Your foreman sends a message at 7:15am: "Sparky confirmed he can start Tuesday but needs the trenches dug first."
That message contains two tasks, a dependency, and a deadline. In most construction teams, none of those get tracked. The message scrolls up. The foreman moves on. Tuesday arrives. The trenches aren't dug. The electrician goes home.
This is what Tai was built to fix.
What Tai Does
Tai is Convoe's task management AI. It reads your team's conversations — in channels, threads, and direct messages — and creates tracked, assigned tasks automatically.
No manual entry. No @-commands. No opening a second app. Your team communicates the way they already do, and Tai handles the rest.
When someone in your site channel says "we need to re-price the tiling variation before Friday," Tai creates:
- Task: Re-price tiling variation
- Assigned to: The person responsible based on conversation context
- Due date: Friday
- Priority: Inferred from urgency cues
- Context: Linked back to the original message
The task appears on your board, your list view, your calendar, and your timeline — without anyone touching a task interface.
Why Manual Task Creation Fails on Construction Sites
Every PM knows the drill. You have a project management tool. It has a task board. It works — in theory.
In practice, here's what happens on a busy site:
Your site supervisor is standing in a half-framed house at 6:45am. A subcontractor asks about the footing spec. The supervisor pulls out their phone, sends a WhatsApp message with a photo and a voice note. Done.
Now ask that same supervisor to open your PM tool, navigate to the right project, click "New Task," type a title, set a due date, assign someone, and save. Five steps of admin work in the middle of a physical job.
It won't happen consistently. It never has. The problem isn't discipline — it's that the tools require manual data entry at the exact moment when field workers are least available to do it.
The result: FMI Corporation found that construction projects waste approximately 52% of working time on non-optimal activities, with poor communication as a primary driver. Not because people aren't talking — they are. Because what they say never becomes what gets tracked.
How Tai Works
Tai uses natural language understanding to identify when a conversation contains an action item. Not every message needs to become a task — "good morning" and "running late" don't qualify. Tai understands context.
What Tai Detects
Tai identifies construction-specific action patterns:
- Material procurement: "Need 40 bags of cement at Site B by Friday" → Task created, due Friday, assigned to procurement
- Defect flagging: "Crack in eastern footing needs inspection before pour" → Inspection task, flagged urgent
- Scope changes: "Client wants the kitchen island moved 300mm left" → Scope change task, assigned to PM for cost review
- RFI creation: "Waiting on architect to confirm window spec" → RFI task, assigned to site contact
- Safety actions: "Scaffold needs re-inspection after yesterday's wind" → Safety task, high priority
- Subcontractor coordination: "Electrician needs to be on site before plumber arrives Thursday" → Coordination task with dependency
What Tai Creates
For each detected action item, Tai generates:
- A task title — Clear, actionable description extracted from the conversation
- An assignee — Based on who was addressed or who is contextually responsible
- A due date — Detected from time references ("by Friday," "before the pour," "end of week")
- A priority level — Inferred from urgency cues, overridable by the PM
- A conversation link — Every task connects back to the message that created it
Multiple Views, One Source of Truth
Tasks created by Tai appear across every view:
- Board view — Kanban-style columns for your workflow stages
- List view — Sortable, filterable task list for bulk management
- Calendar view — See what's due when, across all projects
- Timeline view — Gantt-style view for dependencies and critical path
Every view shows the same data. Update a task in any view, it's updated everywhere. The conversation link stays attached — so when someone asks "why is this task here?" the answer is one click away.
A Day on Site With Tai
Here's what changes when Tai is running on your project:
06:45 — PM checks Convoe before leaving for site. Tai created 7 tasks overnight from the afternoon's messages. Three are flagged urgent. PM reviews and adjusts two assignments. Takes 90 seconds.
07:30 — Site foreman photographs a drainage issue, posts it to the project channel with a note: "Water pooling in the western trench. Need to sort before the slab prep." Tai creates a rectification task, attaches the photo, assigns it to the drainage sub, flags it to PM.
09:15 — Client messages about a variation to the entry facade. Tai creates a scope change task, flags it for cost review, and links the original client message as context.
11:00 — GC messages about a concrete delivery window change. Tai creates a coordination task for the foreman and a secondary task to notify the pump operator.
14:00 — PM checks the board. Every task from the day's conversations is there. No status meeting needed to know where things stand.
17:00 — End of day. PM reviews tomorrow's open tasks. No chasing. No "what happened with that." Clean visibility into what's pending.
Compare that to a day without Tai: the same conversations happened. The same decisions were made. But none of them were tracked. Tomorrow morning starts with a standing meeting to reconstruct what everyone already discussed.
The "No Special Syntax" Design
One of the most common objections to AI tools on construction sites: "My subbies won't use it."
Fair concern. And it's why Tai requires zero changes to how your team communicates.
Some tools require commands: "@bot create task: fix drainage." Others need structured input: "TASK: [description] ASSIGN: [name] DUE: [date]." These approaches fail in construction because the people who most need to create tasks — field workers, subbies, foremen — won't adopt structured communication patterns.
Tai works with natural language. How people actually talk on site. There's no training required. The subbie who sends "I'll have the rough-in done by Thursday" doesn't need to know Tai exists. The task is created. The commitment is tracked.
Tai vs. AI Summarisation
A lot of PM tools now advertise "AI-powered task creation." Most of them are actually AI summarisation — the AI reads your conversation and produces a summary with suggested action items. A human still has to read the summary, decide which suggestions are valid, and manually create the tasks.
That's useful. It's not what Tai does.
Tai creates the task. Not a suggestion. Not a summary. The actual task — in your task system, assigned, with a deadline, linked to context. No human intermediary required.
The distinction matters because the failure mode on a construction site is rarely "the PM decided not to create the task." It's "the PM was dealing with three other things and the task creation step got skipped." Tai eliminates that step entirely.
Tai Works With Maya and Workflows
Tai is one part of Convoe's AI layer. It works alongside:
- Maya — Handles email-to-task conversion. Client sends a request via email? Maya creates the task. Vendor invoice arrives? Maya tracks it.
- Workflows — Turns structured documents (SWMs, SOPs, OH&S forms) into chat-based completion flows. Tai tracks the completion status.
Together, they close every gap between where work originates — conversations, emails, documents — and where it gets tracked.
Getting Started
Tai is included in every Convoe plan. No add-on. No per-user AI surcharge. Your team starts chatting, Tai starts creating tasks.
The setup:
- Create a Convoe workspace for your project
- Add your site team — foreman, subbies, PM
- Start communicating in channels the way you normally would
- Watch tasks appear on your board
No migration. No training. No disruption to how your crew already works.
If your team communicates — and every construction team does — Tai turns those conversations into the task list you never had time to build manually.