Cut backlogs, answer citizens faster, and draft policy — safely and within the rules.
Public-sector AI adoption has climbed sharply, yet only about one in five pilots reaches durable production — and tools alone never transform an agency. You'll learn to diagnose where your own work stalls, separate hype from operational reality, and frame AI as a way to rebuild capacity rather than cut.
You've built the fragments of a system: a use-case shortlist, a backlog workflow, response templates, a drafting method, a compliance checklist, transparency notices, and a 90-day scaling plan. Each piece works on its own. The path to production has five gates, in order.
Now you have to make it run. A backlog doesn't shrink because you bought a model — it shrinks when a queue moves through a sequence of small, checkable steps, and every step that touches a citizen ends at a named human who can be held responsible for it.
You're asked to produce an options paper on whether your agency should extend a permit-fee waiver for small businesses. You have a folder of consultation responses, two prior evaluations, a legal opinion, and a budget annex.
A correspondence-triage tool can satisfy every privacy and procurement rule and still trigger a front-page story if a citizen learns a machine flagged their case and nobody can explain why. Trust is a separate engineering problem, and oversight bodies test it with blunt questions: How was this decided?
The demo worked. The room nodded.
Take fifteen minutes and one real initiative. Answer in plain language, in writing — vagueness here is what hides a stall. For your chosen initiative: What work does this remove or speed up, in one sentence? Who is accountable for the output being correct?
Picture two agencies that buy the same document-summarization capability for case files. The first treats it as a tool: staff get a login and a short training. Summaries appear, but the standard operating procedure is untouched, so caseworkers read the full file anyway and paste the summary in afterward to satisfy a manager.
When you're choosing whether to push an idea forward, you're weighing two forces that pull against each other: how much value the work would deliver if it worked, and how ready your operating model is to actually run it.
Work through this with one real backlog or process you own. List the tasks hiding inside it, breaking each process into separate tasks until every line is a single repeatable action. Mark the AI-shaped ones, then tag each against the five traits: volume, text-heavy, rule-bound, bounded judgment, draft-tolerant.
EPUB, PDF, and HTML are included so the book can work on an e-reader, as a designed copy, or as a searchable desk reference.
For e-readers and reading apps.
The designed edition with diagrams and layouts intact.
Searchable, copy-pasteable, and practical as a reference.
Yes. You get the complete edition, including the chapter sequence and internal materials described on this page.
EPUB, PDF, and HTML are included so you can read on an e-reader, keep a designed copy, or use the searchable browser version.
Because this is an instant digital download, broad change-of-mind refunds are not offered after the files have been accessed. Refund requests are reviewed within 7 days for duplicate purchases, accidental purchases before access, access failures we cannot fix, wrong files, corrupted files, or pages that materially misdescribe the book.