All demos
Interactive Browser Simulation — Cross-Vendor Evidence Composition

Composed Evidence: Closing the Assistant/ERP Audit Gap

A business action can span two systems: a universal AI assistant that takes the request, and an ERP vendor's own execution agent that carries it out. Each one logs only its own half. Send the request below, watch each vendor's log populate on its own, then try to answer a real audit question from each — and see what only a composed record can actually prove. Illustrative composite scenario, not a real client engagement.

Evidence typeBrowser simulation — Synthetic records, local state — no live ERP connection.

The request

Priya Raman, a sales operations analyst, sends this to a universal AI assistant in an ordinary chat session.

Priya R. — chat-88231

“Finalize the material configuration on order SO-4471 now that engineering signed off.”

The real audit question

“Who authorized this, and does the system record prove it matches what SAP actually executed?”

Send the request above, then use each panel's “Try answering from here alone” button below to test it against that vendor's log — before looking at the composed record.

Panel 1

Assistant vendor's own log

Conversation and intent only — no transaction detail, no proof it actually happened correctly in SAP.

Send the request to populate this panel.

Panel 2

ERP's own log

Transaction record only — no proof of who asked or why, no natural-language context.

Populates once the ERP executes the write.

Panel 3

Tioga's composed record

Both halves stitched into one attributable, queryable record — who asked, what was asked, what policy tier, what the ERP actually did.

Composes once both vendor logs exist.

Grounded in a real, vendor-acknowledged gap: SAP's own June 2026 reference architecture for third-party agent access states that the MCP protocol specification is still maturing, with security, identity and governance requirements for enterprise deployments “not yet fully addressed,” and assigns the controls for custom or third-party MCP servers to the customer. This demo dramatizes that gap with a single composite scenario: a universal AI assistant (Claude/ChatGPT-class) that logs conversation and intent, and an ERP vendor's own execution agent that logs the transaction under a shared integration account — neither composes the other's half into one attributable record on its own. Everything on this page runs in your browser; nothing is sent to a server. Illustrative composite scenario grounded in a real, documented gap — not a real client engagement.

Want the full picture for your environment?

A 20-minute intro call puts you in touch with the person who builds these engagements — not a form, a conversation.

Start a conversation