Before AI-assisted development, estimating a project was fairly straightforward: calculate the expected hours, multiply by the hourly rate, and charge the client accordingly. Now agentic coding tools can speed up implementation considerably. If the goal is to maintain a similar revenue stream, how are you explaining and justifying the cost when the actual development takes fewer hours? Saying that the client is paying for the completed feature and the agency's responsibility still feels somewhat vague. We have started pushing monthly retainers more aggressively, but I'm interested in how other agencies and freelancers are handling pricing and scope.
4 Answers
A surprising amount of billable work was never just writing code. Requirements discovery, dealing with changing policies, project coordination, updating tickets, testing, reviewing generated code, and explaining tradeoffs still take time. AI often means the finished feature is more polished rather than simply cheaper, because you can spend the saved time on improvements that would previously have been cut.
I’d avoid billing strictly for effort and focus on the value of the outcome. The client is paying for a working solution, the expertise required to design it, and the responsibility of delivering and supporting it—not just the number of hours someone spent typing code.
A practical approach is to charge a fixed project price or a monthly fee rather than exposing every hour. AI may let you deliver more features, improve polish, or finish sooner, but the price should reflect the scope, business impact, risk, and ongoing responsibility. You can also set clear limits around revisions, support, and monthly capacity so the retainer remains scalable.
AI can reduce implementation time, but it doesn’t eliminate the work. You still have to guide the tools, review the output, make architectural decisions, test everything, handle edge cases, and take responsibility for production quality. I’ve also been able to accept projects that would previously have been too expensive, such as building a CRM on top of an ERP database. A project that might once have cost around $250,000 could now land closer to $75,000–$150,000, depending on scope.
The extra supervision is real. People who haven’t used these tools often imagine that you provide one prompt and receive production-ready code, but keeping the architecture coherent while reviewing generated work can be mentally exhausting.
That’s why we’ve been quoting more fixed-price projects and retainers. If the tools reduce our hours while the client receives the same or better result, the client doesn’t necessarily need to see a timesheet-based discount.
The important part with a monthly fee is defining capacity and boundaries. Otherwise clients may assume the retainer includes unlimited work, no matter how much they request.