Autonomous coding agents are not the future. They are already here. Agents can now design, build, test, and deploy an entire full-stack feature from front end to back end without a human touching the keyboard.
The reality is that while this technology has advanced quickly, Generally Accepted Accounting Principles (GAAP) have not traditionally focused on the cost of tools used in development. Under current U.S. GAAP, you can capitalize certain third-party software costs if they are a direct cost of creating software during the application development stage. Historically, though, developer tools were treated as overhead because their cost could not be directly tied to capitalizable work. Under GAAP, work that meets the criteria should be capitalized. When agents perform that work, they should be treated no differently than salaried engineers.
Link to headingThe current state: GAAP’s phase-based model
Under ASC 350-40 (Internal-Use Software), development costs fall into three stages:
The preliminary project stage includes research, planning, and feasibility. Costs in this stage are expensed immediately
The application development stage covers coding, integration, and testing before launch. Certain direct costs in this stage are eligible for capitalization and can be recorded as an asset on the balance sheet and amortized over time
The post-implementation stage includes maintenance, bug fixes, and minor enhancements. Costs in this stage are expensed
For human developers, salaries incurred during the application development stage meet the criteria for capitalization under ASC 350-40.
For AI agents, the same capitalization rules apply. If you can track their usage and tie the cost directly to capitalizable activities such as coding, integration, and testing, those costs can also be capitalized. Historically, that has been difficult because the cost of developer tools could not be allocated with this level of precision. AI agent usage changes that. Detailed logging lets you measure exactly how much work happened during the application development stage, just like you would for a human developer.
Link to headingWhy this has been overlooked
GAAP does not prohibit capitalization of AI-related development costs. The blind spot is that most accounting teams have not considered developer tool costs in this way because, in the past, they were used across all phases of a project and could not be allocated directly.
With AI agents, usage can be tracked at a feature or project level, making it possible to isolate costs tied directly to capitalizable work. This is a shift from treating all tool costs as overhead to treating qualifying AI agent usage as a direct cost of development.
Link to headingThe case for capitalizing AI agents like salaried developers
Treating AI agents like salaried developers when they perform capitalizable work would:
Match economic substance: Both human and AI capacity create long-lived software assets during development. The accounting treatment should reflect that they are performing the same type of work
Increase transparency: Investors see a truer picture of asset creation when all qualifying development activity, regardless of whether it is human or AI, is capitalized
Improve comparability: Aligning treatment of human and AI development costs avoids distorted cost structures and makes project economics easier to compare
Apply existing GAAP consistently: ASC 350-40 already allows capitalization for qualifying third-party software costs. AI agents should be no different when their work meets the criteria
Link to headingQuantitative impact
Consider a $500,000 AI agent development project that meets GAAP criteria for capitalization:
Link to headingFrequently asked questions (FAQs)
If AI agents are a service and not employees, can their costs still be capitalized?
Yes. GAAP focuses on the nature of the work, not how it is paid for. If an AI agent performs work that meets the criteria for capitalization, those costs can qualify in the same way as human labor.
Isn’t tracking AI usage too difficult?
Tracking was a challenge for human developers too, yet companies managed it when capitalization mattered. AI tools can actually make it easier by generating granular usage logs that are often more accurate than manual time tracking.
Won’t capitalizing AI agent costs artificially inflate assets?
Not if the rules are applied correctly. Only work that clearly meets the definition of the application development stage should be capitalized. The same guardrails apply whether the work is done by a person or by an AI agent.
Link to headingWhy this matters now
The current habit of treating all AI agent costs as overhead is a carryover from a time when tool usage could not be measured and allocated. That era is over. AI makes it possible to measure, allocate, and document the cost of development work at a level that meets GAAP’s existing requirements for capitalization.
If the goal of financial reporting is to reflect economic reality, then AI agent capacity used in development should be accounted for like any other direct development cost under ASC 350-40. The rules already allow it, and finance teams now have the ability to track and apply them.
Ready to put AI agents to work?
Build AI agents with the AI SDK. Create agents that can design, build, test, and deploy entire features.
Build AI agents on Vercel