Project Context That Drafts Itself
Find the right Basecamp-style workspace, read message-board threads, propose check-in responses and Hill Chart positions from real activity — and commit only after the human approves. Every AI proposal is a draft until the user clicks confirm; every direct write is gated by the same confirmation.
Why Project Workspaces Drift
Workspace Agent attacks all four — AI-drafted check-ins, AI-drafted Hill Chart positions, thread summaries, client digests with auto-redaction — and gates every commit on explicit human approval.
Recurring Check-Ins Become Throw-Away Replies
"What did you ship this week?" gets answered with three words on Friday at 4:55pm because nobody remembers what they did Monday. The whole point of the check-in — visibility — collapses.
Hill Charts Sit At The Same Position For Weeks
The scope hasn't moved on the chart since the kickoff. Nobody nudges the dot because nobody wants to be wrong, and the chart that was supposed to show uphill / downhill momentum stops showing anything.
Long Message Threads Out-Live Anyone's Memory
The 28-comment thread from three weeks ago had the answer. Nobody re-reads it, so the same question gets asked again, and a new 12-comment thread starts under the old one.
Client Updates Take An Hour To Hand-Compile
Every Friday someone summarizes the week's posts, redacts the internal-only items, and emails a digest. Skip a week and the client loses visibility; do it every week and someone spends an hour they don't have.
Decisions Get Made In Chat And Never Make It Into The Workspace
The team agreed in Slack to ship the staging migration before the rewrite. The workspace message board doesn't know. A new contributor reads the board and starts on the rewrite. The decision existed; the record didn't follow.
New Team Members Spend A Day Catching Up On Project Context
A designer joins the project mid-flight. The kickoff doc is six weeks old, the message board has 180 posts, and the Hill Chart looks frozen. "Where do we stand?" is a one-hour Loom from the lead — or it would be, if the lead had the hour. A drafted onboarding summary from the actual workspace activity would replace the Loom.
Workspace Agent At A Glance
Workspace AI
Drafted check-ins, hill charts, and digests — committed only by humans.
Inside Workspace Agent — The Actual Capabilities
21 tools across discovery, AI drafting, and gated writes. The agent finds and reads workspaces (read-only), proposes check-ins and Hill Chart positions as drafts (no commit), and writes through 12 confirmation-gated tools. Every AI proposal is reviewed by a human before anything touches the record.
Find Workspaces, Threads, Tasks, And Check-Ins
Six read-only tools cover the workspace browse loop — list workspaces, get one workspace's current state, list message-board threads, list tasks (open / completed / overdue), search threads with semantic + ILIKE, and see active recurring check-ins.
- List workspaces via list_workspaces — active / archived / all, with role and member count.
- Get one workspace via get_workspace — by ID or slug, with recent messages and open tasks summary.
- Search threads via search_workspace_messages — hybrid semantic + ILIKE; handles paraphrased queries.
- Summarize a thread via summarize_workspace_thread — TL;DR of a long post + comments.
AI Drafts For Check-Ins, Hill Charts, And Tasks
Four tools draft content from the user's recent activity — recurring check-in responses, Hill Chart position moves, follow-up tasks from message threads — and return the draft without committing. The user reviews and then calls the matching commit tool to write.
- Draft a check-in response via draft_check_in_response — built from your completed tasks and posts this cycle; returns a draft, doesn't commit.
- Propose a Hill Chart position via propose_hill_chart_position — proposes a new 0–100 position based on completed work; doesn't commit.
- Suggest a follow-up task via suggest_next_task — reads a thread and proposes a task; doesn't create until create_workspace_task is called.
- Two-step pattern — every AI draft requires the user to call a separate commit tool, with confirmation, before anything persists.
Client Digest With Internal-Only Posts Auto-Redacted
Generate a weekly markdown digest of a workspace for an external client. Internal-only message-board posts are automatically excluded. The agent returns markdown — sending the digest is a separate action; the agent never emails on the user's behalf.
- Weekly client digest via generate_client_digest — markdown returned, not sent.
- {"Auto-redact internal_only — posts flagged internal_only" => "true are excluded by design."}
- Week-scoped — defaults to current week start; pass week_start for any other ISO date.
- Read-only output — sending happens separately; the agent never auto-emails.
Twelve Write Tools — All Confirmation-Gated
Create workspaces, add and remove members, post messages, create and complete tasks, set up check-ins and commit responses, create Hill Charts and commit positions, archive workspaces, customize templates. Every one is risky, every one requires explicit confirmation, every one captures audit context.
- 12 risky write tools — workspace lifecycle, membership, messaging, tasks, check-ins, Hill Charts, archive, template — all confirmation-gated.
- Commit tools follow drafts — commit_check_in_response and commit_hill_chart_position finalize AI drafts after human review.
- by_agent flag on Hill Chart commits — when the position came from an AI proposal, the commit emits an audit AppEvent.
- internal_only on posts — workspace messages can opt out of client digests at creation time.
Outcomes Teams Can Measure
The agent shifts three operations from "Friday afternoon scramble" to "drafted from activity, committed by humans" — check-ins, Hill Chart moves, and client digests.
- Check-in response rate — share of members responding by the cycle deadline.
- Check-in draft acceptance — share of AI drafts committed verbatim vs heavily edited.
- Hill Chart freshness — workspaces with positions updated in the last 14 days.
- Thread-summary usage — long threads summarized vs scrolled end-to-end.
- Client digest turnaround — minutes to generate vs hours to hand-compile.
Twelve Write Actions, Four Draft-Only — All Human-Committed
Workspace Agent has 21 tools. Five are read-only (list, get, list-messages, list-tasks, search, summarize-thread, get-check-ins). Four are AI drafts that do not persist anything — they return proposals. Twelve are confirmation-gated writes — including the commit tools that finalize AI drafts. The pattern is consistent: AI proposes, human commits, audit logs.
- 12 risky write tools — every direct write requires explicit confirmation.
- 4 AI-draft tools — draft_check_in_response, propose_hill_chart_position, suggest_next_task, generate_client_digest — return proposals; do not commit.
- Two-step commit pattern — AI drafts require a separate commit tool call (also gated) before anything persists.
- by_agent audit on commits — Hill Chart positions and tasks born from AI proposals are flagged in the audit log.
WHAT TEAMS TRY INSTEAD
The four alternatives — and none of them propose check-ins, Hill Chart positions, and tasks the human commits later
Project tools generate text. None of the alternatives produce a draft Hill Chart position from real workspace activity, route it through a separate commit, and tag the resulting record as agent-born in the audit log.
ChatGPT, Claude, or Copilot drafting a status update
Generic AI producing text the PM still has to paste
- Workspace Agent's draft_check_in_response reads the actual message-board activity — generic AI guesses
- propose_hill_chart_position reads task progress and check-ins to suggest a position — generic AI has no project state
- Two-step commit means the draft never becomes a record until the human approves — generic AI text isn't tied to any record
Basecamp AI features, Asana AI Intelligence, Microsoft Loop Copilot, ClickUp AI
Vendor PM-tool AI — separate seat, separate audit, no by_agent flag
- Reads workspaces alongside HRIS, recognition, schedule, and training — vendor PM AI sees only its project graph
- Frontline contributors without an Asana/ClickUp seat still collaborate in the same MangoApps app
- by_agent audit tag on commits — vendor tools rarely distinguish human-authored from AI-proposed updates
Custom check-in templates and project bots
A PM-ops team's no-code automations, then maintenance
- Already shipped — workspace model, message-board threads, Hill Chart, check-ins, task lifecycle, and audit log in place
- AI proposals are first-class objects the agent returns; templated bots can only paste the same prompt every time
- Two-step commit pattern is part of the framework — no custom approval workflow to build for every project type
"I'll write the check-in tonight"
The status quo — PMs writing status from memory
- Check-in drafted from actual activity in seconds — not an evening status writing exercise
- Hill Chart position proposed from task progress; PM clicks confirm or adjusts — no "guess where the work is"
- Client digests generated from the workspace's own message board — no separate weekly summary email to assemble
PLATFORM LEVERAGE
Workspace Agent inherits everything the platform already runs
A standalone PM tool has to plumb each of these. Workspace Agent gets them for free.
AI proposes, human commits
4 AI-draft tools (check-in, Hill Chart, next-task, client digest) return proposals that don't persist — a separate commit makes the change.
by_agent audit on commits
Hill Chart positions and tasks born from AI proposals carry a by_agent flag — auditors can trace which records started as AI drafts.
Reads real workspace activity
Drafts come from actual message-board threads, task progress, and check-in history — not a template the PM filled out.
Confirmation-gated direct writes
12 risky writes spanning workspace creation, member moves, and status changes — every one requires explicit confirmation.
Permission-aware
The agent only sees and proposes against workspaces the user already has access to — same model the Workspace app uses.
RubyLLM model tiering
Workspace lookups on small tier; check-in drafts on standard; client digests on standard. Per-workspace cost stays bounded.
INDUSTRY FIT
Industries where workspace-driven project work moves the most weight
Workspace Agent shines where projects are the unit of value and check-in cadence is operational, not ceremonial.
Professional Services
Engagement workspaces with Hill Charts proposed from task progress; client digests drafted from the workspace's own threads.
Technology
Squad workspaces with check-ins drafted from message-board activity; engineering leaders see Hill Chart movement without survey-style updates.
Marketing & Creative
Campaign workspaces with proposed Hill Chart positions on each deliverable; agency teams audit by_agent provenance on every status.
Construction & Trades
Project workspaces with safety and milestone check-ins drafted from punch list and field activity — no Friday status spreadsheet.
Manufacturing
NPI and CAPA workspaces with Hill Chart positions reflecting real task closure rate — engineering leadership sees the project shape.
Public Sector
Program workspaces with grant-reporting check-ins drafted from message-board activity; within the FedRAMP-eligible boundary.
WHY MANGOAPPS WINS
An embedded workspace agent beats a Basecamp AI, an Asana copilot, or a custom check-in bot on every axis
The argument PMs, finance, IT, and the contributor on the team all share — and the one a horizontal AI or a vendor PM AI structurally cannot answer.
Cheaper than the alternatives
No Basecamp AI add-on, no Asana Intelligence seat, no Loop Copilot license, no extra PM-ops headcount writing status updates.
More secure
12 confirmation-gated direct writes, 4 propose-only AI tools, by_agent audit on commits, every action logged. Stays inside the tenant.
Easier to deploy
Already deployed if Workspace is enabled. Turn the agent on and check-in drafts / Hill Chart proposals / client digests work the same day.
Easier to use
"Draft my Monday check-in" / "propose Hill Chart for the auth project" — one prompt, the human clicks confirm.
Easier to manage
Per-business commit policies, by_agent retention, and visibility rules sit in the same admin console as every other app's settings.
Easier to extend
New propose-only tools (risk register, retro summary) ship as additions — the agent picks them up the same release.
AI is actually better
A vendor or generic AI can write a status. Only Workspace Agent can also read real activity, propose a Hill Chart from progress, gate the commit, and flag the resulting record as agent-born in the audit log.
At our head office within HR, marketing, and finance, we use ChatGPT quite a bit, and MangoApps AI Helper is a useful tool for us. As I go around to the stores, I’ve been showing them how the AI Intranet Helper can find things and answer plain language questions about our resources, and they’re quite excited about that.
Alison van Zijl
Head of Human Resources
Full House
Customer Success
Related Customer Stories
Customer Case Studies
Customer Case Studies
Customer Case Studies
Customer Case Studies
Customer Case Studies
Customer Case Studies
Frequently Asked Questions About Workspace Agent
21 tools across discovery, AI drafting, and gated writes — list and get workspaces, list and search message threads, list workspace tasks, summarize threads, get active check-ins, AI-draft a check-in response, AI-propose a Hill Chart position, AI-suggest a follow-up task, generate a client digest (internal_only redacted), and 12 confirmation-gated writes including create / archive workspace, add / remove members, create messages and tasks, complete tasks, create check-ins and commit responses, create Hill Charts and commit positions, and customize templates.
No. draft_check_in_response and propose_hill_chart_position return drafts — they do not persist anything. The user reviews, optionally edits, and then calls the matching commit tool (commit_check_in_response or commit_hill_chart_position) with confirmation.
Workspace messages support an internal_only flag at creation time. generate_client_digest automatically excludes every internal_only post from the output. The agent returns markdown — sending the digest is a separate action; the agent never auto-emails.
Workspace tasks live inside a Basecamp-style project workspace and use create_workspace_task / complete_workspace_task. The Tasks app handles org-wide assignments via the separate Tasks Agent. Both are confirmation-gated.
Check-in response rate and draft acceptance rate, Hill Chart freshness, thread-summary usage, and client digest turnaround. Compare against your pre-agent baseline.
Let's Talk
Since 2008, we've been building the workforce platform — earning the trust of 2 million+ users and an NPS of 78.
Why Choose Us?
- AI-Powered Platform: The most unified workforce experience on the planet.
- Top Security: HITRUST, ISO & SOC 2 certified.
- Exceptional UX: Delightful on mobile and desktop.
- Proven Results: 98% customer retention rate.
Trusted by Legendary Companies:
