Brief a process. Get a tested worker.
Explain how a process such as accounts payable or claims intake runs today. Nx·Studio's AI analyst proposes up to five alternatives and estimates each one's monthly cost and hours returned. Your team tests the chosen one on realistic sample data. You then receive the complete software, ready for your IT team to deploy, with connection starter code, a reviewer inbox and optional monitoring.
Your IT team can inspect sample export files from the Trust Center during your evaluation.
Your brief
“Match supplier invoices to their purchase orders, pass mismatches to accounts payable, and flag possible duplicates for review.”
Where your process becomes a worker.
A digital worker handles one business process from first input to final result. Routine steps go to AI agents, decisions that need a person go to reviewers you name, and the worker links into your existing systems. Your team and Nx·Studio design each worker together. Each design option lists an AI-estimated operating cost and hours returned as planning inputs, and payback is estimated once, in the generation plan.
Fit the automation to the process, not the reverse.
You do not rent a hosted AI platform or reshape a process to suit one. Your team explains the current way of working and receives a worker built for it, measured in test runs and handed over as software it owns.
Everyone with a stake in the decision.
You and your operations leads judge the results, architects review the design, and IT and AI teams confirm they can run it. Everyone works from the same studio.
Six stages, from requirements to exported code.
Brief the priority process
Answer a guided form of ten steps, share your website link and process notes, or write a short summary. Uploads and eight starter packs speed up the brief.
Shortlist the AI's designs
Nx·Studio's AI analyst prepares up to five alternatives, cautious through ambitious, each with a monthly cost estimate and the hours of staff time it should free up. It scores them, repairs its own drafts and recommends one.
Specify rules and reviewers
Fine-tune your pick by dragging steps on the canvas, check responsibilities in the lane-by-lane view, and add the reference data and knowledge it will rely on.
Measure it in evaluation runs
Test it on computer-generated sample data built for this worker, with no customer records required. Walk through it step by step, try a full run that stops at each approval, or run several workers together.
Receive the complete software
Before download, an itemized list shows everything in the package. It holds the complete software for your IT team to host, with approval steps, connection starter code and, if wanted, monitoring and deployment files.
Adopt it in your operation
Your IT team runs it on infrastructure you select, with the AI model service you choose. Running it requires no Nxgyle software or license check, and Nxgyle charges nothing per decision.
Checks, not promises.
When AI models perform alike, the difference lies in how a worker is built and checked. The Foundation under each Nx·Studio worker gives your operations, IT and risk colleagues four mechanisms to test, read and verify, rather than take our word for it.
See the results before it handles live work.
Each worker gets synthetic test data shaped around its own process, so evaluations use believable cases while customers, suppliers and the ledger stay untouched. Runs can cover one step or the whole flow, and they stop at each approval until the reviewer answers. You can also test workers side by side on accelerated time, using stand-in responses or, if you prefer, the real workers. Scorecards cover accuracy, cost and safety for each run. Production data is never a prerequisite.
Policy checks with no AI call.
List the rules your business must follow, such as approval limits or payment terms, and the AI analyst encodes each one as a fixed check wherever it can. Nothing it drafts takes effect until you accept or edit it, and the analyst reports the rules it was unable to encode. Given the same inputs, a fixed check always reaches the same result and never consults an AI model. A separate product, Nxgyle's Cognitive Engine, covers routine decisions no one has documented: it learns them from confirmed outcomes, outside Nx·Studio workers.
Judge the design before it runs.
Each worker is described in plain language: every step, approver, escalation point and connected system, with nothing hidden in prompts. Every save is stored as a version, so any two can be diffed, merged or reverted. Business teams see the same design as a lane chart of the teams, people and systems involved, while the export screen gives an itemized list of the build. The blueprint is yours to keep, as is everything else the platform generates.
Confirm what was delivered.
You see the full contents on the export screen before downloading. Inside is a signed inventory of everything in the software, including its AI components, which your IT team can verify offline with the bundled checker. It also contains a security analysis for your IT security team, plus draft documents for the European Union's AI Act. Once delivered, it needs nothing from Nxgyle to run, with no license check, no metering and no calls home.
AI handles judgment calls; your people make the decisions you reserve for them. At each approval step the worker waits for the assigned reviewer. When a workflow has approval steps, the export contains a reviewer inbox for your IT team to connect, and Slack can be added as a channel. Confidence thresholds are configurable too: a low score raises a warning by default, or can send the case to a person.
Options scored before you choose.
No fixed catalog limits the choice. The AI analyst studies your teams, systems, goals, rules and current pain points, then shapes digital workers to fit them. Each design shows roles, steps and review points, with an AI-estimated operating cost and hours returned as planning inputs.
-
Your choice of intake
Work through the ten-step form, paste in a website and notes, or summarize how it works in one paragraph. The eight starter packs run from finance month-end close to IT incident triage.
-
A shortlist of up to five
Conservative, balanced, ambitious, compliance-first and cost-optimal options. Each is scored for business fit, safety, compliance, cost, scalability and simplicity, with its savings estimated.
-
Self-reviewed drafts
The analyst checks and corrects each draft before it reaches you, and can expand any step of the chosen design in more detail.
-
Written rules, enforced
List your business rules and the analyst turns each one it can into a fixed check, giving identical results on every case without an AI call. Rules it cannot convert are flagged for review.
Reviewable as a diagram.
A drag-and-drop canvas makes the workflow visible to every reviewer, technical or not. A lane-by-lane view (swimlanes) groups the same steps by team, individual and system, so process owners can check responsibilities.
-
Forty step types
Agents, rules, approvals, system connections and loops sit on one canvas, with reusable workflow patterns to start from. Building the design requires no code.
-
Lanes by role and system
Every step sits under whoever is responsible for it: a person, a team or a system.
-
Checks while you edit
The Studio flags mistakes and policy gaps during editing; a readiness score shows how near the design is to testing.
-
Every save retained
Saving records a new version, which you can compare with others, merge or roll back.
Tested before it touches live work.
Nx·Studio can generate sample data shaped to each worker's process. You see how it deals with realistic cases, exceptions included, before it touches customers, suppliers or the ledger, and no live records are needed.
-
Sample data per process
Computer-generated (synthetic) records match the data structure of each worker, and you can upload your own samples, as large as 200,000 rows.
-
One step at a time
Run a single step, restart from any point you choose, and inspect what went in, what came out and what was logged.
-
Reviews tested too
A test run halts at each review point and continues once approved. Decisions taken during testing act only on simulated systems, and any reversible action can be undone.
-
Several workers at once
In the simulation lab, multiple workers operate together on an accelerated clock. Stand-in versions run by default, or the real workers if you prefer, and the lab flags record conflicts, badly formatted hand-offs and escalation chains.
Test runs earn scores for accuracy, coverage, speed and robustness, which together form the evaluation report your procurement colleagues will request. An example based on a simulated run is available on request.
Example test report, on requestYours, in full.
You receive the complete software, ready for your IT team to host, not just diagrams. Each worker asks your reviewers for approval where you require it, operating on infrastructure and an AI model service of your choosing. Your IT team completes and tests its connections to your systems, starting from a catalog of integration connectors and starter code.
-
System connections
Starter code and catalog entries cover SAP, Oracle EBS, Workday, Salesforce and ServiceNow; your IT team adds each system’s details. API and AI-gateway connections need compatible interfaces, credentials and integration testing.
-
Reviewers you assign
Assign a role or team to each review point. Workflows with review points include a separate inbox application for reviewers, and your IT team links it to the worker. Slack requests are optional; email works only through the one supported service.
-
Hosting and model choice
Optional deployment files are available for Databricks, container hosting, Snowflake, AWS, Google Cloud and Azure, and your IT team carries out the deployment. Pick a supported AI model service, local models included, after confirming prerequisites.
-
Monitoring, if selected
Optional at export: performance dashboards and alerts, a warning if a worker's behavior gradually changes, and auto-generated operating guides (runbooks).
Compare running costs before you commit.
“A rule that settles a case costs nothing in AI fees.”
Two estimates per option
For each design, the AI estimates monthly running cost and hours returned, and the generation plan estimates payback. You can weigh options on economics, not only on features.
How much runs on fixed rules
Each business rule you state becomes a fixed check where possible. The Studio then reports what share of a worker always gives an identical result, with ways to raise it.
Fees that do not scale
Nxgyle charges no fee per decision after export. Plan separately for infrastructure, AI usage, integration, maintenance and any agreed support fees.
Review the logic before you approve it.
A blueprint sets out everything a digital worker does: agents, roles, system connections, review points, routing and rules, all on view. None of the logic hides inside AI instructions or a vendor's platform.
Saved work is versioned, so your IT colleagues can diff and merge designs the way they handle code. Business reviewers see the same design as lanes of responsibility, and an itemized list before download shows what was built. You end up owning the asset rather than renting a setting in someone else's platform.
- Expense intake — receipt checked against company policy
- Standard claims — written rule, same result each run
- Claim approval — line manager, any amount
- Unclear receipts — sent to a named reviewer
- Expense intake — receipt checked against company policy
- Standard claims — written rule, same result each run
- Claim approval — finance review above $5,000
- Unclear receipts — sent to a named reviewer
Illustrative example: a single edit between versions of an expense worker, in plain words. In the Studio, the same edit appears on the design in its version history.
Build on existing process knowledge.
Process logs, existing automation and the software bots you already operate can each seed a design.
Rank candidates from logs
Optional module: load event logs from Celonis, UiPath or similar tools to see which processes rank highest as automation candidates.
Convert existing designs
Optional module: convert existing AI agent code or standard process diagrams into a digital worker you can edit.
Bots stay in the loop
Flag the steps that software bots (RPA) handle in tools such as UiPath, Automation Anywhere or Power Automate. Your team routes the worker's hand-off requests to them, and the bots keep running on your RPA platform.
Discovery from logs and code / diagram import are optional modules, used only with the material you choose to share.
Manage workflow changes over time.
Each improvement stays in designs and code you own, and every change can be traced, reviewed and undone.
Lessons from test runs
Optional: when reviewers correct a worker during testing, those fixes can feed the knowledge base it carries, which stays your organization's property.
Scores tied to versions
Compare and merge a worker's changes, go back to a saved checkpoint, and see which version earned each test score.
Regenerate, keep your edits
Optional: include an upgrade starting point in the export, and future regenerations can be merged with your team's own code edits, with conflicts flagged.
Measured suggestions
Using test-run results, the Studio suggests improvements: a less expensive AI model, a fixed rule where AI is not needed, or steps that run in parallel. Nothing changes until you accept.
What to assess, group by group.
Nine groups, in the order a process moves from brief to working software.
Designs from your brief
Reads your business context and proposes workers to fit it.
- Four ways in: guided form, quick paste, paragraph or starter pack
- Five scored options at most, with estimated savings and costs
- Self-reviewed drafts, expanded in detail on request
- Business rules converted to rule-based steps where possible
Canvas and swimlanes
One visual design that business and technical reviewers read the same way.
- Drag-and-drop canvas, 40 step types
- Swimlanes by team, person and system
- Live checks, plus a readiness score
- Saved versions to diff, merge or revert
Data inputs
Supply the facts and files each worker relies on.
- Its own data structure for each worker
- Synthetic test data or your own samples
- Documents, a knowledge map and seed memories
- Packaged with the export when used
Evaluation runs
Measure behavior before production use.
- Single-step runs, restarted from any point
- Review points: approve, reject or undo
- Several workers simulated together
- Scores and A/B tests, summarized in an evaluation report
Integration scope
Starter code and connectors that your engineers complete.
- Entries for SAP, Oracle EBS, Workday, Salesforce and ServiceNow
- Compatible APIs and AI gateways, after integration testing
- Requests passed to your existing RPA bots
- Trigger starter code for schedules, web requests and event streams
Review and approval
People decide wherever your policy requires it.
- A reviewer inbox app your engineers connect; Slack optional
- Reviewers named by team or role
- Confidence thresholds: a warning by default, or a reviewer on request
- Approval placed ahead of actions that cannot be reversed
Safeguards
Controls designed in, not added later.
- Personal data hidden, if you enable it
- Every export includes a security analysis for IT (a threat model)
- Add policy steps to block actions you prohibit
- Compliance rule sets and a governance console, if enabled
Delivery options
Packaged for the way your teams will consume it.
- A connection for your systems (API), a chat widget, an AI gateway
- Deployment files for AWS, Azure, Google Cloud, Databricks, Snowflake or container hosting, if wanted
- Eleven AI model services, local options included
- Optional dashboards, alerts and operating guides
Ownership
Code your team deploys, runs and maintains.
- The generated code is yours in full
- An ownership report, line by line, ahead of download
- A signed inventory of contents, checkable offline
- Nothing from Nxgyle needed to run it: no license check, no metering
Evaluate Nx·Studio against your requirements.
Assess each capability with your risk, technology and business teams in a scoped evaluation.
Workers designed for your context rather than chosen from a catalog
A shortlist of up to five scored designs, reviewed before it reaches you
Stated rules run as rule-based steps where possible, without an AI call
Sample data generated per worker, so no live customer records are needed
Review points that really stop the worker, checked in testing first
Every change to a worker kept, comparable and reversible
Start from logs, diagrams, agent code or existing bots (some optional)
Your choice of infrastructure and AI model service, including local models
Each export signed, inventoried and backed by an ownership report
Recent capability additions.
- Data
Worker data, built in
A data model and generated test data for each worker, with space for your own samples, documents, a knowledge map and seed memories that can travel with it.
- Design
Drafts reviewed by the analyst
A Rules & Decisions step in the guided form, self-correcting drafts, extra detail on request and a readiness score.
- Review
Review points in test runs
Test runs pause at approvals and pick up once cleared. A list shows runs on hold and simulated actions waiting for a reviewer, and any reversible action can be rolled back.
- Views
Lane views for business owners
Swimlanes for each role, team and system, and an itemized pre-download list of the package.
- Integrations
Connection starter kits
Integration starter code in every exported worker, links to compatible web APIs or AI gateways, and secure sign-in methods for enterprise systems.
- Delivery
Delivery and hosting
Delivery as an API, a chat widget or a gateway for AI tools, with deployment files for Snowflake, Databricks, Google Cloud, Azure or AWS if wanted. Trigger starter code lets your IT team link it to schedules, web requests and events.
Optional capabilities and export settings.
Add-on modules, per deployment
- Simulated attacks before rollout
- Voice input and an AI chat assistant at intake
- An AI assistant for editing designs
- Compliance rule sets with a governance console
- Discovery from process logs; import of code or process diagrams
- Lineage tracking for data, and mathematical checks of the workflow
Chosen at export, otherwise off
- Approval requests by Slack, or email via its single supported service
- Hiding personal data in exports
- An upgrade base for merging future regenerations
For now, every export uses a single programming language.
Who takes part in an evaluation.
Sponsors, designers, builders and operators each have a role.
Process owner
COO, business-unit heads
Rank processes by expected value
Solution architect
CTO, chief architect
Judge whether the design will hold up
AI builder
Applied AI and ML teams
Improve workers between test runs
Platform owner
Infrastructure and SRE
Keep exported workers running
Product lead
Product owners
Turn automation into a roadmap
Department analyst
Analysts, power users
Automate recurring team tasks
“Will we truly own it?” Test the claim.
Before download you see an itemized list of the package, plus a plain PDF summary to share with stakeholders. The package carries a signed bill of materials, meaning an inventory of everything in the software that your IT team can verify offline. It also holds a security analysis for your IT security team. A starter document set covers the documentation the EU AI Act requires (not a conformity assessment).
Once exported, a worker runs with nothing from Nxgyle: it has no license check, sends no usage data and never phones home. Hosting, updates and upkeep then sit with your IT team.
Can we lock Nxgyle out?
It has no way in. There are no Nxgyle keys inside, no license check and no phone-home. You hold the keys and the code, on infrastructure you select.
Which files do we get?
The complete software, ready for your IT team: code, tests, a container build file, a README and, if selected, deployment files and runbooks. It starts locally with one command, and your IT team runs any cloud deployment. No proprietary platform is needed.
What if Nxgyle ceased operating?
Exported workers have no run-time dependency on Nxgyle, so they keep running while your team maintains infrastructure, credentials and dependencies. Hosting the Studio itself on your premises requires a separate Nxgyle license, and an air-gapped install still needs private AI model endpoints.
Your evaluation team can ask for example test results, the README and a sanitized file listing via the Trust Center.
What you receive at delivery.
Counts reflect published catalog records, not customer deployments or measured business results. Each digital worker runs a defined business workflow under configured rules and approvals. Nx·Studio exports its source code, which your organization then operates and maintains.
Nx·Studio designs digital workers for the processes you select. Nx IDE gives your engineers extra capacity for the systems that surround each worker. The Cognitive Engine decides routine cases once each decision type passes its exam.