover.
Each download ships with a threat model: a plain map of the security risks and how each one is dealt with. We run our own platform to enterprise standards.
- Data encrypted on the move and in storage
- People see only what their role permits
- Every download maps its risks and data flows
- Ongoing monitoring, activity logs and a plan for incidents
- A simple, responsible way to tell us about security issues
Your records stay in your hands.
You can try workers entirely on test data, so no real records need to leave your hands. Once delivered, the software runs on your own systems, not in a Nexgile cloud.
- Build and try a worker with no real data
- Your own servers or cloud, not ours
- Data residency, meaning where data is kept, across our Herndon, VA and Hyderabad, India operations
- In line with the GDPR and CCPA privacy laws, including keeping or deleting data on request
- Ask us for the list of outside firms that handle data on our behalf
AI that can explain itself.
We treat AI oversight as a core feature, not a footnote. Each worker is built so your team can explain and check every decision it makes.
- A readable plan instead of “the AI decided”
- Your rules built in wherever possible: same facts, same answer
- At each sign-off step, the worker waits for a person
- Every sign-off logged, plus an optional log that exposes later changes. Read the audit-trail specification
- Optional hiding of personal details when the worker uses its tools (off by default)
Four promises your team can test.
Everything we promise on this site comes back to these four. Check any of them in a live session, in a document your auditors ask for, or by replaying a past test run.
Rehearsal before reality
Every worker does its job on realistic test data, made up so none of your real records are needed, before it touches anything real. Run it step by step, watch it stop for sign-offs, test several workers together and even fast-forward time. An optional test attacks it the way a hacker would. With no real data involved, your security team can approve a trial that starts from a written description, with no data-sharing agreement.
How Nx·Studio does itRules you can rely on
Wherever possible, the rules you set are built right into the worker, and we flag any that can’t be. A built-in rule never calls on AI, so identical cases always get identical answers. Some routine decisions are never written down. Our separate Cognitive Engine learns those from past examples, and acts only once it passes an exam on cases it hasn’t seen before.
See how the Cognitive Engine learnsBlueprints you can read
Every worker comes with a blueprint anyone can follow: what it does, who signs off on what, and when a case is handed to a person. Each version is kept, so you can compare changes or roll back. Business teams can see it as a simple chart of who does what. Like everything the platform makes, it belongs to you.
How oversight worksProof in every download
Each download contains a signed list of everything inside it (a bill of materials) and a tool your IT team can use to check it offline. It also carries the threat model and a guide that maps it to the documents the EU AI Act (Europe's law on AI) asks for. The Studio adds a full record of what you received. The worker runs without us: no Nexgile software, no license check, no usage meter, and nothing sent to us. Your team can confirm “no lock-in” on its own.
See the proof for yourselfYour IT team can check it all, offline.
Every download carries proof your team can check: a signed list of every part of the software and its AI (a bill of materials). Your IT team checks it offline, on your own computers, without ever contacting us. The Studio adds a full record of what you received. Ask us for a sample with sensitive details removed, and judge it for yourself.
Every worker is built to run without us. There’s no Nexgile software inside, no license check, no usage meter, and nothing is sent to us.
Rules hold up the same way. In any live session, ask us to run one case twice, and you’ll see your rules give the same answer both times.
- Signed bill of materials Every part of the software and its AI, listed and signed, with an offline checking toolincluded
- Threat model The security risks, how each is handled, and where data goesincluded
- EU AI Act cross-reference Links the Act's documentation items to the bill of materials. This is not an official compliance assessmentincluded
- Ownership report A complete record of what was delivered, plus a PDF summaryfrom Nx·Studio
Your computers, your check, your verdict.
See every step before a worker takes it.
Nothing hidden. Your team gets plain documents it can read, compare version by version, and keep.
One plan everyone can read
Each worker comes with a blueprint anyone can read, with every version kept. It lays out the worker’s AI agents and what each one does, its connections, its sign-off steps, where each case goes and its rules. Nothing is tucked away in AI instructions or a supplier’s software. Developers compare versions, your compliance team reviews it before launch, and auditors check against it afterwards.
A record of every sign-off
Each sign-off is written to an audit trail: who approved what, and when. Two extras are optional and off by default. One is a log that exposes any later change. The other is a record of where each piece of data came from (data lineage). Built-in rules give the same answer to the same case every time.
Sign-offs happen in software you own
Sign-offs take place in a review inbox built into the worker, and if you switch it on, requests can also reach people in Slack, Microsoft Teams or email. You choose who approves each step, such as a role or a team. Because it’s all code you own, your approval rules are enforced in your software, not by a setting on our servers.
Where your data goes, and where it doesn’t.
Every worker gets test data shaped like its own work, so trial runs, test runs and attack tests need none of your real data. If you choose to upload your own sample data instead, it stays in your workspace and only its members can see it.
What you share while we design is handled on its own terms. You might share documents, process records or sample data so your workers fit the real job. You decide what to share, it’s used only to design your workers, and a data-processing agreement (a contract on how your data is handled) is signed first.
Begin with a written description, and that’s all you’ll need to share.
What’s fixed, what can change, and who has the final say.
Ask an AI model twice and you may get two answers.
So your rules don’t rely on AI. Wherever possible, they’re built into the worker and give the same answer every time, and you can replay a past test run step by step to confirm it. Some routine decisions are never written down. Those can go to our separate Cognitive Engine, which learns from past examples and acts only after passing an exam on cases it hasn’t seen.
Where AI judgment is needed, a quality check can score each answer. You set the minimum confidence score and pick what happens below it: flag it, stop the run, or send the case elsewhere, such as to a person. By default, it only raises a flag. At a sign-off step, the worker waits for a person to decide.
Your team decides when a person steps in.
What your auditors can ask us for.
We share each one on request, with sensitive details removed. To ask for any of them, email security@nexgile.com.
What a worker will do
The who-does-what chart and download summary for one sample worker, so your team sees its actions before it takes them.
How a test run scored
A report from a trial run on test data, the kind purchasing teams usually ask for.
A look inside a download
Its folder layout, a page from its operating guide and a test report, so you can see what “ready for real work” means.
What the threat model covers
An outline of the threat model in each download, without the set-up details.
The audit-trail specification
Each event the audit trail records, and how the optional log exposes any later change.
Proof of what you own
For one sample download: the complete list, the signed bill of materials and the checking tool’s output. The verdict is yours.
Your auditors can review every step before the worker takes it.
Three promises we can prove.
One sentence each, covering only what the platform can prove. Better a small promise you can check than a big one you can’t.
You can build and try a worker without sharing any real data.
Your workers run with no Nexgile software, no per-decision fee to us and nothing sent our way.
Stop paying us, and your workers keep working.
Each number on this site tells you how it was measured. If we can’t measure something honestly, it doesn’t go on the site.
The same goes for other people’s numbers: we quote them only with their source and date. Our delivery counts explain how we counted, and we check them again whenever we publish. If a figure can no longer be confirmed, we take it off every page.
Staying available
The software we deliver runs on your own servers or cloud, so you decide when it’s available. We monitor our own platform continuously and keep a plan for handling incidents. A public status page for Nexgile services is on the way.
Reach the right people
- Security questions, or to report a problem security@nexgile.com
- Privacy and your data rights privacy@nexgile.com
- Accessibility requests accessibility@nexgile.com