Ask a general AI assistant which of your projects are off track and it will answer politely and at length. Look closely at the answer and you find the statuses that the project owners typed into the tracker, summarized in good prose. If the owners wrote "green" three weeks ago, the AI says green.
That is the central problem with the question. The AI is as current as the last person who updated a status field, and few people enjoy updating status fields.
Why do most AI assistants give the wrong answer?
An assistant inside a project tool sees that tool. It can summarize tickets, and it can tell you which tasks are past their due date. It cannot see the Slack thread in which the engineer wrote that the vendor API is two weeks late, or the meeting in which the team agreed to pause the project.
An enterprise search assistant sees more sources, but it treats them as documents to find. It returns the passages that match your question. Deciding from those passages whether a project is off track remains your job.
What does an AI need to answer the question?
It needs four things, and each one is a requirement on the data underneath the AI:
- The plan as a structure. Which strategy a goal serves, which projects serve the goal, and which tasks belong to each project.
- Activity from every tool. Ticket changes, completed tasks, threads, emails, documents and meeting transcripts.
- A status that comes from evidence. The system works out a project's state from that activity and keeps it next to the status the owner typed.
- Changes to the plan itself. A due date moved for the third time is a finding, even if every status field is green.
With these in place, "off track" has a definition the AI can check. Without them, the AI guesses from wording.
Which kinds of AI tools have this?
| Kind of tool | What it knows | Can it name off-track projects? |
|---|---|---|
| General chat assistant | What you paste or upload | Only from what you tell it |
| AI inside a project tool | Tasks and owner-written status in that tool | Overdue tasks, yes. Stalled projects with a green status, no |
| Enterprise search assistant | Documents and messages across tools | It finds mentions. The judgment stays with you |
| Context layer for strategy execution | The plan, activity from all tools, and a derived status | Yes, with the reason and the date |
We compare specific products in Atlassian Rovo vs Vindaris, Asana AI Studio vs Vindaris and Glean vs Vindaris.
How does Vindaris answer it?
Vindaris is the AI context layer for strategy execution. Its Work Graph maps your strategy to the goals, KPIs, projects and tasks under it, and it analyzes work activity across the tools your teams already use, such as Jira, Asana, Slack, Microsoft 365 and meeting transcripts. Nobody writes a status update for it.
When the evidence and the owner's status disagree, Vindaris raises an alert. A project that is blocked, a project whose activity has slowed, a KPI behind the pace it needs, and a due date pushed several times are all separate alert types. Each alert names its reasons and goes to the owner, weeks before a monthly review would raise the same issue.
You can ask Vindaris directly, or ask your own AI. MCP clients such as Claude and ChatGPT query the same graph through the Vindaris MCP server, so the answer to "which projects are off track?" is the same whichever AI you ask. The setup is on the AI agents page, and What Is an AI Context Layer? explains the concept.
How can you test your current AI?
Pick one project that you know is in trouble and whose status still says on track. Ask your AI which projects are off track. If the project is missing from the answer, the AI is repeating typed statuses, and every other answer it gives about your projects has the same weakness.
FAQ
Can ChatGPT or Claude tell me which projects are off track? On their own they know only what you paste in. Connected to a context layer such as Vindaris over MCP, they can answer from current work activity and name the reason.
Does this require people to update their status more often? No. Vindaris derives the status from activity in the tools your teams already use. The owner's typed status stays visible next to it.
How early does an alert arrive? An alert opens when the evidence appears, for example on the day a blocker is mentioned in a thread. That is typically weeks before a monthly or quarterly review.
Which tools does Vindaris connect to? Jira, Asana, Slack, Microsoft 365, Google Workspace, Notion and others. The full list is on the integrations page.