Many teams delay using AI because they believe their data is not ready.
The backlog is not clean.
Estimates are inconsistent.
Capacity changes every week.
So the conclusion becomes:
“We should wait until the data is better.”
In practice, this often delays progress.
You don’t need perfect data to start using AI in project management.
You need structured enough data to support better planning and better conversations.
What you actually need
To get useful results from AI, you only need a few basic elements.
1. A clear goal
What are you trying to achieve?
Not a long description. Just the intent.
Example:
Complete the core work needed for user authentication.
2. Rough team capacity
You don’t need precision.
Just:
• number of people
• availability
• known absences
Example:
4 developers, 1 QA. One developer at 50%. QA is available only in week 2.
3. A simple backlog
This is the most important input.
Each item should have:
• a short description
• a rough estimate if available
• a priority
If you are using a project management tool, your existing work items with title and description are already enough to get started.
Example:
• Feature enhancement, 3 points
• Backend update, 5 points
• Integration task, unknown estimate
The estimates don’t need to be perfect.
4. Known constraints
This is where AI becomes more useful.
Include:
• dependencies
• risks
• deadlines
• external pressure
Example:
Integration depends on another team. Demo in 10 days.
What AI can do with this
AI will not replace your planning.
But it can:
• organize information
• highlight gaps
• flag risks
• suggest questions
• create a first draft
This helps teams move faster from discussion to decision.
A practical tool you can try
If you want to experiment with this approach, tools like PMI Infinity can support this type of analysis.
You can paste your sprint goal, backlog, capacity, and constraints, and ask for a structured draft of your planning.
PMI Infinity also includes a prompt library, which can be very helpful for teams that are just getting started and are not sure how to ask the right questions.
The goal is not to rely on the tool for decisions, but to use it to organize information, identify gaps, and prepare better conversations with your team.
A simple step-by-step example
Step 1. Prepare your input
Sprint goal:
Complete core authentication functionality
Team capacity:
4 developers, 1 QA
1 developer at 50%
QA available only in week 2
Backlog:
• Feature enhancement, 3 points
• Backend update, 5 points
• Integration task, unknown
• Regression testing, 3 points
Constraints:
• Integration depends on another team
• Demo in 10 days
Step 2. Ask AI for a draft plan
You can use a simple request:
“Based on this information, suggest a realistic sprint plan, identify risks, and highlight what should not be committed yet.”
Step 3. Review the output
A useful response might include:
• recommended items to commit
• items that should be delayed
• risks and dependencies
• missing information
• questions for the team
This is already valuable.
Not because AI made the decision, but because it helped structure the conversation.
What improves accuracy over time
As you use AI more, you can add:
• past sprint performance
• velocity
• estimation patterns
• definition of done
But none of this is required to start.
A simple way to introduce this to your team
Start small.
Use:
• one sprint goal
• a few backlog items
• rough capacity
Run a quick demo and ask:
What was useful?
What was missing?
This keeps the discussion practical.
Final reflection
Waiting for perfect data often delays progress.
Using AI with imperfect data helps teams see what is missing, improve clarity, and make better decisions earlier.
The goal is not perfect answers.
It is better questions and better conversations.
If you already have a goal, a backlog, and rough capacity, you have enough to start.
Rosana Inacio — PM Insights
Leave a Reply