Use cases · Real-time learning
A policy changes mid-shift.
The AI adapts in seconds — and tells you why.
A working test of genuine, real-time learning: drop a company document into Senua AI, ask a question, then drop in a contradictory update. No retraining, no re-indexing, no cloud — the answer changes on the spot, grounded in the exact line it read.
The scenario
“Senua Air” changes its travel policy without warning.
A fictional airline, Senua Air, has a standard travel-booking policy. Finance issues a one-page memo: flights under five hours must be booked in economy. Later that day, Operations issues an emergency amendment: flights departing Gatwick are upgraded to business, bypassing the duration limit. Both are ordinary internal memos — the kind a real company circulates as PDFs.
The test is deliberately adversarial. Anyone can summarise a document. To prove real learning, you introduce an instant logic shift: teach a rule, rely on it, then contradict it — and see whether the system genuinely re-reasons or just parrots the first thing it read.
How Senua AI handles it, end to end
Everything below happens live, on the device, through the Console — no model retraining and no data leaving the machine.
1 · Learn the memo
An admin drops the travel-policy PDF into the Console’s Knowledge → Learn workflow. Senua AI reads the document, comprehends the governing sentence, and turns it into a rule it can apply — no templates, no hand-written parsers.
2 · Apply the rule
Ask “what class for a 3-hour flight?” and it answers economy — and shows the exact policy line it used. Ask about an 8-hour flight and it doesn’t force a match; it only answers what the rule actually covers.
3 · Adapt in real time
Drop in the amendment PDF. Immediately, a 2-hour flight from Gatwick becomes business — the newer, more specific rule wins — while other flights still return economy. The system explains the conflict and which memo decided it.
Why it’s hard — and why this matters
This is learning, not retrieval.
No retraining, no re-indexing.
The update takes effect the moment the memo is ingested. There is no training run, no fine-tune, no vector re-index — the new knowledge is live in seconds.
It resolves the conflict itself.
Two rules now disagree for a Gatwick flight. Senua AI doesn’t need a hand-coded priority table: the more specific, freshly-asserted rule wins, and it says so.
It grounds every answer.
Each answer carries the sentence that produced it. When nothing it has learned applies, it says so rather than guessing — it doesn’t hallucinate.
Private and on ordinary hardware.
The documents and the reasoning stay on the machine. No GPU, no cloud, no per-token bill — the whole run happens on an everyday computer.
The documents used
The two Senua Air memos from the recording.
These are the exact PDFs ingested in the video. Download them to follow along or try the same flow yourself.
See it on your own documents.
This is one of an ongoing series of real-world use cases. If your team needs AI that learns from your files, privately, and keeps up with change — let’s talk.