Detective Ana Reyes finished this level of the program on a Friday evening, closed the laptop, and spent the weekend thinking about everything she had just learned: that a police report is evidence, not a memo; that an AI model can invent a detail the footage never showed and present it in perfect professional prose; that Brady v. Maryland (the Supreme Court ruling requiring disclosure of exculpatory evidence) and Giglio v. United States (the ruling requiring disclosure of evidence that could impeach a government witness) frame AI use as a constitutional matter; and that the King County (WA) prosecutor had already drawn a governance line by barring AI-written reports. On Monday morning she was back at her desk with twelve open cases, a CAD (computer-aided dispatch) system that had just been upgraded to include an AI transcription module, and no written policy from her agency about any of it. The awareness was solid. The path from awareness to competence was not obvious. This lesson is the map.

What the 90-Day Arc Is For

Completing Level 1 of this program means you understand the landscape: what AI can and cannot do in a public-safety context, why the evidence standard changes every rule, what the specific failure modes look like (gap-fill hallucination, softened facts, invented quotes), what Brady and Giglio require, what disclosure policies look like, and what roles in the agency carry the highest AI-competency stakes. That is a genuine and valuable foundation. It is not yet a credential in the operational sense, because knowledge is not the same as demonstrated skill.

The 90-day on-ramp is the bridge between the two. It is a structured progression through three phases, each with specific tasks, each building on the last, and each oriented toward a concrete standard: the ability to verify an AI-assisted draft to court standard. Not "good enough to submit," but "good enough that a defense attorney's question about the AI's role in the report has a documented, accurate answer." That is the standard the program is built around, and it is the standard this 90-day arc is designed to reach.

The arc works whether your agency has already deployed AI tools or has not deployed any yet. If your agency is already using AI drafting assistance, the ramp provides the structured practice that turns exposure into competence. If your agency has not yet deployed AI tools, the ramp provides the preparation that makes you the most capable voice in the room when deployment begins. Either way, the skills and knowledge you build over 90 days are applicable on day one of whatever your agency's actual AI timeline turns out to be.

The 90-day on-ramp is not about waiting for your agency to hand you a tool and a policy. It is about building the competence that makes you the professional who helps write the policy.

Days 1 Through 30: Understand and Map

The first month is about completing the knowledge foundation and mapping your own context. This does not mean reading every article about AI in policing. It means doing four specific things well.

Complete the evidentiary-risk inventory for your role. Go through your current documentation workflow and identify every step that could potentially involve AI, now or in the near future. For a patrol officer, that means report writing and BWC (body-worn camera) footage review. For a dispatcher, that means CAD entry and call transcription. For a records clerk, that means redaction and public-records response. For an analyst, that means crime pattern reports and case summaries. For command staff, that means every AI-assisted product that carries your name on a policy. For each touchpoint, ask: what is the legal standard the output has to meet? What is the verification standard that satisfies that legal requirement? What does disclosure look like if AI touched this product? Write that down. Not a formal memo, just a working document that names your specific exposure. This is the foundation of everything that follows.

Locate your agency's current AI policy, if one exists. You may be surprised to find that there is more documentation than you expect, or you may be equally unsurprised to find that there is none. Either way, knowing the current state of written policy in your agency is essential. If a policy exists, read it with fresh eyes after completing this program. Does it address verification? Does it address disclosure? Does it address the Brady and Giglio implications? If a policy does not exist, you now have both the competency to contribute to writing one and the context to explain why one is needed. That is a professional standing very few people in most agencies currently occupy.

Identify the AI tools your agency has already deployed or is actively evaluating. Many officers are surprised to discover that their agency has already signed contracts or is in active procurement discussions for AI tools that will directly affect their work. Check with your records division (the department responsible for managing incident reports, evidence, and other agency records) about body-camera platform contracts. Check with your dispatch supervisor about any AI modules in the CAD system. Check with your command staff about any pending RFPs (requests for proposal, the formal solicitation documents agencies issue when seeking vendor bids). Understanding what is already in the pipeline, before it arrives in your workflow, is the posture of a professional who is ready rather than surprised.

Begin Level 2 of this program. The first month of the on-ramp runs parallel to the beginning of Level 2. The two reinforce each other: the practical prompting, verification, and disclosure skills in Level 2 provide the specific techniques that make the mapping exercises in the first month concrete rather than theoretical. You do not need to complete Level 2 to finish the on-ramp; you need to begin it, so that the learning is live and applied rather than abstract.

Days 31 Through 60: Practice and Verify

The second month is about building the core verification skill. This is the month where knowledge becomes practice, and practice becomes a reliable habit.

Run a verification pass on an actual AI-assisted draft. If your agency has deployed drafting tools, this is straightforward: use the tool on a real call, generate a draft, and then execute a structured verification pass. Check every factual claim in the draft against the BWC footage, the CAD entry, and your own direct knowledge. Build a simple checklist if the tool or your agency has not provided one: call time, location, subject description, sequence of events, officer actions, subject statements, witness statements, items recovered. For each element, ask: is this supported by the footage? Is the footage timestamp consistent with the CAD entry? Did the subject actually say the thing the draft attributes to them? Document every change you make and why. Keep a log of the types of errors you catch. Over thirty days of regular practice, you will develop an intuition for the failure modes specific to the tool your agency uses.

If your agency has not yet deployed drafting tools, use training scenarios. Many body-camera platform providers offer demo environments. Several agencies that have already deployed tools have made their training scenarios available through law enforcement professional associations. The skill you are building, structured comparison of a written account against an audio-visual record, transfers regardless of the specific tool, because the failure modes of generative AI drafting are consistent across platforms.

Write or refine your personal verification checklist. By the end of the second month, you should have a checklist that is specific enough to your role and call type that you can run it consistently in under fifteen minutes. The checklist is not bureaucratic overhead; it is your evidence that the verification was done. In a deposition or a cross-examination about an AI-assisted report, "I used a structured verification checklist and I have it in my case file" is a substantially better answer than "I read through it." The checklist is also the document that closes the gap-fill loop: you cannot catch a gap-fill detail by reading the draft in isolation, because the draft will sound plausible. You catch it by checking the draft against the footage for the specific item.

Practice the disclosure answer. This sounds less tangible than the verification work, but it is equally important. The disclosure answer is the response you give to: "Officer, did you write this report, or did a computer write it?" The correct answer is not "I wrote it" (if AI generated the draft) and not "the computer wrote it" (because you reviewed, corrected, and adopted it as your own sworn account). The correct answer is something like: "An AI drafting tool generated a draft narrative from the body-worn camera audio. I reviewed that draft against the footage, made corrections where the draft was inaccurate or incomplete, and adopted the corrected version as my report. The adopted version reflects my review and my sworn account of the incident." Practice that answer until it comes out naturally and completely. Then practice answering the follow-up questions: Did you make any corrections? What kind? How long did the verification take? Those are the questions that determine whether the answer holds up.

Identify a peer group. The on-ramp is significantly more effective when it is not done alone. In month two, identify two or three colleagues who are also engaging with this material, either through this program or through their own professional development. The peer group does not need to be formal or official. It can be as simple as agreeing to share verification checklists, compare notes on the error types you are catching, and talk through the scenarios in the practice questions. The review discipline of catching a gap-fill in a colleague's practice draft, or explaining the Brady implications to someone who is not yet clear on them, accelerates your own fluency in ways that solo practice does not.

Days 61 Through 90: Apply and Document

The third month is about applying the competence you have built in a context that has operational significance, and creating the documentation that makes that competence visible and verifiable to the professionals who need to see it.

Execute a full AI-assisted report from draft to adoption. If you are a patrol officer, this means using the drafting tool on a real call, running the full verification pass, documenting the verification, and submitting the adopted report with the disclosure statement your agency policy requires (or drafting the disclosure statement your agency policy should require, if one does not yet exist). If you are a dispatcher, this means reviewing and correcting an AI-generated CAD entry, documenting the correction, and filing the verified entry. If you are a records clerk, this means running a verification pass on an AI-assisted redaction, documenting each frame or passage you reviewed, and logging the result. The specific task depends on your role. The principle is consistent: execute the full workflow, document every step, and produce a record that demonstrates the human review.

Write a one-page summary of your verification protocol. This is not a policy document; it is a personal professional standard. It describes what you do when AI is involved in a document or record you are responsible for: what verification steps you run, in what order, how long they take, what you look for, and what you document. Keep it in your professional portfolio. If your agency asks you to contribute to a formal AI policy, your one-page summary is the starting point for the verification section. If you are ever deposed about an AI-assisted report, your one-page summary is the document that demonstrates a systematic professional standard, not an ad hoc hope that the draft was probably accurate.

Complete Level 2 of this program. The 90-day on-ramp and the Level 2 curriculum are designed to run in parallel. By day 90, you should have completed Level 2's core content: prompting basics for responders, getting accurate output under an evidence standard, recognizing bad AI output in a report, reviewing and adopting an AI draft, the footage-grounded verification pass, and disclosing AI assistance. Each of those modules deepens one of the competency areas the on-ramp is building. The combination of Level 2 content and 90 days of applied practice is what produces a practitioner who can operate these tools to the evidence standard, not just a professional who knows the theory.

Contribute to your agency's policy discussion. By day 90, you have the knowledge and the documented practice to be a credible voice in your agency's AI governance conversation. That contribution looks different depending on your role and your agency's current state. For a patrol officer, it might mean volunteering to participate in the agency's AI policy working group, or sharing your verification checklist with your watch commander as a starting point for a unit standard. For command staff, it might mean presenting a one-page briefing to the chief or the city attorney on what Brady and Giglio disclosure require in the context of AI-assisted reports, and what written policy the agency should have in place before the next court challenge arrives. The contribution does not need to be dramatic. It needs to be informed, documented, and offered from the posture of a professional who has done the work.

The Court Standard: What You Are Aiming For

The endpoint of the 90-day on-ramp is not a certificate, though the Level 2 certification you complete during this period is a professional credential. The real endpoint is a professional posture that can withstand the specific pressures the public-safety context places on AI use.

The court standard, as this program uses the term, means: every AI-assisted document you are responsible for has been verified against the record, every verification has been documented, every disclosure has been made, and you can answer questions about that process, in a deposition, at a suppression hearing, in a cross-examination, or in front of an oversight board, with complete, accurate, documented answers. That standard is not unreachable. It is what a rigorous professional who has done the 90-day work can achieve.

The specific things you can say at day 90 that you could not say at day zero are: "Here is the verification checklist I used." "Here are the corrections I made to the AI draft." "Here is the disclosure statement I included in the report." "Here is the documentation that I reviewed the AI-assisted redaction and that every protected element was identified." "Here is the written policy my unit is using for AI-assisted documentation, and here is the role I played in developing it." Those statements are the output of the 90-day arc. They are also the statements that close off the lines of attack a defense attorney, a plaintiff's counsel, or an oversight investigator would otherwise use.

The professionals who complete this arc are not just better protected individually. They are better for the agencies they serve, because they are the people who can staff the working groups, train the new officers, review the vendor contracts, answer the prosecutor's questions, and keep the agency on the right side of governance decisions that are being made right now, often without enough informed voices in the room. The goal is to be one of those voices.

Key Takeaways

• The 90-day on-ramp bridges the gap between Level 1 awareness and Level 2 operational competence. The target is the ability to verify an AI-assisted draft to court standard: every claim checked against the record, every correction documented, every disclosure made.
• Days 1 through 30 focus on understanding and mapping: completing the evidentiary-risk inventory for your role, locating your agency's current AI policy, identifying tools already in deployment or procurement, and beginning Level 2 of this program.
• Days 31 through 60 focus on practice and verification: running verification passes on real or training drafts, building your personal verification checklist, practicing the disclosure answer to court-standard questions, and identifying a peer group for shared learning.
• Days 61 through 90 focus on application and documentation: executing a full AI-assisted workflow with documented verification, writing your personal one-page verification protocol, completing Level 2, and contributing to your agency's AI policy discussion from a position of informed competence.
• The verification checklist is not bureaucratic overhead. It is your evidence that the human review was done, and the specific document that catches gap-fill hallucinations that a read-through would miss because the hallucinated text sounds plausible.
• The disclosure answer requires explicit practice. "An AI tool generated a draft. I reviewed it against the footage, corrected it, and adopted the corrected version as my sworn account" is the answer. The follow-up questions about what corrections were made and how long verification took also need clean, accurate answers.
• The 90-day arc runs parallel to Level 2 content, not after it. The applied practice and the curriculum reinforce each other. By day 90, you should have both the Level 2 certification and 90 days of documented verification practice.
• The professionals who complete this arc are better positioned to contribute to agency AI policy, participate in vendor evaluation, train peers, and answer the governance questions that command staff, prosecutors, and oversight boards are asking now. That is the career value of the credential this program builds.