Project Manager Agent
Does the chasing and the reporting so your PM can run the project.
What is project manager agent?
A project manager agent collects progress updates from the people doing the work, keeps the plan current, flags tasks that have slipped or are about to, and produces the status report, leaving priorities and decisions to your project manager.
Project management splits cleanly into two jobs. One is judgement: what matters, what to drop, which risk is real, how to tell a client the date moved.
The other is collection. Chasing eleven people for updates, moving dates in the plan, noticing that a task due Tuesday has not started, and assembling the same status report every Friday.
The second job consumes most of the week and needs none of the skill. An agent does the chasing and the assembly. It does not decide what matters, because that is the part your project manager is actually for.
Why we do not automate reprioritisation
Duara builds project agents from Nairobi for teams running several projects at once, where the plan is out of date because updating it competes with delivering.
We do not build agents that reprioritise work. A model that reshuffles a plan optimises for whatever it was told to measure, usually dates, and quietly sacrifices the things nobody wrote down. Priority is a judgement with politics in it, and it stays with a person.
Who this is for
- The plan is out of date within a week
- Status reports take a day to assemble
- Slippage is noticed at the deadline
- Your PM spends more time chasing than managing
Industries we serve
- Construction and engineering
- Software delivery
- Marketing and creative agencies
- Consulting firms
- NGO programme delivery
- Manufacturing projects
Problems this solves
These describe projects where the administration has crowded out the management.
- The plan goes stale immediately. Updating it competes with delivering, so it is accurate on the day it is written and wrong by Thursday.
- Status reporting eats a day. Assembling the same report each week is manual, repetitive and out of date by the time it is circulated.
- Slippage noticed at the deadline. A task that never started is discovered on its due date, when the options have narrowed to bad ones.
- Chasing crowds out managing. Your project manager spends the week collecting updates rather than resolving the problems the updates reveal.
What we deliver
Updates collected from the people doing the work, the plan kept current, early flags on tasks at risk, and the status report assembled automatically for your PM to edit.
Progress updates collected from the team
A plan that stays current without manual updating
Early flags on tasks at risk of slipping
Status reports assembled for review
Dependency warnings when a blocker moves
A record of what changed and when
Features
- It never reprioritises. Reordering work is a judgement with politics in it. The agent reports and your project manager decides.
- Chasing where people already are. Updates are requested over the channel your team actually uses rather than in a tool they avoid.
- Early warning, not deadline warning. A task not started three days before it is due is flagged then, while options still exist.
- Reports drafted for editing. The status report is assembled and your PM changes the framing, which is the part that needs a person.
- Dependencies tracked. When one task moves, the things waiting on it are flagged rather than discovered later.
Benefits
- The plan reflects reality. A current plan is the difference between managing a project and describing one.
- A day a week returned. Report assembly and chasing are the two largest recoverable costs in most project offices.
- Problems surface with time to act. Early flags convert a crisis into a decision.
- Your PM does the job they were hired for. Time moves from collection to judgement, which is where their value actually is.
Technologies we use
- Update collection. Requesting and capturing progress over the channels your team already uses.
- Plan synchronisation. Writing collected updates back into your project tool so one version stays current.
- Risk detection. Flagging tasks not started or running late against their own history rather than at the deadline.
- Dependency mapping. Tracking what is waiting on what, so a slip propagates visibly.
- Report assembly. Drafting the status report from current data for a person to edit and send.
Update collection is the hard part in practice. People do not update project tools, so an agent that only reads the tool learns nothing.
How the process works, step by step
A project agent usually takes four to six weeks, and it is worth starting on one project rather than all of them.
-
Step 1
Free project review
We look at how updates are collected today, how stale the plan gets, and what reporting costs you.
-
Step 2
One project first
We start with a single project, because conventions that work across a portfolio are discovered rather than designed.
-
Step 3
Update collection live
Chasing begins over your team existing channels, which is the change they notice most.
-
Step 4
Flags tuned
Risk thresholds are adjusted so warnings are early enough to act on and rare enough to read.
-
Step 5
Reporting added
Status report assembly follows once the underlying data is genuinely current.
Quality assurance and how we work
The commitments attached to a system inside your delivery process.
- It does not reprioritise work. A model that reshuffles a plan optimises for what it was told to measure and sacrifices what nobody wrote down. Priority stays with a person.
- It does not report on individuals. Flags are about tasks and dependencies. We will not build a productivity score for your staff.
- Reports are drafted, never sent. Your project manager edits and sends. Framing a status report is a judgement about the audience.
- We start with one project. Rolling across a portfolio before conventions have settled produces a system everybody works around.
- We report when chasing is the wrong fix. Where updates are missing because the plan is unrealistic, we say so rather than automating the chasing.
Frequently asked questions about project manager agent
Does it replace our project manager?
No. It does the chasing and the assembly. Judgement about what matters and how to tell a client the date moved stays with them.
Will it reprioritise our work?
No, and we will not build that. Reshuffling a plan optimises for what it was told to measure and sacrifices what nobody wrote down.
How does it get updates from people?
It asks over the channel your team already uses. An agent that only reads the project tool learns nothing, because people do not update project tools.
Does it work with our project software?
Most common tools, yes, writing updates back so one version stays current rather than creating a second plan.
What if the team ignores it?
Then the plan is usually unrealistic rather than the team uncooperative, and we report that instead of automating the chasing harder.
Will it monitor individual performance?
No. Flags are about tasks and dependencies. We will not build a productivity score for your staff.
When does it flag a task?
When it has not started or is running late against its own history, early enough that options still exist.
Does it send the status report?
No. It drafts it and your project manager edits and sends, because framing a report is a judgement about the audience.
Can it handle several projects?
Yes, though we start with one, since conventions that work across a portfolio are discovered rather than designed.
How soon before the plan stays current?
Four to six weeks, with update collection going live first because it is the change teams notice most.
Does it track dependencies?
Yes, so when one task moves the things waiting on it are flagged rather than discovered later.
Can it estimate how late we will be?
It projects from current progress and labels it as projection. It does not present that as a commitment.
Does it work for construction projects?
Yes, and site updates over WhatsApp are a common arrangement where nobody is opening project software.
What about agile teams?
It works alongside a board rather than replacing ceremonies. Standups and retrospectives are conversations, not reporting.
Can clients see the reports?
Only what you send them. The agent drafts internally and sharing stays a deliberate decision.
Will it chase people too often?
Frequency is set by you and tuned during setup, since an agent that nags gets muted like any other.
Does it record what changed?
Yes, a dated record of plan changes, which is frequently useful in a dispute about what was agreed.
Can it flag budget as well as time?
Where you connect cost data, yes, though budget judgement stays with your finance and project leads.
Is our project data confidential?
Yes. Plans and updates stay in your systems and are never used to train public AI models.
What does a project agent cost to run?
Project and team numbers determine it, quoted against your own portfolio before anything is agreed.
Glossary of terms
Project terms in plain language.
- Dependency
- A task that cannot start until another finishes, which is how one slip becomes several.
- Critical path
- The sequence of tasks that determines the project end date.
- Slippage
- A task running later than planned, worth catching before its due date rather than on it.
- Status report
- The periodic summary of progress, risks and decisions needed.
- Blocker
- Something preventing work from continuing, where speed of escalation matters most.
- Baseline plan
- The original agreed schedule, kept for comparison as the plan changes.
Related services in AI Agents
Talk to us about project manager agent
Add up the hours your project office spends chasing updates and assembling reports in a month. That is the recoverable figure.
Start your project