Meet Ask Sentry: AI that drafts, people decide
Ask Sentry joins FlowSentry's AI assistants, all of which draft and explain while people keep every decision.
tocOn this page
AI can save investigators hours of reading. It should never decide whether a customer is reported, cleared or approved. Every assistant in FlowSentry is built on that line: it drafts, summarises and explains, and a named person takes every decision. This release adds Ask Sentry, a copilot available on every page, and brings all of the assistants under one set of labels and controls.
Meet Ask Sentry
Press Ctrl+J on Windows or Linux, ⌘J on a Mac, or select the sparkle button in the app bar. A drawer opens beside the page you are on.
Ask it about the case, customer, screening hit or rule in front of you, or about records you are allowed to see. Answers come with Sources: links to the records it used, so you can check every statement. Start afresh with New conversation.
Ask Sentry carries a Read-only badge, and it means it. If you ask it to approve, close or file something, it declines and gives you a link to the screen where you can do it yourself. Its tools check your own permissions first, so it can never show you a record your role cannot open.
Every assistant, in one place
| Assistant | Where you find it | What you get |
|---|---|---|
| Alert triage | Alert drawer → Summarise with AI | Why it fired, risk context and a suggestion; Use as rationale fills the rationale and reason code |
| Sentry Agent memo | Case → Overview → Investigate with AI | Summary, cited evidence, typology, recommendation and confidence |
| Case summary | Case → Overview → Summarise case | Key evidence, open questions, next steps and a handover note |
| STR narrative | Case → Suspicion & STR → Draft narrative with AI | A Who/What/When/Where/Why/How draft; Use draft fills the narrative for you to edit |
| Match explainer | Screening hit → Explain this match | What agrees and differs; Use narrative fills the decision narrative only |
| Risk explainer | Customer → risk tab → Explain this rating | What drives the rating and what would change it |
| KYC review | KYC changes → Review with AI | What changes, the risk and what to check before approving |
| Ask Sentry | Sparkle button or Ctrl/⌘J | Read-only answers with sources |
| File analysis advisor | File analysis | Mapping suggestions and a findings memo draft |
Labelled, every time
Every draft is labelled. AI draft — advisory comes with a confidence. If AI is switched off, times out, cannot be reached or returns an answer that fails validation, you get Generated without AI — rules-based summary, with the reason. Your work never stops because a model is unavailable. Discard removes any draft, and nothing is ever submitted automatically.
When you add a draft to case notes, the note starts with "[AI draft (advisory), added by the investigator]", so a reader always knows where the text came from.
How your data is protected
- Pseudonymisation. Names, national IDs, phone numbers, account numbers, e-mail addresses, KRA PINs and long digit runs are replaced by tokens before any model call, and restored only on your screen.
- Less is sent where less is needed. The match explainer sends only whether a date of birth or ID agrees, not the values. The risk explainer sends nothing identifying. KYC review sends only whether an identifier was added, replaced or removed.
- Self-hosted for real client data. A self-hosted model tier keeps processing in your environment. Institutions holding real client data are pinned to that tier, or receive rules-based drafts. Cloud tiers are used only for synthetic and demonstration data.
- Every call is logged for examiners: the feature, prompt version, model, tokens, cost, latency, outcome and the person who asked. What you did with each draft (used, edited, added to notes or discarded) is recorded too.
Kill switches
Your administrators can turn AI features off for the whole institution from Tenant settings, with a reason and a fresh second factor. The service operator can also switch off any single assistant. In both cases, the assistants fall back to rules-based drafts straight away.
What AI never does
No assistant can close an alert, decide a screening hit, file an STR, approve a KYC change, edit a list, or change or activate a rule. Those actions stay behind your institution's roles, maker-checker and step-up.
How to try it
- Open any case and press Ctrl+J (or ⌘J).
- Ask "What are the main risk indicators on this case?" and follow a link under Sources.
- Ask it to close the case, and see it point you to the right screen instead.
- On the case Overview, select Investigate with AI and read the cited evidence.