The attachment is titled "Data Readiness Assessment v3 FINAL.pdf" and it is sixty pages long. It took four weeks to produce and it is, by any technical standard, excellent: a complete data-source register, annotated gap exhibits, a defensible severity matrix, a clean privacy pre-flight. It lands in the steering committee's inboxes on Tuesday. On Thursday, the committee gives the topic five minutes, and those five minutes go to a verbal summary from a program manager who skimmed the executive overview on the train. The summary mentions the vendor-master duplicates. It does not mention the retention blocker on page 41, the one explaining that the resolution emails the pilot needs to learn from are auto-deleted after twelve months. The pilot is approved on schedule. It starts on time, runs beautifully for a quarter, and dies in month five, exactly where page 41 said it would, when the training set turns out to have no answers in it. The audit was right about everything. It changed nothing. This closing lesson of the chapter is about the one page that would have saved it.

The Audit Is Only as Good as the Decision It Produces

Across this chapter you built four artifacts. The data-source register told you what data exists, where it lives, and who owns it. The gap exhibits documented what is missing, duplicated, or dirty, with the evidence attached. The severity matrix triaged those gaps into blockers, parallel fixes, and accepted risks against one specific pilot. The privacy pre-flight flagged what must never reach a model and routed every flag to an owner. That is a real audit, the kind Gartner is implicitly asking for when it reports that 63 percent of organizations lack, or are unsure they have, AI-ready data practices, and that through 2026, 60 percent of AI projects without AI-ready data will be abandoned. Your organization exits that statistic in one of two ways: by luck, or deliberately. The audit is the deliberate part. But an audit does not exit anything by existing. Decisions exit statistics, and decisions are produced by documents leadership actually reads.

Here is the uncomfortable operational truth this lesson is built on: leadership reads one page. Not because executives are lazy or shallow, but because their calendar is a triage system and your pilot is one of eleven items on Thursday's agenda. A sixty-page PDF does not compete for those five minutes; it forfeits them to whoever speaks. The steering committee in the opening scene did not reject the audit. They never encountered it. What they encountered was a verbal paraphrase, and the paraphrase, like all paraphrases, lost the load-bearing detail.

Unread evidence is not evidence. It is archaeology.

So the final artifact of this chapter, and the named deliverable of this lesson, is the One-Page Data Readiness Report: the four artifacts of the audit, compressed to a single severity-ranked, decision-oriented page. And the skill this lesson teaches is the compression itself, done ruthlessly but without losing defensibility. The design principle is borrowed from how airlines dispatch aircraft. Before every flight, the captain signs a one-page dispatch release: a verdict, the conditions, the fuel numbers, the exceptions. Behind that page sits a stack of weather charts, maintenance logs, and load sheets that nobody reads at the gate but anybody can pull if challenged. The page states conclusions. The stack carries proof. The captain decides in ninety seconds precisely because someone did the work of making ninety seconds enough. Your report works the same way: the page states, the appendix proves, and you, the assessor standing in the room, can answer any challenge by turning to the exhibit behind the sentence being challenged. That is "evidence or it didn't happen" completing its loop: the page is not a substitute for the evidence, the page is the evidence's interface.

The Anatomy, Part One: Verdict, Frame, Findings

The One-Page Data Readiness Report has a fixed anatomy of six sections, and the order is not negotiable, because the order is doing psychological work. We will walk it slowly, section by section, with the actual language patterns.

Section 1: The Verdict Line, First, Not Last

The report opens with one sentence delivering one of exactly three verdicts:

• READY. "The data supports the pilot as designed; proceed." No conditions, no hedges buried in subordinate clauses.
• READY-WITH-CONDITIONS. "The data supports the pilot as designed once the following named blockers clear, by the following dates." Every condition named, every condition dated, right there in the opening block.
• NOT READY. "The data does not support the pilot as designed; here is specifically what would change the verdict, and what that costs." A NOT READY without a path is a complaint; a NOT READY with a priced path is a plan.

Why first? Because executives read the first line to decide whether to read the second. That is not cynicism; it is how professional readers process documents under load, and you should write for the reader you have. A verdict at the top means the reader who stops after one sentence still leaves with the audit's actual conclusion instead of a mood. A verdict at the bottom, after four paragraphs of methodology, means the skimming reader supplies their own verdict from tone, and tone is not an audit output.

Two calibration notes on the verdicts themselves. First, READY-WITH-CONDITIONS is the workhorse and the honest one. Real week-one audits of real operational data almost always find something, so an unconditional READY from a first-pass audit should worry everyone in the room, starting with you; it usually means the audit sampled too little or the assessor blinked. Second, remember the chapter's standing principle: a documented NOT READY is a win. It converts a nine-month, six-figure discovery into a one-page, four-week discovery. The verdict line is where that win either gets claimed in plain language or mumbled away.

Section 2: The Pilot Frame

Two lines, immediately under the verdict, restating what the pilot needs from data. Something like: "The pilot will classify and route three invoice-exception types, targeting 90 percent routing accuracy. It requires 24 months of exception tickets with resolution codes and the vendor master they resolve against." This restates the data charter you wrote at the start of the audit, and it exists because a verdict is only meaningful against a stated design. "The data is ready" is not a sentence that means anything by itself; ready for what? The same warehouse of tickets can be READY for a routing pilot and NOT READY for a payment-prediction pilot. The frame also protects you six months later, when someone tries to stretch your verdict across a pilot you never assessed. The frame is the fence around the verdict.

Section 3: The Top Findings Block

Maximum five findings. Each one gets a single line with six fields: the gap, its evidenced impact on this pilot, the fix, the cost, the owner, the date. Not the gap in general, its impact on this pilot, with the evidence sitting one exhibit away. You do not write "vendor master has quality issues." You write "14 percent of vendor records are duplicates; in the dirty-sample test, one in five duplicate-vendor exceptions misrouted; fix is a dedup and merge-rule pass; $21,000; AP operations; four weeks." One line, and a hostile reader can pull the exhibit and recompute every number in it.

Where do the five come from? Straight from the severity matrix: the blockers first, then the parallel fixes worth leadership's attention. And here the triage discipline you learned earlier in this chapter holds at the report layer too: resist finding number six. It is real, it is documented, and it lives in the appendix. Every finding you add past five dilutes the ones above it, because the reader's attention is a fixed budget and you are spending it. If everything is on the page, the page is the PDF again, just smaller.

The Anatomy, Part Two: Money, Privacy, and the Ask

Section 4: The Money Line

One comparison, in dollars: the total cost of remediation versus the cost of proceeding unready. The left side is easy; it is the sum of the fix costs in your findings block. The right side is the line most assessors never write, and it is the line that funds remediation, because it converts "data hygiene" from a virtue into an investment with a payback period. You build it from two numbers you already own: the degradation rate from your dirty-sample test (what accuracy actually did when the model saw the data as it really is) and the run-rate numbers from the baseline pack (what the process costs per year). Degradation times exposure equals risk, priced in dollars, framed as an illustration with the math shown. We will run the full calculation in the worked example below. The pattern to internalize now: without a money line, remediation is a cost and the CFO's job is to cut costs; with a money line, remediation is the cheaper of two options and the CFO's job is to pick it. The CFO here means the chief financial officer, and this is the one line on the page written primarily for them.

Section 5: The Privacy Attestation

One line: "Privacy pre-flight run on [date]; [N] flags identified and routed to owners; no unresolved red flags." Or, when it is not clean: "One unresolved red flag: customer bank details present in dispute attachments; masking fix included as Finding 2; pilot must not ingest attachments until cleared." Notice what that second version does: it states the red flag plainly, on the page, attached to a fix. What it never does is bury it. A privacy flag that surfaces later, after having been documented but omitted from the page leadership saw, does not read as an oversight. It reads as concealment, and it converts your entire audit, including all the parts that were right, into a credibility liability. PII, personally identifiable information, is the category regulators and counsel care about most; one line of attestation on the page is cheap insurance against the most expensive kind of surprise.

Section 6: The Ask

The page ends with the specific decision requested at the specific meeting: "At the September 2 steering committee: approve funding for Findings 1 and 2 ($34,000 total), confirm the named owners and dates, and approve a pilot start of October 13." Three shapes cover almost every case: fund these fixes, move the start date, or kill this version of the pilot and revisit the design. What the ask is never allowed to be is absent. A report without an ask is a memo, and memos do not produce decisions; they produce appreciation. The steering committee, a body that exists to decide, will happily receive information all year without deciding anything, unless the document in front of it names the decision, the options, and the date. Write the ask last when drafting, place it last on the page, and make it the second thing you polish after the verdict line, because the verdict opens the reader and the ask closes the meeting.

The Worked Example: The Invoice-Exception One-Pager

Here is the actual report for the audit this chapter has been running: the invoice-exception process with the roughly $428,000 annual run rate from the baseline pack. Every figure is illustrative, built for the pattern, not the precision. This is the block to copy.

Walk the money line's arithmetic once, because it is the line that gets challenged. The baseline pack priced the process at about $49 per exception and roughly 8,700 exceptions per year. The dirty-sample test, run earlier in the audit, showed routing accuracy dropping 8 points when the model saw the data uncleaned. Eight percent of 8,700 is about 700 additional misrouted items a year. Each misroute costs a second handling pass plus, on a fraction of items, a late-payment penalty: call it $124 average, all-in. Seven hundred times $124 is roughly $86,000 a year, risked to avoid a $34,000 one-time fix. Presented that way, the steering committee is not being asked to care about data quality. It is being asked whether it would like to spend 34 to avoid 86, annually. That meeting takes four minutes, and the answer funds your remediation backlog. Every number in the chain came from an artifact you already built: degradation rate from the gap exhibits, run rate and unit cost from the baseline pack. The report invents nothing; it only compresses.

And notice the ask's final sentence, the quiet one: if the fixes are not funded, the verdict reverts to NOT READY. That is not a threat; it is the logical structure of a conditional verdict made explicit. READY-WITH-CONDITIONS means the conditions are load-bearing. A committee that wants the pilot without the conditions is asking you to falsify the audit, and the sentence makes that visible before anyone can do it by accident.

Three Readers, One Page, and the AI Red Team

A one-page report is read by three different people in three different ways, usually in under a minute each, and good compression writes for all three simultaneously. The CFO scans for the money line: what does this cost, what does it save, what is the payback. The COO, the chief operating officer, scans for owners and dates: who is on the hook and when do I check. The sponsor scans for the verdict and the ask: can I take this to my boss, and what am I asking for. Test your draft against all three scans: can each reader find their line in ten seconds without reading the others? If the money is smeared across three findings instead of totaled in one line, the CFO scan fails. If the fixes say "IT" instead of a name, the COO scan fails. If the verdict has grown dependent clauses, the sponsor scan fails. The page is one document with three landing strips.

AI's three jobs on this page

AI, which drafted half your audit artifacts this chapter, has exactly three jobs at the report layer, and all three are assistive, not authorial.

• Compression passes. Your findings arrive from the severity matrix as paragraphs; the page needs lines. Prompt pattern: "Compress this finding to 25 words without losing the number, the owner, or the date." Run it per finding, then verify that nothing quantitative fell out, because compression is exactly where models drop qualifiers and round away the inconvenient digit.
• Red-teaming the page. Before the meeting, hand the model the draft and ask: "You are a hostile CFO who thinks this pilot is a distraction. What are the five hardest challenges you would raise against this page?" Then rehearse the answers, each of which should be a turn to a specific exhibit. The three challenges the invoice-exception page should expect: where does $124 per misroute come from (Exhibit E, the baseline pack's unit-cost table plus the penalty-rate sample), why should we believe the 8-point drop (Exhibit C, the dirty-sample test protocol and results), and why can't we start now and fix in parallel (Exhibit B, the 31 percent of tickets with no resolution rationale, which is the training set's missing answer key). You will not be surprised in the room, because you already held the meeting against a tireless simulated skeptic.
• Consistency-checking page against appendix. Every number on the page must recompute from an exhibit behind it. Prompt the model with both and ask it to list every figure on the page and where in the appendix it comes from, flagging any orphan. Then check its checking. This is the cite-your-source discipline from the first lesson of this level closing its loop: you began Level 2 refusing to accept an AI claim without a source, and you end this chapter refusing to issue one.

One more line travels with the report wherever it goes: the verification note, as shown in the worked example. Which figures were AI-extracted, what was spot-checked by hand, what the spot-check found. It costs one row of the table and buys quiet credibility: the reader who wonders whether the machine made up the 14 percent finds the answer already waiting. In a year when most organizations still cannot say which of their numbers a model touched, the note is a small structural advantage that compounds.

Zoom out to close the chapter. MIT found 95 percent of GenAI pilots deliver no measurable return, and the autopsies keep finding the same organizational absences: no baseline, no integration, no honest gate. The data readiness audit you can now run is one of those gates, and this report is the gate's handle. A pilot that proceeds through a documented READY-WITH-CONDITIONS, with funded fixes and named owners, is playing a different game than the 60 percent Gartner expects to be abandoned for un-ready data. And when the audit says NOT READY and the page says so plainly with a priced path forward, you have not failed; you have delivered the cheapest kill or the best-justified delay in the program's arsenal. Next, Chapter 2.4 turns from the data to the people side of readiness, starting with AI-assisted interview synthesis at scale. The One-Page Data Readiness Report, meanwhile, takes its place beside the process inventory, the process map, the SOP, the baseline pack, and the scorecard in the assessment pack you will assemble in the Level 2 capstone. The audit is done. Make it decide something.

What to Do Monday Morning

You have the four chapter artifacts, or you are about to. Here is the sequence that turns them into a decision.

1. Draft the one-pager from the four artifacts. Verdict line first, ask last, findings pulled straight from the severity matrix's blockers and top parallel fixes, capped at five. If you are tempted by finding six, it goes to the appendix, and the temptation goes in your notes.
2. Price the money line. Remediation total on the left; on the right, your dirty-sample degradation rate multiplied through your baseline pack's volume and unit cost, shown as illustrative math with every input traceable to an exhibit.
3. Run the hostile-CFO red-team prompt against the draft, pick the three hardest challenges, and rehearse the answers out loud, each answer ending at a named exhibit. If any challenge has no exhibit behind it, the page has a claim it cannot defend; fix the page, not the rehearsal.
4. Attach the verification note. One line: which figures were AI-extracted, what you spot-checked by hand, what the spot-check found. Never ship the page without it.
5. Book the ten minutes. Get the report on the agenda of the specific meeting where the ask gets decided, with the decision named in the agenda line itself. A report waiting for a meeting to happen to it is the sixty-page PDF with better formatting.

Key Takeaways

• Treat the audit as unfinished until it produces a decision, because leadership reads one page and unread evidence is archaeology, not evidence; the One-Page Data Readiness Report is the four chapter artifacts compressed into the page that gets read.
• Open with the verdict line, never close with it: READY, READY-WITH-CONDITIONS with named and dated blockers, or NOT READY with a priced path, and distrust any unconditional READY from a week-one audit.
• Frame the verdict with two lines restating what this pilot needs from data, because a readiness verdict is only meaningful against a stated design and the frame stops it being stretched to pilots you never assessed.
• Cap the findings block at five one-line findings, each carrying gap, evidenced impact on this pilot, fix, cost, owner, and date, pulled from the severity matrix; finding six lives in the appendix.
• Write the money line as remediation cost versus the priced cost of proceeding unready, built from the dirty-sample degradation rate and the baseline pack's run rate, because that comparison is what converts data hygiene into a funded investment.
• State the privacy attestation in one plain line and never bury a flag, since a documented-but-omitted privacy issue discovered later reads as concealment and destroys the credibility of everything else on the page.
• End with a specific ask at a specific meeting (fund the fixes, move the date, or kill this version), because a report without an ask is a memo and memos do not produce decisions.
• Use AI for compression passes, hostile-CFO red-teaming, and page-versus-appendix consistency checks, then ship the verification note with the page: every number must recompute from an exhibit, and the reader should be told which figures were machine-extracted and spot-checked.