Practice 04
AI & Automation
Assess a workflow, set AI use rules, and test a limited pilot.
We start with a task staff repeat: sorting incoming requests, comparing contract terms, or pulling information from reports. We check whether a simpler process or system change would solve it first. Where AI or automation makes sense, we agree on what records a tool may use, test it on a small sample, and have staff check the output before recommending broader use. Alongside the assessment, we help the agency set its direction: which kinds of work are candidates, which are off-limits, what approvals a new tool needs, and how the agency will judge whether a tool is worth keeping.
Talk with an advisorOutcomes
What success looks like.
- A shortlist of staff tasks worth testing, with expected time savings and review needs
- A written AI use policy the agency can adopt: permitted uses, prohibited uses, approval steps, and staff responsibilities
- Data, privacy, records, accessibility, and purchasing requirements identified before a tool is selected
- A sequenced plan for the next one to three use cases, with rough cost and staffing estimates
- Pilot results compared with the current way of working
- A clear decision to expand, revise, or stop the pilot
Deliverables
What you receive.
- Workflow review and prioritized use-case list
- Draft AI use policy and tool approval checklist
- AI strategy summary: candidate work areas, off-limits areas, and a sequenced plan with cost and staffing estimates
- Small pilot plan with baseline, test cases, and staff reviewers
- Results log covering errors, exceptions, staff time, and costs
- Recommendation and handoff instructions for the agency
Engagement model
How we work together.
Start with a defined assessment and strategy session. If a use case holds up, agree separately on a limited pilot and any continuing support.
Ideal for
Who we serve.
- Departments spending hours sorting or re-entering information
- Agencies receiving AI proposals without a way to evaluate them
- Teams that need a controlled test before changing a public-facing process
Questions to resolve before the work begins
Scope, staffing, and cost.
Plain-language answers on scope, timeline, staffing, and cost — the same ones we give in the first procurement conversation.
- Do we need to buy an AI product before starting?
- No. We first look at the task and the systems you already use. A process change or existing feature may be enough. We only recommend a new tool when the test gives your agency a reason to use one.
- Will staff or resident records be put into a public AI tool?
- Not without agency approval. We agree on permitted data, access, retention, and review requirements before testing. Sensitive agency or resident records do not go into unapproved public AI tools.
- What does the strategy part cover?
- It sets the agency's direction before tools are purchased: which kinds of work are candidates for AI or automation, which are off-limits, what approvals a new tool needs, and a sequenced plan for the next few use cases with rough cost and staffing estimates. Leadership can adopt the policy and plan even if no pilot follows.
- How do you decide whether a pilot worked?
- We record how long the task takes today, test representative cases, and track errors, exceptions, staff review time, and costs. Staff check the results before the agency decides whether to expand, change, or stop the work.