AVRO - Bluefort

ARVO · BC Intelligence · Governed, Audit-Grade AI.

Months of Build.
Delivered in Days,
Governed & Audit-Grade.

Forward deployed engineering for BC, delivered with AI.

From one approved requirement, ARVO generates the design, AL, tests, UAT scripts and documentation in hours, with the evidence and approval receipt attached. Your team stays in control, shaping and reviewing the work in about five days instead of thirty.

Become a Founding Partner Today Book Your Walkthrough View on Marketplace

Join us at
Directions EMEA

27-29 October

Paris, France

Stand S33

~6×

build throughput per consultant, with the same team

30 → 5

consultant-days on a major customization

-83%

lower delivery cost per
project

7 Gates

on every build, with
no exceptions

Illustrative for a major Business Central customization of about thirty consultant-days. Actual results vary with scope and review depth.

What Actually Gets Faster.

The change is clearly measurable. On a typical mid-sized Business Central
customization, here is where the effort falls, workstream by workstream.

−92%

AL development

−92%

Testing and UAT scripts

100%

FDD, TDD & user guides, generated with the build

−67%

Project management, review and sign-off

−60%

Functional analysis and design

Thirty consultant-days become about five. The remaining five are the work that still needs human judgment, review and approval.

Illustrative for a major Business Central customization of about thirty consultant-days. Effort reduction only. Planning guide, not a quotation.

If the Build Gets Faster, the Evidence Has to Get Stronger.

Every build keeps the whole record: the requirement, the designs, the tests and their results, the person who approved it, and the exact app that was promoted. Roughly six weeks of delivery work — and the evidence a customer needs to sign it off.

Compiled Solution App

Tested against the agreed ladder

Functional & Technical Design

FDD and TDD, generated with the build

User & Persona Guides

Ready to hand straight to the client

Best-Practice Evidence

Against Microsoft’s own BC conventions

Test Evidence Pack

Compiler output, analyzers, AL tests, UAT scripts

Approval Receipt & Audit Trail

Append-only: who approved what, and when

Scale Clients. Not Consultants.

When a thirty-day build becomes five, the same team can serve more clients without adding delivery
headcount. Capacity stops setting the ceiling. Demand does. The question becomes how many projects you
can win, not how many you can staff.

~6× Build Throughput

Per consultant

Unlimited Seats

Across your delivery team

White-Label Evidence

under your brand

Your Azure

Bluefort never bills AI

Get in Touch

Not Another AI Code Generator.

Not This

A prompt that returns AI.

Another autonomous code generator.
A black box pointed at a live ERP.

This

Governed delivery automation,
with the proof attached.

Design, tests and approvals built into every build.
A delivery multiplier your team controls and your client can audit.

One Governed Lane, From
Approved Requirement to
Governed Evolution.

Every change follows the same seven stages, with a named gate at each step. Nothing reaches Production without approval, and governance continues after go-live.

1
Intake
Approved requirement
2
Design
FDD/TDD and tests drafted
3
Generate
AL built, compiled and analyzed
4
Test
Sandbox tests and UAT scripts
5
Approve
Human sign-off and receipt
6
Promote
Gated release to Production
7
Evolve
Regression-tested across BC release waves

Human-in-the-loop ApprovalAppend- Only Audit LogBusiness Central Role & Permission MappingKill SwitchConsumption Visibility

What Changes When
You Run ARVO.

One governed lane, intake to evolution

Every change follows seven stages, each with a named gate. Scope is locked before generation. Designs and tests are approved before code. Compile, test and sign-off must all be green before promotion. Two bounded repair loops fix code first against compiler diagnostics, then against the frozen test suite. If a loop reaches its cap, ARVO parks the run with full diagnostics for human triage. Nothing is forced through. After go-live, Evolve keeps the extension governed across Business Central release waves.

Substantiated by:

Seven stages, each with a named gate
Two bounded correction loops with hard caps
Human triage when a cap is reached
Scope, object ranges and target environment locked at intake
Evolve continues governance after go-live

The tests are written first, then frozen

If the same model writes the code and the test, it can mark its own homework. ARVO does not allow that. Test scripts are generated from the approved design in a separate model call, before any code exists. A person approves them, then they are frozen. From there, the tests stay fixed and the code must adapt. When a test fails, ARVO regenerates the code, never the test, and the full suite runs again after every repair.

Substantiated by:

Separate model call at design time
Human-approved, then frozen
Compiler first, frozen suite second
Full suite re-runs after every repair
Oscillation detection

Tested with the permissions your users actually have

A feature can work perfectly for the consultant who built it and still fail for the person who needs it. ARVO runs each scenario under the real Business Central permission sets mapped to that role. Positive cases prove the user can complete the job. Negative cases prove they cannot perform actions outside their role. Permission defects are caught before go-live, not on deployment day.

Substantiated by:

Persona scenarios under real Business Central permission sets
Positive and negative cases
Five named test levels, L0 through L4
Deterministic test data with lineage and cleanup
UAT scripts generated with the build

A production gate nobody can bypass, with a receipt to prove it

Promotion is enforced server-side. ARVO checks that every required gate is green, the right people approved, the entitlement is valid, and the package hash matches the exact release candidate that passed testing. Production receives that same approved app, not a rebuilt copy. ARVO Studio stays in the sandbox. Every model call, gate evaluation, approval and deployment is written to the append-only audit log and sealed into the evidence pack.

Substantiated by:

Server-side gate check, unreachable from any client or script
Package hash verified against the approved candidate
Sandbox-only by design
Append-only audit log and approval receipt
Tenant-wide kill switch

Your evidence. Your brand. Your Azure.

The evidence pack carries your brand, so your client receives the design, tests and approval trail as part of your delivery. AI runs on your own Azure subscription, where you can see and control consumption directly. Bluefort never resells or marks up AI capacity. That is part of the contract, not a pricing policy that can change later.

Substantiated by:

White-label, partner-branded evidence pack
Unlimited seats for your delivery team
AI on your own Azure subscription
Never resold or marked up by Bluefort
Consumption visible per build

What Changes When
You Run ARVO.

Group 1

The Governed Lane

Group 2

Testing & Evidence

Group 3

Trust & Data Boundaries

Group 4

Licensing & the Platform

Capability

What it Means in Practice

Intake and scope lock

Captures the approved requirement in Business Central and locks exclusions, object ranges, dependencies and the target environment before generation begins.

Grounded requirement analysis

Uses live capability evidence through the gateway, not model memory, and turns each requirement into a clear, decision-ready record.

Precedence ladder

Configure standard Business Central first, reuse existing IP second, generate new code last. The guardrail recommends before it builds.

Design generation

Drafts the FDD and TDD with confidence flags for named human review and recorded approval.

Live symbol acquisition

Pulls symbols from the target environment for every build, so compilation reflects the apps that environment actually runs.

Pinned compile and analyzers

Runs a pinned compiler, CodeCop and analyzers, then stores structured diagnostics as evidence.

Bounded repair loops

Repairs code first against compiler diagnostics, then against the frozen test suite. Both loops are capped; on cap, the run parks for human triage.

Gated promotion

Promotes only after server-side checks and package-hash verification, through the Business Central Admin Center API.

Evolve

Regenerates and regression-tests the extension across Business Central release waves, with documentation refreshed as part of the run.

Capability

What it Means in Practice

Five test levels

L0 compile and analyzers · L1 unit and integration · L2 persona scenarios · L3 human code review by risk tier · L4 acceptance testing with sign-off.

The frozen test oracle

Created from the approved design in a separate model call, approved by a person, then locked for the life of the build.

Data factories

Creates deterministic test data with lineage and cleanup, so test runs are repeatable instead of dependent on whatever is already in the sandbox.

Permission-true persona testing

Runs scenarios under the real Business Central permission sets held by each role, including both positive and negative cases.

UAT script generation

Generates acceptance scripts with the build, assigns them to named persona users and captures sign-off in place.

Recorded approvals

Captures design, technical design, review waiver, acceptance and promotion against named approval authorities.

Evidence pack

Assembles the compiled app, design documents, persona guides, best-practice evidence, test results and approval receipt as the build runs.

White-label branding

Puts your brand on the evidence pack and client-facing documentation, not Bluefort’s.

Capability

What it Means in Practice

What Bluefort holds

Bluefort holds the code-signing certificate in an HSM-backed key vault and the entitlement-token signing key. That is the complete list.

What Bluefort never holds

No passwords, Business Central credentials, refresh tokens, business data or binaries.

Session-scoped identity

Consultants sign in with their own delegated identity through a public-client flow. No stored secret leaves the session behind.

Per-operation tokens

Each environment operation mints a token, uses it and discards it.

Sandbox-only by design

Enforced in code. Production is a deployment target, never a test target.

Consent as the kill switch

Your administrator grants environment access in Microsoft Entra and can revoke it at any time. Revocation stops access.

EU hosting and residency

Bluefort’s runtime is hosted in the European Union. For EU customers, inference defaults to EU Data Zone deployments.

Your artifacts, your storage

Designs, source, packages and evidence stay in your own storage. Bluefort keeps metadata and hashes.

Capability

What it Means in Practice

Bring your own AI

Inference runs on your own Azure subscription and Microsoft bills you directly. Bluefort never resells or marks up AI capacity.

Consumption visibility

See inference consumption by build and by environment, so AI spend maps to the work that produced it.

Entitlement at the gate

A valid, signed, short-lived entitlement token is required at promotion, when the build reaches the point of value.

Microsoft platform

ARVO consumes Entra ID, the Business Central development endpoint, the Admin Center and Automation APIs, and Microsoft Learn grounding rather than rebuilding them.

ARVO Studio on AppSource

A free “Get it now” listing. Install ARVO Studio into a sandbox and activate in-app; licensing and billing run directly with Bluefort.

Partner Edition

€1,800 per month, including four customer environments and unlimited seats. Additional environments are €500 per month. Annual term.

Direct licensing

€1,000 per environment per month, self-serve and activated in ARVO Studio. The direct route never undercuts the partner route.

Twenty Places. First Come,
First Served.

ARVO is live today. At Directions EMEA we put it on stage in front of 3,200 attendees. Become a Founding Partner today.

Founding partners get more than early access: they full onboarding kit, priority onboarding with dedicated partner support through the first projects, and a founding rate held well beyond launch — locked in early before public pricing moves.

When the twenty are taken, the program closes — whether or not we have reached Paris.

Apply for a Founding Place

Native Business Central extensionCertified for Microsoft Dynamics 365EU-hosted runtimeSandbox-only by designYour AI on your own Azure

Install It. Bring in a Real Requirement. Check the Evidence.

Run a real customer requirement through ARVO’s governed lane and inspect what comes out: design, code, tests, evidence and approvals. Every claim on this page is built to be checked.

Show Me View on Marketplace