Preliminary Assessment of Significant Changes to State’s Technology Project Approval and Oversight Processes

Summary

The administration is implementing a new process to approve certain types of technology projects called the Project Delivery Lifecycle (PDL). This new process will likely replace the state’s current information technology (IT) project approval process—the Project Approval Lifecycle (PAL)—over the next few years. The administration anticipates the new PDL process will be a more incremental and a more iterative approach to technology project planning than the current PAL process thereby better meeting state entities’ needs, reducing risk to the state, and incorporating potential generative artificial intelligence (GenAI) project needs. State entities request hundreds of millions of dollars through the annual budget process to plan, develop, and implement technology projects. As such, a new approval process that is expected to launch for all GenAI projects in just a few months warrants close legislative scrutiny.

Thus far, the administration has used only part of the new PDL process on a few technology projects. However, those projects largely have not completed more than the planning phase. As such, the Legislature does not have the information necessary to evaluate the new process and whether it would address any shortcomings of the existing process. This lack of information, especially given the implementation time line, is problematic because PDL might differ in significant ways from PAL. Some information currently available through PAL—such as a project’s total baseline cost, schedule, and scope—might not be as readily available in PDL while other information might be made confidential sooner during project planning. Therefore, to help the Legislature evaluate and understand the new process over a more reasonable time frame, we recommend the Legislature require reporting on completed project planning activities, and limit the new process to a small subset of projects (with additional reporting) over the next fiscal year.

Background

Current State IT Project Approval Process Approved Nearly a Decade Ago

The California Department of Technology (CDT) is the administration’s central IT entity with broad authority over most aspects of technology in state government. CDT—in consultation with other departments, such as the Department of Finance (DOF)—is responsible for the review and approval of IT projects proposed by state entities, and for the oversight of their development and implementation until completed. In 2016, CDT fully implemented the state’s current process for planning and approving IT projects—the PAL process. The primary purpose of PAL was to improve the planning of state IT projects through four discrete stages, all of which taken together create a complete project plan with a baseline project cost, schedule, and scope that is approved by CDT and DOF. Figure 1 provides a graphical depiction of PAL.

For each stage of PAL, a department is required to submit an associated planning document to CDT for review and/or approval. The Legislature also receives a copy of these documents, which can give the Legislature a better understanding of—and more confidence in—the project cost, schedule, and scope prior to approving funding for project development and implementation through the annual budget process. The time line to complete PAL varies based on a project’s complexity and cost, ranging from several months to several years.

Recent Advancements in GenAI Technology

Over the past several years, advancements in artificial intelligence (AI) technology—and, in particular, GenAI—have introduced both opportunities and challenges for state entities. State law generally defines AI as a system that varies in its level of autonomy to take a variety of inputs and generate outputs that, based on certain objectives, influence both physical and virtual environments. GenAI is subsequently defined by state law as an AI system that can generate text, images, video, and audio that generally emulates the AI system’s inputs. Some examples of potential opportunities for state entities to use GenAI technology include improved analysis of complex data; more effective communication with state residents about government programs and services; and personalization of information provided to state residents based on their language and service needs. Some possible challenges for state entities in using this technology include the delivery of false and/or misleading information to state residents; increased cybersecurity and privacy risks; and opaque decision-making by the AI system.

Governor Issued Executive Order (EO) to Study and Deploy GenAI Technology in State Government

In September 2023, the Governor issued EO N-12-23 directing a number of state entities to study the use of GenAI technology within state government and to develop and implement new guidelines, processes, and procedures for its deployment. Over the past year and a half, the state entities named in the EO have completed a number of deliverables to accomplish the Governor’s objectives. Key deliverables include an initial report on benefits and risks of state government use of GenAI technology, updates to guidelines and training for government employees on GenAI use, and several rounds of procurement documents for GenAI projects.

Several State Departments Developed GenAI Projects

Several of the deliverables completed pursuant to the EO involve what is referred to as a “proof of concept” (POC). A POC is generally defined in administrative policy as a test of the functionality of a solution or technology. The GenAI POCs started with several state entities identifying opportunities for GenAI technology to help solve particular state program or service problems. The administration—led by CDT, the Department of General Services, the Government Operations Agency (GovOps), and the Office of Data and Innovation (ODI)—picked four state entities with a total of five POCs for the first round. GovOps then issued a request for innovative ideas (RFI²) for each of the GenAI POCs, containing a problem statement that potential bidders would address through a combination of demonstrations and written responses. The administration then selected two vendors for each of the five GenAI POCs, and has worked with each of the four state entities over the past year to test various GenAI technologies. GovOps announced a second round of POCs with three state entities at the end of last year, with procurement through RFI²s still underway as of April 2025. Figure 2 provides a description of each of the current GenAI POCs.

The New PDL Process for GenAI Projects

Consistent with the requirements of the EO, the most recent deliverable completed in February 2025 makes changes to the state’s technology project procurement and approval processes. The changes are expected to facilitate additional procurement of GenAI projects by state entities. Among these changes is a new process for planning and approving GenAI projects called the PDL. The PDL, led by CDT in coordination with DOF and ODI, is anticipated to be both a more incremental and a more iterative approach to technology project planning that is informed by the administration’s experience with the first round of GenAI POCs. (An incremental and iterative approach is one that delivers system functionalities to a department as they are developed, so a department can use and improve them over time instead of at the end of system development.)

Under PDL, state entities would begin by (1) identifying and refining a business problem to be solved by GenAI technology, (2) using an RFI² or comparable procurement method to solicit and select one or more vendors for a POC, and (3) analyzing the POC results to select one or more vendors for a “minimum viable product” (or MVP). (An MVP is generally defined in state administrative policy as a solution that only includes minimum capabilities to confirm customer needs and demands prior to full development.) If the MVP is successful, the state entity would continue to work with the vendor(s) to plan, develop, and implement the remainder of the project until completed.

The administration launched an initial version of the PDL process in February, and expects to launch a final version of the PDL in July. While the process currently only applies to GenAI projects, the administration expects to add additional technology projects (including state IT projects) to PDL in the future. Figure 3 provides a high-level graphical representation of the new PDL process.

Preliminary Assessment of New PDL Process

PDL Process Likely to Replace the PAL Process Over Time. The state’s current IT project approval process, PAL, helps state entities plan dozens of IT projects each year. State entities then collectively request hundreds of millions of dollars each year through the annual budget process to plan, develop, and implement these projects. Therefore, the introduction of a new process for GenAI projects, which the administration and the Legislature have not yet had the opportunity to fully evaluate, is an important development that warrants close legislative scrutiny. Moreover, the explicit intent of the administration to expand PDL to other technology projects, including at least some state IT projects, suggests the administration likely will replace PAL with PDL over time. How PAL and PDL will operate together as two parallel but separate approval processes in the meantime—particularly as GenAI technology becomes more ubiquitous across projects—remains uncertain.

New PDL Process Could Improve Technology Project Outcomes… Based on our preliminary assessment of the new PDL process, we acknowledge that PDL—in concept—could improve the outcomes of GenAI (and potentially other) technology projects. PDL is more incremental and iterative than PAL, which means state entities could determine more quickly whether a vendor’s technical solution to their program or service problem actually works. If not, CDT could suspend or terminate the project before additional resources are expended, thereby reducing risk to the state. Also, PDL appears to measure success based on state entities receiving business value from projects, as opposed to PAL which emphasizes completion of system requirements. While business value may be a better measure of success, it is unclear at this time exactly how business value is defined and/or determined by the administration. It is possible that, if successful, PDL could help improve the state’s experience with planning, developing, and implementing technology projects over time.

…But Premature as First Round of GenAI POCs Remain Underway. However, the launch of a final version of the new PDL process for all GenAI projects by July is very aggressive and would be premature. This is because the administration cites its experience with the first round of GenAI POCs as a key reason for the introduction of PDL but, as of April 2025, only some of the POCs are expected to lead to subsequent contracts for state entities to procure GenAI technology. As a result, nearly all information relevant to assess the POC process will either be newly available or remains confidential because several of the GenAI POCs are in active procurement. POCs are a central component of the new PDL planning process but, due to the delayed time line for the first round of GenAI POCs, the Legislature does not have any information to evaluate the POC process. The absence of this information makes it very difficult for the Legislature to know whether a key feature of the new PDL process works well, and to understand how it will differ in practice from the current PAL process.

POC Outcomes Could Provide Insights to PDL Improvements and Challenges. Provisional budget bill language adopted in the 2024-25 Budget Act made clear the Legislature’s intention to better understand how the first round of GenAI POCs would impact its role through the annual budget process in evaluating and approving technology project funding. Monthly meetings with the administration to discuss EO deliverables and progress on GenAI POCs, as well as required disclosure of the funding sources for GenAI POCs and subsequent contracts, were meant to help the Legislature understand how GenAI technology would be evaluated and impact state entity operations over time. Though information about a number of EO deliverables, including high-level overviews of GenAI POCs, have been shared with the Legislature, detailed information about the outcomes of the first round POCs has not been shared. These details will be critical to understanding how subsequent rounds of POCs will work, including if and when funding for the first set of GenAI projects will be requested through the annual budget process. Before the PDL process is fully implemented, the Legislature may wish to assess how it works based on the outcomes of the GenAI POCs. In particular, the Legislature may wish to consider whether changes to its own oversight of this new process may be warranted given differences in the planning processes and the types of projects being planned through each process.

PDL Could Be Refined Through Second Round of GenAI POCs. In addition, the second round of potentially more impactful GenAI POCs (and the first round to fully use the new PDL process) will provide a clear opportunity for the Legislature to see how PDL works in practice, as well as better understand how it differs from PAL. If the PDL process were pilot-tested based on the first and second POC rounds, the Legislature would have time to evaluate the new PDL process as it is refined by the administration based on actual experience with GenAI POCs and subsequent contracts. This time line would be more reasonable than the administration’s plan for a full launch of the PDL process this July, giving both the administration and the Legislature additional time to adjust their own processes (including the annual budget process) to the new realities of GenAI technology.

Legislative Oversight Through PDL Likely Will Differ From That Through PAL. One of the adjustments the Legislature may need to make in response to the new PDL process is in how it will oversee technology projects. Currently, the Legislature uses approved planning documents to understand the baseline cost, schedule, and scope of a state IT project often before approving the funding for development and implementation through the annual budget process. While the administration expects a similar level of detail from projects planned through PDL as planned through PAL, some of the differences between the two processes make this uncertain. For example, the incremental and iterative nature of the PDL process means baseline cost information might not be available until at least the MVP of a project is completed and the subsequent project iterations are better understood. Moreover, the use of POCs might mean a project goes into confidential procurement sooner than previous state IT projects, limiting what information is available for the Legislature to review ahead of funding requests through the annual budget process. To make informed decisions on these requests, the Legislature must have detailed information about the cost, schedule, and scope of a project—whether or not it involves GenAI—before approving their development and implementation. As the new PDL process is fully implemented and refined, the Legislature will need to understand and adapt to these differences to maintain robust oversight through its existing processes.

Recommendations

To help the Legislature evaluate and understand the PDL process, particularly how its role in overseeing technology projects will change as PDL replaces PAL, we provide recommendations aimed at gathering additional information about the GenAI POCs going through the new process. These recommendations, together with the continued use of existing opportunities for legislative engagement such as monthly meetings, will enable close legislative scrutiny of this new process.

Require Administration to Provide Report on Outcomes of First Round of GenAI POCs. We recommend the Legislature approve provisional budget bill language in the 2025-26 Budget Act requiring the administration to provide a report on outcomes of and lessons learned from the first round of GenAI POCs. Outcomes information about the first round of GenAI POCs will allow the Legislature to make a preliminary assessment of how the new PDL process works. This is because the POC process is a central component of the new PDL process. The Legislature also could consider requiring reporting on the funding sources for any POCs with subsequent contracts for GenAI technology. Knowing if and when additional funding might be requested through the annual budget process for GenAI POCs will better help the Legislature assess and prepare for discussions on these types of projects. We recommend the reporting be provided no later than the end of the calendar year, to inform the next annual budget process and any potential requests for funding for subsequent GenAI POCs and MVPs.

Limit Use of New PDL Process to Current GenAI POCs, and Require Reporting on Outcomes From the Pilot. In addition, we recommend that the Legislature direct the administration to limit the use of its new PDL process to only the first and second round GenAI POCs through the end of 2025-26. The Legislature could consider requiring reporting on the outcomes of this “PDL pilot” and any lessons learned. This limitation on the use of the new PDL process would give the administration and the Legislature more time to evaluate and refine the PDL process based on experience and information from GenAI POCs and subsequent contracts for GenAI technology. In particular, the Legislature may want to consider whether changes to the PDL process are necessary to enable effective legislative oversight of the technology projects to which the process applies.

Require Administration to Continue Monthly Legislative Staff Meetings. We recommend the Legislature continue to require monthly meetings with the administration in 2025-26 on the GenAI POCs and any other forthcoming GenAI-related changes. These monthly meetings will provide an opportunity for the administration to explain its evaluation and refinements of the PDL process, as well as address other important topics such as state workforce readiness and training for adoption of GenAI technology.