Question 1

What is exnatonAI?

Accepted Answer

exnatonAI is the agent layer on top of the Exnaton platform: a library of agents scoped to concrete utility workflows — billing edge cases, regulatory change, market communication, master data synchronisation, and energy and market data analysis. Wattson coordinates them; all of them read from the same energy data backbone that powers our tariff, community and prosumer modules.

Question 2

Does exnatonAI replace SAP or powercloud?

Accepted Answer

No. Exnaton is a modular add-on that integrates via API into any ERP, CRM, core system or national energy data hub — explicitly not an ERP replacement. The SAP Joule integration exists so agents can work across both environments rather than migrate anything out of them.

Question 3

Can we use our own agents alongside yours?

Accepted Answer

Yes — that is what the interoperability layer is for. Agents you build get governed API access to energy data, and the Joule connector lets agents in your SAP environment and ours coordinate over agent-to-agent protocols.

Question 4

Where is our data processed?

Accepted Answer

On EU-hosted infrastructure, on a platform certified to ISO/IEC 27001:2022 by Bureau Veritas. Customer data is not used to train models. Written hosting and processing documentation is available for your legal and information-security review — several customers have run exactly that review before activation.

Question 5

How are these agents classified under the EU AI Act?

Accepted Answer

They perform commercial, advisory, billing and administrative functions and are not safety components in the management or operation of electricity supply, which places them in the limited-risk tier: transparency obligations under Article 50 and AI literacy under Article 4 apply; high-risk obligations such as conformity assessment and CE marking do not. Classification is re-run per agent and per version — an agent that dispatches grid-relevant load would be assessed separately. This is our working baseline, not legal advice.

Question 6

Can an agent get a bill wrong?

Accepted Answer

Billing and settlement remain deterministic, auditable algorithms. Agents explain, detect and propose — they do not compute the bill. That separation is deliberate: anything physics- or settlement-related has to be exactly right, not approximately right, and a language model is the wrong tool for it.

Question 7

How do we start?

Accepted Answer

The Bill Explainer needs nothing from your systems — upload a bill and see the output. Wattson activates on an existing Exnaton tenant. Beyond that, our discussion paper's recommendation holds: consolidate the data foundation first, then pick two or three workflows where determinism and ROI are provable, rather than multiplying pilots.