# An AI assistant that waits for your yes > Ask about your projects, hours and team in plain language. Every number names the record it came from, every change waits for you to confirm it, and your organisation decides which models answer and where they run. - **Sources under every answer** — A figure the assistant cannot point to a source for is never shown. - **Nothing changes on its own** — Every change it prepares waits on a confirmation card until you accept it. - **Your permissions, nothing more** — The assistant reads only what the person asking could open anyway. - **Runs on models you approve** — Use Akollo's models, your own provider key or a model server inside your network. ## AI help inside every Akollo screen Small, checkable help inside the screens people already use. - **Assistant** — Questions about your records, with steps, sources and the model named under each answer. - **Writing help** — Write from notes, expand, shorten, fix grammar or change tone in descriptions, comments and replies. - **Thread summaries** — Summarise a ticket or a work item's comments. Summaries are read-only and not stored. - **Plain-language search** — “My open items due this month” becomes ordinary search filters, run with your permissions. - **Staffing suggestions** — Up to three best fits plus a growth pick per item, ranked by Akollo, confirmed by a manager. - **Plans and breakdowns** — Draft a project plan or break a work item into children, then review and create. - **Access control** — Off, admins only, admins and managers, everyone, or people with a seat, plus a monthly limit per person. - **Pay-as-you-go AI** — Akollo models start at $14 per million tokens and get cheaper with volume. Your own key, model server or local model has no AI token charge. ## Press ⌘J and ask The assistant opens beside any page with ⌘J or Ctrl+J and knows where you are, so “What is at risk in this project?” just works. It shows the steps it took, lists its sources and names the model that wrote the answer. - /plan, /staff, /summary, /free and /explain commands - @ for people and # for projects - Up to 50 saved prompts per person - Missing-time hints and status-update drafts for your own work ## Plans and staffing suggestions a manager confirms Describe a goal and the assistant drafts epics, stories and tasks with estimate ranges. Ask “who should take this?” and get the best-fitting people from your own teams, with skill, availability, familiarity and workload behind each suggestion. Pay is never used. - Draft plans you edit before anything is created - Large plans need a second person's approval - Choosing someone else than the top pick asks for a reason ## Your data, your choice of model By default Akollo's models answer. An administrator can connect the organisation's own provider key, its own model server, or a server inside a closed network through the AI Connector. Data classified as restricted never goes to an outside provider. - Choose a model per job: everyday answers, planning, search - Personal data such as IBANs and ID numbers masked before the model sees it - Requests logged with who, which feature and when, but not the content - An emergency stop for every AI feature at once ## Five commands and one rule Type / for a command, @ for a person and # for a project. Whatever you ask, the answer names its sources. - **/plan a mobile app release** — Epics, stories and tasks with estimate ranges and dependencies. Nothing is created until you choose Create. - **/staff the API migration** — Up to three best-fitting people and one growth pick, each with the reasons. - **/free next week** — Free hours per person, with leave shown only as “on leave”. - **What should I do today?** — Reads your queue: approvals, assigned and overdue items, each linked to its list. One answer prepares at most one change, and that change waits for your Confirm. ## AI across every step The assistant drafts plans in Projects, suggests people from Resource planning, finds missing time in Time tracking and summarises in Reports, always with your permissions. ## Works with the rest of Akollo - [Work & projects](https://akollo.com/en/products/work) - [People & HR](https://akollo.com/en/products/people) - [On-premise & data control](https://akollo.com/en/solutions/on-premise) - [Security & trust](https://akollo.com/en/trust) ## Questions about AI assistant ### Can the assistant change data on its own? No. When you ask for a change it prepares one and shows a confirmation card with what would change. Nothing happens until you confirm, and one answer prepares at most one change. ### Does our data leave our organisation? Your organisation decides. Requests go only to the models it has approved, an administrator sets the highest data class an outside model may receive, and restricted data is answered only by a model inside your installation. On-premise & data residency covers the installation options. ### Does the AI judge or rank employees? No. It never infers health, personality, mood or intent, and it does not score people. Staffing suggestions are ranked by Akollo from skills and availability, and a manager decides. ### How are AI tokens counted? Each question and answer is measured by the text the model reads and writes, weighted by the model's price and converted into AI tokens. On Akollo's models the organisation's use is totalled at month end and billed from $14 per million tokens, less with volume; an organisation of 1,000 people pays about $160 in a typical month. Requests through your own provider key, your own model server or a local model carry no AI token charge. Frontier models are a separate add-on at $99 a month and off by default. ## Ask your own questions in a demo We show the assistant on your kind of projects and explain where its models can run in your setup. - [Book a demo](https://akollo.com/en/demo) - [Read the docs](https://docs.akollo.com/en/modules/ai) --- Source: https://akollo.com/en/products/ai · [Türkçe](https://akollo.com/tr/urunler/yapay-zeka-asistani), [Deutsch](https://akollo.com/de/produkte/ki-assistent)