← Blog Automating a process
AI in project management: how to use it without creating more work
AI in project management helps when it reduces operational work and improves project visibility. It doesn't help when it becomes one more channel where the team has to copy data. This guide explains what to automate, what to leave to human judgment, and how to move from loose prompts to a system you can measure.
Key takeaways
- Start with a real pain. Late reports, hidden risks, or meeting notes nobody logs are better candidates than a vague idea to “use AI.”
- AI needs project data. If Jira, Asana, Excel, Teams, or the ERP tell different stories, you need to organize sources and permissions first.
- The project manager doesn't disappear. AI can suggest risks, summarize progress, and detect delays, but scope, priority, and negotiation decisions stay human.
- Measure before and after. Define how you'll evaluate report quality, update time, useful alerts, and reduced manual work.
What is AI for in project management?
It helps read project signals, summarize scattered information, and alert the team when something drifts. The hard part isn't writing prompts. It's connecting AI to real work.
PMI defines project management as applying knowledge, skills, tools, and techniques to meet requirements. It also includes defining scope, identifying deliverables, managing risks, and communicating across teams, as explained in PMI’s definition of project management. That sets the point well: AI helps when it improves one of those tasks.
In practice, we see three uses with a clear return. First, status reports built from tickets, meeting notes, and comments. Second, alerts about risks, blockers, and delays. Third, project information search without asking the team to remember where everything was left.
Adoption is no longer marginal. In an APM/Censuswide survey published in 2025, with 1,000 project professionals, 70% said their organization already used AI and 29% planned to adopt it. Among those who saw benefits, the most cited functions were task and schedule automation, resource allocation, risk analysis, reports, and dashboards. Each was close to half of responses, as APM reports on AI use in project management.
The same survey shows the blockers: technical training, data security, integration with workflows, and inaccuracy. That’s why it’s better to treat AI as an operating system for the work, not as a list of tools.
Illustrative example, not a client. A construction company tracks site progress in Excel, runs meetings in Teams, and manages purchasing in an ERP. Every Friday, the project manager spends hours building the report. A system with AI automation can read progress, commitments, and pending orders. Then it proposes a report with status colors, risks, and pending items by owner. The manager reviews, corrects, and approves it. The expected result is less manual work and better traceability of decisions.
What tasks can you automate, and which ones should you not automate?
AI works best when the task has clear inputs and an output you can verify. If the task requires negotiation or political judgment, it should stay with people.
| Task | What AI can do | What a person should do |
|---|---|---|
| Meeting notes | Summarize agreements, dates, and owners | Validate sensitive commitments |
| Status reports | Propose progress, risks, and blockers | Approve the message to leadership |
| Schedule | Detect delays and dependencies at risk | Change priorities or scope |
| Resources | Cross-check workload, availability, and due dates | Resolve conflicts between areas |
| Risks | Find patterns in tickets, costs, or dates | Decide mitigations and budget |
| Communications | Prepare drafts for stakeholders | Manage expectations and agreements |
A good rule: if the error can be detected and corrected quickly, AI can propose. If the error affects a contract, budget, or reputation, a person decides.
How to apply AI in project management without improvising
The healthy path starts small. One well-measured use case teaches more than a large rollout with no control.
- Choose a visible pain. Look for a task that takes time today or arrives late: a weekly report, risk follow-up, blocker classification, or decision search. Avoid starting with “we want AI in the PMO.”
- Define the decision you want to improve. “Summarize information” isn’t enough. Define what someone will do with the output: approve a report, escalate a risk, reassign a task, or request missing data.
- Map sources and permissions. List where the information lives: Jira, Asana, Microsoft Project, Teams, Slack, Excel, email, or the ERP. Define who can see what. Without this, the system can mix projects, leak data, or answer with incomplete information.
- Create an evaluation before building. Build real examples: good reports, correctly detected risks, false positives, and cases the system must not answer. We use evaluation sets because they let us compare versions and measure whether the system improves. The same approach applies when measuring whether an AI assistant answers well.
- Integrate where the team works. If the team uses Jira and Teams, AI should read and write there when possible. If it forces people to copy data into another screen, adoption drops and data quality gets worse.
- Keep human review. On June 9, 2026, PMI published The Standard for Artificial Intelligence in Portfolio, Program, and Project Management, presented as the first global standard for applying AI to professional project work. PMI says the standard includes human-in-the-loop oversight at every stage, as summarized in its announcement of the AI standard for projects.
What data does AI need to help the project manager?
AI doesn't guess the project status. It needs reliable signals, correct permissions, and a way to compare what it says against reality.
- A schedule with tasks, owners, dependencies, and updated dates.
- A history of scope changes, blockers, decisions, and approvals.
- Meeting notes, commitments, and stakeholder communications.
- Budget, spend, purchase orders, or invoices when financial risk matters.
- Project definitions: scope, deliverables, acceptance criteria, and constraints.
- Role-based permissions, so each person sees only what they should see.
- Stable connections with the team’s tools. If you need to connect Jira, Teams, Excel, or an ERP, it’s worth designing the integrations and APIs well.
- Traceability: which source the system used, when it read it, and what it recommended. ISO/IEC 42001:2023 applies to organizations that develop, provide, or use AI-based systems and defines an AI management system for policies, objectives, and processes, as stated on the ISO/IEC 42001 page.
- Oversight of generated outputs. NIST states that organizational use of generative AI may require extra human review, tracking, documentation, and more management oversight in its generative AI profile.
Which option fits: tool feature, automation, or custom system?
Not everything needs custom development. The decision depends on risk, the integrations required, and the level of control you need.
Option 1
Features in the tool you already use
This fits when you need summaries, simple search, or help writing reports inside one platform. It’s the fastest path when the data is already organized.
Depends on licenses, permissions, and configuration limits.
Option 2
Automation platform
This works when you want to move information between tools: create tasks from emails, send reminders, or consolidate reports. It works well with clear rules and low risk.
Depends on connectors, volume, and error handling.
Option 3
Custom system
This makes sense when there are several sources, complex permissions, sensitive data, or quality evaluation. Here, the flow, measurement, and operation are designed together.
Depends on integrations, data, security, and testing.
Option 4
Agents for project workflows
This can help when the system has to execute several steps: review tickets, request missing information, generate a report, and escalate risks. It requires controls, limits, and action logs. If you’re at that point, review how we think about AI agents.
Depends on permissions, allowed autonomy, and human review.
What we do
Frequently asked questions
How is AI used in project management?
It’s used to summarize meetings, prepare status reports, detect delays, classify risks, and search project information. What matters is connecting it to the team’s real sources and defining which decision it improves. If it’s used only as an isolated chat, it helps little and can create more work.
Can AI replace a project manager?
It shouldn’t replace the project manager. It can reduce operational tasks and show signals that would take a person longer to find. But the project manager still negotiates scope, sets priorities, resolves conflicts, and takes responsibility for decisions with leadership, clients, or vendors.
Which AI tools help with schedules and tracking?
It depends on where your team works. Some features already come inside project management tools. Others require automations between systems. Some require custom development. The right question is which workflow you want to improve and which data is available without double entry.
How do you measure whether AI improves project management?
Define a baseline before building. Measure report preparation time, detected errors, useful alerts, false positives, team satisfaction, and decisions made with better information. Also check whether the system reduced manual work or only moved the work somewhere else.
What are the risks of using AI in projects?
The most common risks are privacy, wrong answers, outdated data, overconfidence, and lack of traceability. There’s also the risk of integrating AI poorly and creating another source of truth. That’s why it’s worth limiting permissions, keeping evidence, and maintaining human review.
Written by
Quarl engineering team
We are an AI engineering team in Medellín. We build and run systems that reach production — automations, assistants and agents — and for three years we ran an assistant for two million users across ten countries.
See how we work →Want to apply this in your company?
In fifteen minutes we tell you whether it is worth it, free and with no pitch. The proposal arrives within 48 hours.