Capability is commoditized.
Ownership is not.
Anyone can demo an agent. What you can't buy off a shelf: workers designed for your business, proof before production, decisions your auditors can read, and code you keep.
Workers and software, built to order.
No catalog to pick from. Describe your business and an AI business analyst proposes five alternative worker designs — each with its steps, the systems it needs, its approval points, an estimated running cost and an expected return — then checks, scores and recommends one. Designed in minutes, not months, and tested on sample data before anything goes near production.
Rules before AI
Design is where policy and judgment are separated: decisions that follow fixed business rules become rule steps that run instantly and identically, with no AI call — and AI is kept for the steps that genuinely need judgment.
How the Foundation worksBuilt from your context
Your organisation, team, systems, pain points, goals, rules and approvers shape every design — not a generic template.
Multi-modal intake
Describe your business with a 10-step guided wizard, pasted notes or a single prompt, and add documents — design-time inputs you choose to share, used only to design your workers, with a data-processing agreement in place before anything moves.
Five designs, not one
Conservative to ambitious, compliance-first to cost-optimal — each design that passes validation is scored on fitness, safety, compliance, cost, simplicity and scalability.
Refine visually or in words
Shape the design on a drag-and-drop canvas, write rules in plain English, or deepen a design step by step.
A workforce designed for your business, where fixed rules cost no AI tokens — and every design shows its estimated running cost before you commit.
Proof before production.
Test every worker on sample data generated for it — or on files you import — with a real AI model, while you watch each step live. Pause at approval steps, build evaluation suites, compare versions and rerun scenarios after every change — before a single real record is touched.
Tested on sample data
Preview runs on sample data, and Studio holds no credentials for your live systems — so an evaluation that starts from a description of your process needs none of your data — and no data-sharing agreement.
How the Foundation worksSample data for every worker
A data model and realistic records generated for each worker — or your own CSV, Excel or JSON files.
A score for every run
Evaluation suites score runs on six dimensions, with golden datasets, A/B tests and regression checks.
Compare & rerun
Compare versions side by side and rerun saved scenarios after every change — rule steps return the same answer every time.
Adversarial testing
Optional red-team runs attack a worker during preview, across six categories of attack, before it ever goes live.
Production comes only after testing and your approval gates — so your security team can say yes sooner.
Decisions your auditors can read.
Every worker is a readable, versioned workflow your governance team can compare, test and review before deployment — the workflow is the audit artifact. It ships inside the export, so what was reviewed is what runs.
Readable workflows
What a worker does, who approves what, where it escalates — captured as a readable, versioned flowchart, not a tangle of prompts. Your engineers compare it like code, your governance team reads it before deployment, your auditors check against it after.
How the Foundation worksVersioned like code
Compare versions side by side, fork and merge them, and return to any earlier version.
Approval steps, designed in
Approval gates assigned by role, team or variable — and Studio flags any sensitive action that is left without one.
Audit-ready exports
A threat model and a signed AI bill of materials in every export, plus an optional tamper-evident audit log in the running worker.
Audit-trail specification on the Trust CenterEscalation as a control
Route low-confidence cases to a human-approval step and they wait for a person instead of proceeding on their own.
Your auditors read what a worker will do before it does it — and the artifact they read is yours.
Software you keep and can defend.
The platform delivers a complete, self-contained Python project — not slides: the workflow, its prompts and code, the runtime, tests and deployment files — and you run it on your own infrastructure. No vendor runtime. No per-run fees. No calls home.
What's in an export: details on the Trust CenterYours to run
Every export includes a security threat model and an AI bill of materials listing what's inside, with a signed attestation your engineers can verify offline. The code carries no licence keys and makes no calls to Nexgile — check it on your own hardware.
How the Foundation worksConnect through APIs
Workers call your systems through their APIs: upload an OpenAPI description — or let AI draft one — and Studio creates a private connector.
Signed exports
Each export carries a signed hash of its contents, so you can confirm the package you deploy is the one you reviewed.
Human oversight in the box
Approval steps travel with the worker, and a reviewer-inbox app (still in progress) can be included in the export.
Trust built in
Optional guardrails, personal-data redaction, policy checks, tamper-evident audit logging and monitoring dashboards — chosen when you export.
100% yours
A complete Python project, self-hosted on your servers, a private cloud or a public cloud — with no license keys and no runtime dependency on Nexgile.
Nothing stops working if you stop paying us — and you don't have to take our word for it.
Three ways to get a digital workforce.
Each lane is a real choice with real trade-offs. Here is how we'd describe them — including ours.
Build it yourself on open frameworks.
You own the code — and the years of plumbing, governance, and maintenance you'll build before the first worker ships.
Rent a hosted agent platform.
Fast start — on their runtime, behind their meter, inside their black box, for as long as you use it.
Generate it and own it.
Designed for your business, tested on sample data, delivered as code in your hands. This is the lane we built.
Artifacts, not adjectives.
The capabilities above produce things you can read. Two examples — both illustrative, both standing in for artifacts your team receives.
- Intake: vendor invoices from email and the ERP queue
- Check: purchase-order match, payment terms, duplicates
- − Approval: mismatches route to one reviewer
- + Approval: mismatches above a limit route to finance
- Low confidence: routed to human approval
Illustrative. A policy change reviewed like code — read before deployment, traced after.
How workflows work, on the platform page- Scenarios run on generated sample data
- Evaluation suite scored on six dimensions
- Low-confidence cases routed to a person
- Regression checks rerun after the change
Generated from a preview run on sample data — evidence for your procurement file.
See how testing works in Nx·StudioThe delivery record.
Portal and worker counts are computed from our delivery records. A digital worker is a generated, governed software worker delivered as owned source code — not a rebadged bot or script.
The same capability, two very different endings.
- Their runtime.
- A meter that punishes success.
- A dashboard over someone else's black box.
- Cancel, and the workforce vanishes.
- Your infrastructure.
- No meter — and rule steps that never call a model.
- Workflows your auditors read before deployment.
- Cancel, and nothing happens — it's your code.
We charge to design your workforce, not to run it. Our incentives point at your independence — ongoing help is optional, on your terms.