This week, more work can enter OKiDO without someone copying it from a form or meeting note. Request forms turn submissions into tasks, and a new Fathom connection brings recorded meeting action items into a review queue.
We also made the rules around that work clearer. You can control what a board column requires, see where approvals stand, and inspect the AI work behind a task.
Turn requests into ready-to-work tasks
You can build a request form from the fields of one or more smart labels and share it internally or with clients and partners. Each submission becomes a task in your chosen column, with its answers on the labels, its submitter and form recorded, and notifications for the form and project owners. Public form links are available on paid plans.
Set forms up under Project settings, Request forms. A new Forms tab puts every project form in one place, with Fill in and Copy link for everyone and New form and Edit for project owners. The board also shows each form beneath Add Task in the column where its submissions land. You can give incoming requests an assignee, a team and a priority.
Other board controls help your team keep work ready for its next step:
Set column rules for required task fields, labels or smart-label fields, and choose who may move tasks into a column. An invalid move is undone with an explanation. The field picker now groups label fields in a shorter, easier-to-use list.
Choose default project labels, or create one with fields, under Project settings, Labels. New tasks start with those labels and empty fields; you can remove them from individual tasks.
Move a task to another organization by choosing its destination project and stage. Before you confirm, you will see who would lose access and which labels, sprint, milestone or dependencies will not carry over. Non-member followers stop following it, and a non-member creator no longer receives its completion or approval notifications.
Moving a task to another project now takes its subtasks into the same stage instead of leaving them behind. The Blocked by picker also finds subtasks at any depth, shows their parent tasks, keeps completed tasks separate unless requested, and reports search failures clearly.
Plan launches and make approvals count
Project templates can now count backward from a launch or end date. Give tasks offsets such as 10 days before launch, then supply the date when creating the project. Milestones are marked at risk when estimated remaining work and dependencies no longer fit before their date, or late when the date passes with work open. You can spot this on milestone and project views, filter for at-risk milestones, and the milestone and project owners are notified.
You can request approval from several people at once without changing the task assignee. Every requested person must decide, can leave a note, and can be asked again after changes. You cannot approve your own request or mark the task done another way while approval is pending. Approval states are visible on boards, lists, Sprints and Milestones; My tasks also shows decisions waiting for you. Approving an approval task now completes it and releases the work waiting on it.
SOP gates now hold later steps, AI agents and automations until the required number of people approve. Approvers can approve, request changes or reject on the run page. Agents can request task approval, but cannot decide it. They also wait for a person before sending email or writing to a connected app, including for existing agents, and cannot take outside action on a task blocked by another task. Reading connected apps needs no approval; admins can exempt specific write tools for an agent.
Bring meeting follow-ups into the right project
Connect Fathom under Settings > Meetings to review action items from recorded meetings. A top-bar Meetings icon flags items for you; on a meeting page, you can accept one into a project in any of your organizations or dismiss it, with a link to its moment in the recording. One Fathom account works across your organizations because the organization is chosen with the project. You can also import meetings from the last 7, 14, 30 or 90 days; hand-added items from older meetings wait for review rather than becoming tasks immediately.
Items assigned to colleagues appear in their review lists, while items for people outside your organizations become follow-ups for you. Hand-added items in new meetings become tasks directly. If several colleagues recorded one meeting, OKiDO merges the recordings and shows each item once. Checking for repeated items and suggesting projects uses a small amount of the organization's AI tokens. Imports also work when your Fathom and OKiDO email addresses differ: OKiDO identifies your Fathom account by name or asks you, and recognizes colleagues invited under their Fathom addresses.
Meeting pages now include Fathom's summary, bookmarked moments and a transcript that only attendees can open. Turn a highlight or transcript line into an action item with one click, or write your own. Search finds meeting titles and summaries, and your AI assistant can find meetings and accept action items in OKiDO chat or over MCP, but can reach a meeting only when an attendee asks. Settings lists other planned note-takers and lets you tell us which to support next. If Fathom access stops working, a daily connection check triggers a notification and a Reconnect button.
See what AI did, and control what it does next
Every task now has an always-visible AI work tab beside Comments and Activity. It shows task-specific agent runs, produced files and who approved or rejected agent actions, with a count when work exists and an explanation when it does not. This includes work from client runners and links earlier runs to their tasks. The Activity tab no longer mistakes recent runs from the whole project for runs on your task.
New MCP tools let assistants list, create, update and submit request forms, and create projects from templates. They can fill smart-label fields with add_label, read task AI work with get_task, retrieve milestone risk status and set column rules through stage tools.
You can import skills from a zip on the Skills page. Your browser checks it before upload, finds folders containing SKILL.md, lets you select which skills to import, and lists skipped scripts, programs and binary files. GitHub imports likewise bring in text files rather than executable or binary files. Refused skill files now explain why, long SKILL.md imports work, duplicate names and versions get a clear message, and descriptions have two lines when you add a skill to an agent.
AI token reporting now includes memory searches, saves and nightly duplicate cleanup. Agent runs, task-comment Copilot replies and Capability Factory jobs count tokens used before a failure; coding jobs count every fix or fallback attempt using measured tokens. Agents on the default OpenAI Sol 6.1 model work again, and agents and chats using reasoning models now call tools in the way each model requires while retaining reasoning between steps.
Smoother sprints, onboarding and everyday details
Dragging a card between others in a sprint column now joins the current sprint, records the move in activity and wakes agents watching task changes. Tasks join their own project's or team's current sprint, not another sprint in the organization. Closing a sprint immediately updates its status and remaining-task count, and the Sprints page opens the next sprint. Human-worked runs show In progress after their first completed step.
New accounts get video-led onboarding with Docs links and steps that complete in any order as you do them. You can skip steps you do not need, and an unfinished step tells you what is missing. The journey respects plan limits: free accounts see the PRO trial before project and team setup, paid customers do not see a trial step, and plan details are under Organization settings, Plans. New organizations no longer arrive with 20 Playbook pages consuming the Starter document allowance.
Your signup page or browser language now carries through to your account, onboarding and initial organization name; founding customers use the language of their purchase page. Dates, times and notifications, including Inbox, browser and email notifications, follow your language setting. Existing notifications keep their original text. Signed-out pages and account-error messages now use your browser language too.
We also tightened translations across the app: organization roles, search results and filters, run variable types, recurring schedule suggestions, sprint-meeting note headings, public-form requester lines, audit events, dashboard activity, Google Tasks confirmation and new AI chat names. AI action names, tool impact labels, built-in skills, Copilot text and approval-wait messages are translated, while MCP catalog app descriptions remain as their makers wrote them. German approval terms and daily standup wording, French approval and SOP terms, and long task-status layouts are now consistent. Approval buttons line up, and a new skill's description is easier to read.
Several smaller fixes remove interruptions. Recordings and task briefings save again, and stored briefings play in tasks. The Playbook review bar stays visible after saving or acting, and Compare Versions wraps long lines. Decision trees show the answer you chose and the tree's name. Search prioritizes title matches, even for older tasks. Accented filenames survive upload, the Theme preference reflects and immediately applies your choice, and run names are clearer. Completed-run notifications reach your Inbox, system-run approval notifications open the right run, and escalation titles and sprint-note email subjects no longer contain an unnecessary long dash.
Try a request form for your next intake process, or connect Fathom to review the follow-ups from a recent meeting. If your team runs work through approvals or launch milestones, use the new board and project signals to make the next decision easier to see.
