Talk to Elise, and see what she can do
Chat with Elise from anywhere, ground a question on a ticket, teach her a preference, rename her, and read what she is allowed to do and what she is costing.
What you will have
- A message answered from the client or ticket it names, without opening that record first.
- A personal preference Elise remembers across every chat, plus a sign-off she now uses.
- Elise opened from a page other than her own, with a prompt already written for that page.
- The assistant renamed from Settings.
- One destructive action and one safe action compared side by side in the Actions catalogue.
- This month's API calls, tokens, and cost, broken down by feature.
Why it works this way
Elise's own tabs live in the left sidebar under Elise, not in a strip across the top of the page - the sidebar is the only place to switch between Chat, Dashboard, Action History, Actions, Agent Runs, On Demand Runs, and Review.
A personal memory and a voice preference are saved by asking in plain language; neither has a settings page of its own.
The Ask Elise button seen on other pages opens the same floating widget as the sidebar's Chat, pre-filled with a starter question for wherever you clicked it from.
Steps
Open Elise in the left sidebar.
Elise's own pages are not a strip of tabs across the top - they are the sub-items listed under Elise in the sidebar: Chat, Dashboard, Action History, Actions, Agent Runs, On Demand Runs, and Review. Opening Elise lands on Chat.
Open Elise in the left sidebar. Type a ticket number into the message box, such as #24.
Elise resolves the number and answers from that ticket's own record, with a link back to it. The same works for a client's name, or for a ticket or client already open when you started the chat.
Note: The figures Elise states here (ticket number, subject, client, and who fixed it) match that ticket's own record; they are not invented.Type a ticket number into the message box, such as #24. Ask Elise to remember a preference and to use a sign-off.
Elise saves both a personal memory and a reply sign-off and confirms what it now holds. These carry into every chat thread, general or ticket-scoped, and into the floating widget on other pages.
Note: The first attempt at this on this tenant answered that the save had failed; asking again in the same wording saved both items. Treat a failed save as worth a retry, and report it if it repeats.Ask Elise to remember a preference and to use a sign-off. On a different page, click Ask Elise.
The button appears in card and page headers away from Elise's own section - here, the Workflow Library. It opens the same floating chat, pre-filled with a starter question for the page you clicked it from, and keeps whatever Elise already remembers about you.
On a different page, click Ask Elise. Go to Settings > AI Assistant > Identity to change the name.
The AI Assistant Name field is the one place the name is set. The new name flows into the chat widget, the sidebar's pages, and anything Elise generates; the sidebar's own AI Assistant label stays fixed.
Go to Settings > AI Assistant > Identity to change the name. Open Elise > Actions and search for a destructive capability.
Every capability lists its authority level next to its name. A Destructive one, like deleting every item on a ticket, always carries an amber warning that it needs an explicit confirmation before it runs.
Open Elise > Actions and search for a destructive capability. Search for a Safe Write capability to compare.
Add or remove ticket tags carries a Safe Write badge instead: it runs on its own, no extra confirmation. Every one of Elise's capabilities is catalogued this way, with its permission and what it needs from the conversation, searchable from the box at the top.
Note: This is the same capability behind a tag request typed in Chat; it never creates a tag that does not already exist.Search for a Safe Write capability to compare. Open Elise > Dashboard to see this month's usage.
Four figures cover the current calendar month: API calls, tokens used, cost, and average cost per call. The figures are read-only and reset with the calendar month.
Open Elise > Dashboard to see this month's usage. Scroll to Usage by Feature for the breakdown.
Each feature (chat, ticket triage, awareness training, KB drafting, and more) gets its own row: calls, tokens, a cache hit rate, and cost. A feature with no cache hits at all shows a dash instead of 0%.
Scroll to Usage by Feature for the breakdown.
Other ways to do this
Ticket-scoped chat
Open a ticket and use its own Ask Elise (about this ticket) button, or its thread in the Chat rail.
If it did not work
- If Elise answers that nothing is stored after a remember request, ask again in the same wording before assuming the feature is off - a save can fail once and succeed on retry.
- If a capability's tier is not what you expected, its card in Actions names the exact permission and context it needs; a missing one is why it refuses rather than runs.
Questions this page answers
How does Elise know which client or ticket I mean?
If you opened chat from a ticket or client detail page, Elise already has that context. Otherwise type the name and Elise will resolve it via search. You can also paste a ticket number (#1234) or a client name to ground the conversation.
Does Elise remember things across chats?
Yes. Elise keeps two kinds of memory that persist across all your chat threads: personal memories tied to you (your preferences and working style) and client memories tied to the client the chat is currently bound to. Just tell Elise to remember a fact, ask her to list what she's remembering, or tell her to forget one. Each kind is capped at 50 active memories to keep her focused, so once you hit 50 a new memory is politely declined with a prompt to forget one first, you choose what to drop. Nothing ages out or is discarded automatically, and forgetting is reversible.
Can Elise write replies in my voice?
Yes, Elise reads your voice blueprint (greeting, sign-off, tone notes, and sample phrases) when drafting ticket replies, closure summaries, or client emails. If you switch on auto-learn, Elise adds your sent replies to your sample phrases (kept to the most recent 50) so it sounds more like you over time. There's no settings page for the blueprint: just ask the assistant to update your greeting, sign-off, or tone and it takes effect on the spot.
How do I change the AI assistant name?
Go to Settings → AI Assistant → Identity and update the Assistant Name. The new name flows into the assistant's chat widget, its generated replies, and the page titles and tab descriptions that reference it. The left sidebar section keeps its fixed AI Assistant label.
What are action tiers and why does Elise sometimes ask twice?
Each capability has a tier: read (no confirm), draft (returns text for review), safe_write (one confirmation, reversible), sensitive_write (double confirmation), destructive (dangerous-mode required + double confirm). Higher tiers ask more so a misclick or an Elise misunderstanding can't take an irreversible action.
What is the Actions reference page for?
A catalogue of every capability Elise can invoke, grouped by category. Each entry shows its name, a description, its authority level (Read, Draft, Safe Write, Sensitive Write, or Destructive), the permission it requires, and the context it needs (a ticket, a client, or a user). Search the top box to filter across every action. Useful for checking what Elise can touch and why a capability isn't available in a given context.
What are the action tiers?
Every action carries one authority level. Read runs immediately (nothing changes). Safe Write runs on its own (Elise writes; an admin can tighten it). Sensitive Write proposes via chat and waits for your one-tap Approve. Draft prepares an artifact for you to apply by hand. Destructive is gated: it needs an explicit confirm plus a typed reason. These map to the badges on this page, the cards in Chat, and the queue in Review > Gated.
What can I ask Elise to do?
Elise has 60+ capabilities across tickets (close, assign, reply, escalate, create), clients (sync M365/Google, draft emails), billing (push to QB, send estimates), docs (draft KB, link assets), and team voice. Visit /admin/elise/actions for the full searchable list. Capabilities are tier-gated - destructive ones require dangerous-mode + double confirmation.
What does the Elise Dashboard show?
The dashboard summarises this month's AI usage and cost: total API calls, tokens used, cost this month, and average cost per call across the top. Below, two tables break the spend down by feature (Ticket Triage, Documentation, Script Proposals, and so on, each with a cache hit rate) and by client, so you can see where the assistant's usage and cost are going. It covers the current calendar month only, and the figures are read-only.
What does AI token usage tell me?
Token usage tracks the input + output tokens spent per Claude call, broken down by rung (chat, draft, classify, etc.). Spikes usually mean either a long-running agent loop or an unusually large context payload. Per-rung cost is calculated from the active provider's rate table.
What is the Elise page?
Your AI assistant's home page opens on Chat. Its pages sit in a group on the left sidebar: Chat, Dashboard (this month's usage and cost), Action History, Actions, Agent Runs, On Demand Runs, and Review. Voice, tone, prompts, and model choices moved to Settings → AI Assistant.
How much is Elise costing in API tokens?
Visit /admin/elise/dashboard for this month's token usage and cost. Per-feature breakdown shows which capabilities (triage, KB drafts, draft replies) consume the most. Cache hit rate per feature is also shown - high hit rates mean prompt caching is keeping cost down. The dashboard covers the current calendar month only; there is no daily or weekly view and no projected cost.
Was this helpful?