Find out whether AI can do the job.

A good fit when

• A buyer deciding whether a local AI deployment is technically and economically worth funding
• A valuable workflow constrained by data location, hardware, connectivity, power, latency, or provider dependence
• A team that wants visible demo outputs and failures on its own target or an approved reference machine
• An organisation prepared to accept deploy, change, remediate, rescope, or stop

Another path is needed when

• Autonomous or safety-authoritative decisions, live OT/SCADA/control integration, or public production exposure
• Classified, export-controlled, patient-identifiable, credential, or otherwise unauthorised material
• A request for regulatory, legal, privacy, security, safety, or engineering certification
• An exact field device that cannot be accessed or qualified inside the agreed scope

• One repeated, valuable, non-safety-authoritative workflow and its economic owner
• One customer-controlled target host or approved reference machine
• Three to ten representative, lawful, non-sensitive test cases
• Known data, network, connectivity, output, and prohibited-use boundaries
• Permission to retain enough non-sensitive evidence to reconstruct the decision

Start requirements

One repeated non-safety-authoritative workflow, one decision owner, an approved target or reference host, three to ten lawful non-sensitive cases, clear data/network/output limits, and a buyer review inside the delivery window are required before work starts.

The steps we follow

1. Define the workflow, current alternative, decision owner, consequence, and stop rule.
2. Inspect data rights, network, host, memory, storage, power visibility, software, and prohibited uses.
3. Run the local preflight and record the deployability envelope, unavailable measurements, and blockers.
4. Demonstrate the practical ceiling on approved customer-relevant cases and retain failures.
5. Recommend staged deployment, remediation, a model/hardware change, rescope, or stop.

• The decision owner confirms that the report accurately records the tested target, approved boundary, demonstration, limitations, and recommendation.
• The local demonstration pack preserves the agreed cases, outputs, failures, target identity, and stated measurement limits.
• The hardware and runtime envelope distinguishes measured findings from unresolved questions and unavailable telemetry.
• The customer selects proceed, remediate, change model or hardware, rescope, or stop.

AI System Fit Audit & Local Demo · NF-A0

Audit and local demonstration only. No production deployment, full-corpus ingestion, bespoke integration, continual learning, certification, penetration test, production SLA, or guarantee of accuracy, savings, deployability, or ERAIS performance.

What this engagement establishes

The offer is a founder-led market test with no observed demand or delivery history. An audit or demonstration does not establish production readiness, broad device support, customer traction, or general ERAIS quality or efficiency.