Soldier DataAI Advisory Request a call
Sample report

What is on your desk at the end.

This is the shape of the report the AI Readiness Assessment produces. The office, the people and the numbers are invented. The sections are the ones you get.

Sample · invented organization · not client data

AI Readiness Assessment · Final report

Example County Emergency Management

Six interviews · three weeks · one office of six

1 Summary

Your staff already use AI in six ways. None is approved. Most are harmless and save real time. One, the wording of public alerts, should stay with people.

We found 18 candidate uses. Of the top eight, three are worth starting now, three pay off but need a sign-off rule first, one waits, and one should be left alone.

2 Ranked inventory top 8 of 18

UseWhoStaff hours a yearExposureCall
After-action report first drafts from activation logsPlanner, director60LowStart
Grant application and progress report draftsDirector170LowStart
Exercise scenario and inject draftsPlanner90LowStart
Hazard mitigation plan section draftsPlanner220MediumWith controls
Continuity plan refreshPlanner, department liaisons120MediumWith controls
Situation report drafts, once exercisedPlanning section60HighWith controls
Volunteer and partner outreach emailsCoordinator60LowLater
Public alert wordingDirector, duty officer20HighLeave alone

Exposure is what could go wrong: private data, a wrong answer to the public, or a decision about a person. Public alert wording is left alone because a wrong alert reaches the public before anyone can check it.

3 Data and rules review

  • No written rule for public chatbots. Four of six staff use one. None could say what is allowed. Write the rule first.
  • The office cannot issue a rule alone. County IT, risk management and the prosecutor’s office have a say. Ask first whether the county already has an AI rule, and build on it.
  • Three after-action reports are unwritten. The activation logs exist. This is the first project: the exposure is low and the backlog is the pain.
  • Nothing says AI stays out of public alerts. Say it in writing. A person writes and sends every alert.
  • There is no plan for when the tool is down. Any use during an activation needs a paper fallback that staff have practiced.
  • Prompts and drafts are public records. Decide how long they are kept before the first project starts, not after a request arrives.
  • The mitigation plan update is fourteen months out. AI can shorten drafting. It does not shorten the public process, state and FEMA review, or local adoption.

4 Three first projects

01

After-action backlog

Three reports and their improvement plans drafted from your activation logs, each reviewed by the director.

Owner
Director, with the planner
Estimate
Scoped and quoted after the assessment

02

One-page AI rule

What is allowed, what stays out, and a flat rule on alerts. Drafted by the director and cleared with county IT and counsel.

Owner
Director
Estimate
2 weeks, staff time

03

Mitigation plan drafting workflow

First drafts of the community profile and capability sections from sourced data, for each participating jurisdiction. The risk assessment and strategy stay with the planner.

Owner
Planner
Estimate
Runs alongside the update

5 90-day plan

PolicyDirector

TrainingDirector

PilotPlanner

ReviewDirector

DecisionDirector

Week 1Week 13

Decision in weeks 12 to 13: keep the after-action workflow, widen it, or stop. The review in weeks 10 and 11 gives the director the numbers to decide.

6 An AI rule one page · a starting draft for the director to clear with the county

  • A person writes and sends every public alert. AI is not used to write them.
  • Nothing that identifies a resident goes into a public AI tool.
  • During an activation, AI drafts are used only where a paper fallback exists and has been practiced.
  • A named person checks every figure in a plan or report against its source.
  • Security-sensitive details about infrastructure stay out of AI tools.
  • Prompts and drafts are public records. Keep them under the retention schedule.

7 What we did not look at

  • We did not test any tool against your data. The first project does that.
  • We did not review your alerting systems or how they are configured.
  • We did not assess whether this fee or later work is allowable under your grants. Your grant manager confirms that.
  • Hours are staff estimates for a year from interviews, not measurements. This work comes in bursts, so a monthly figure would mislead.