Will Software Development Eventually Be Estimated in Tokens?
Recently, I had an interesting thought while working with AI coding tools.
For years, we’ve estimated software projects using hours, days, sprints, and story points. That’s how most teams plan and track work.
But AI is changing how we build software.
Today, many developers use AI to generate code, write tests, explain existing code, create documentation, and even help with system design discussions.
A task that previously took a day might now take a few hours with the help of AI.
This made me wonder:
What are we actually estimating now?
Traditionally, we estimated human effort.
For example:
- Build a new API: 2 days
- Add a dashboard: 3 days
- Integrate a payment gateway: 5 days
But when AI is involved, the effort starts looking different.
The engineer may spend less time writing code and more time:
- Understanding requirements
- Reviewing AI-generated code
- Validating business logic
- Testing edge cases
- Making architectural decisions
Could AI usage become part of estimation?
I’m not saying story points will disappear.
Business complexity, dependencies, testing, and communication will always matter.
However, I can imagine a future where teams also think about:
- AI consumption
- Compute costs
- Human review effort
- Testing and validation effort
Instead of:
This feature will take 5 developer days.
Could we someday hear something like:
This feature requires significant AI assistance, several review cycles, and a day of testing.
Engineers are still the decision makers
Personally, I don’t think AI will replace software engineers.
If anything, AI makes engineering judgment more important.
Understanding the problem, making trade-offs, designing systems, and ensuring quality are still human responsibilities.
The tools are changing.
The responsibility isn’t.
What do you think?
Am I overthinking this, or do you believe AI could eventually influence how software projects are estimated?
I’d love to hear how other engineers see this evolving.
Top comments (0)