From the Public Safety edition of October 15, 2026
Cost-benefit study finds AI police-report tools can show positive NPV under per-officer pricing but negative under bundled contracts, without valuing report quality
Erik Alda (SocArXiv / OSF preprint) · Policing / AI-assisted report writing evaluation · Manchester, New Hampshire, United States (medium-department modeling)
- Publisher
- SocArXiv (OSF)
- Original publication
- June 12, 2026
- Source retrieved
- 2026-10-29
What happened
Alda's June 12, 2026 preprint models agency-budget cost-benefit for AI-assisted police report writing at Manchester PD. Under per-officer subscription pricing, average NPV is about $2.3M (favorable productivity reading) or $0.8M (conservative reading anchored to the department's RCT null on speed) for a medium department, with positive return likely in both; attributing a bundled vendor contract instead yields a negative return. The author warns productivity evidence draws on Draft One field results while modeled price is PoliceNarratives.AI at Manchester, report quality is not valued, and results should be read as an upper bound on the labor-savings case. Newly relevant amid ongoing U.S. Draft One pilots and quality debates.
Why it matters
Gives procurement teams a quantitative caution: pricing structure and how much time savings are credited to AI can flip the ROI sign even before quality risks are priced.
Evidence and measured results
Opened the SocArXiv PDF via OSF download. Preprint modeling study; not peer-reviewed here; mixes products and does not measure downstream justice outcomes.
Limitations and uncertainty
Preprint; quality unvalued; Draft One evidence vs PoliceNarratives.AI price; Manchester-specific.
Put this evidence to work
Lighthouse Advisory interpretation, grounded in this source. Enriched 2026-10-29; this does not change the original publication date. Each role follows the same checklist so takeaways compare across sources.
Sales
Role takeaway
- Customer problem
- Departments are buying AI report writers without a clear, quality-aware ROI under their actual contract structure.
- Stakeholders
- Police chief
- Budget/finance
- Patrol command
- City procurement
- Vendor account team
- Discovery questions
- Is pricing per seat or bundled with cameras?
- What local time-savings evidence exists?
- Will quality metrics enter the go/no-go?
- How many officers would be licensed?
- Value hypothesis
- Contract-aware CBA can prevent false-positive ROI from bundle accounting; it cannot guarantee field time savings.
- Potential engagement
- Two-week AI report-writing ROI model using the agency's wages and contract.
- Unsupported claims
- AI always pays for itself
- Manchester results transfer unchanged
- Quality is unaffected
Pre-sales engineering
Role takeaway
- Fit
- Medium agencies evaluating Draft One-class tools with either seat or suite pricing.
- Architecture and integration
- Capture timestamps from CAD/RMS draft-to-submit, seat license counts and cost centers separating AI from BWC storage.
- Prerequisites
- Baseline report duration data
- Wage/benefit loaded rates
- Contract price schedule
- Constraints
- Preprint methods
- Cross-product evidence/price mismatch
- Quality omitted
- Security
- Keep BWC audio and drafts in approved CJIS/cloud boundaries when instrumenting pilots.
- Proof of value
- Replicate NPV under local prices
- Sensitivity on time-savings credit
- Include a quality-failure cost scenario
Delivery
Role takeaway
- Work
- Design pilot metrics, allocate costs correctly and brief leadership on subscription vs bundle NPV.
- Dependencies
- RMS/BWC vendors
- Finance
- Union/policy on AI drafts
- Ownership
- Patrol administration with budget office
- Skills and adoption
- Supervisors need to score AI-draft quality, not only speed.
- Governance checkpoints
- Pilot go/no-go criteria set
- Mid-pilot metrics review
- Procurement award memo with CBA
- Acceptance criteria
- Documented cost attribution method
- Local sensitivity analysis delivered
- Quality sample plan approved
- Risks
- Over-crediting AI for organizational learning
- Ignoring prosecutor rejection costs
- Bundle accounting hides true AI spend
Implementation considerations
Lighthouse Advisory interpretation across the operating dimensions a public-sector buyer must settle before this evidence becomes a design. Each note answers the question under its heading for this specific source.
Architecture and integration
What must connect, and where does the AI sit in the workflow?
Separates subscription SaaS costing from bundled suite contracts; implies agencies should meter AI report seats independently of body-camera bundles when testing ROI.
Governance
Who approves, reviews and stays accountable for outcomes?
Require pre-registered success metrics including quality and prosecutor outcomes, not labor hours alone.
Security and privacy
What data, permissions and controls need testing?
Paper focuses on labor economics; does not assess BWC-audio privacy or model-provider data flows.
Accessibility and workforce
Who is affected, and what skills or accommodations follow?
Assumes officers still review drafts; adoption curves affect NPV more than headline vendor time-savings claims.
Procurement
What should contracts, pricing and exit terms secure?
Compare per-officer subscription versus bundled attribution in any AI report-writing buy.
Operating model
Which teams own the service once it runs?
Operational owner: budget office with patrol command evaluating AI report tools.
Publication history
- 2026-10-15Public Safety · Issue 516 resources
Stable resource ID: alda-cba-ai-police-report-writing-20260612