10 red flags in consultant reports — and how AI audits them

Consultant reports are written to be persuasive. That’s not a dig — persuasion is the job. But it means that any report you receive has been optimized, consciously or not, to land with the reader. The assumptions most likely to be questioned have been pre-softened. The data most flattering to the recommendation has been foregrounded. The alternatives most likely to be chosen instead have been dismissed with a paragraph of rhetorical cover.

Your job as the reader is to spot that cover. Here are the ten red flags that show up most often in consultant decks, and how to use Synthesis to do a structured audit in about five minutes.

Red flag 1: The “three options” trick

Consultant decks often present three options: a clearly bad one, a clearly outrageous one, and the one they recommend. The structure is designed to make the middle option look obvious. McKinsey did not invent this — it’s a rhetorical form older than the firm — but it’s the default shape of most strategy decks.

The audit: Ask whether the three options are genuinely representative of the decision space, or whether they are constructed to make one look right. Usually the honest option space is 5–7 alternatives, of which 2–3 are plausible. A three-option deck is a compressed answer already pointing at the conclusion.

Red flag 2: The single-figure benchmark

A consultant report that anchors the recommendation on a single benchmark number (“top-quartile companies achieve X”) is almost always cherry-picked. The benchmark is real, but the sample that produces it is specific. Whether your company resembles that sample is the actual question, and it’s almost never addressed.

The audit: For any headline benchmark, ask: what was the sample size, what were the selection criteria, what was the confidence interval, and what was the second-quartile number? If the report doesn’t answer those, the benchmark is decoration, not evidence.

Red flag 3: The implementation hand-wave

Strategic recommendations often have a slide labeled “implementation” that is 60% diagram, 30% bullet points, and 10% substance. The diagram shows boxes connected by arrows. The bullets describe a multi-phase plan. The substance — who does what, with what skills, by when, at what cost — is absent.

The audit: For each phase in the implementation plan, ask: what team owns this, what headcount is required, what is the calendar duration, and what is the cost? If the answers aren’t in the deck, the plan isn’t real — it’s a cartoon of a plan.

Red flag 4: Survivorship data

“Companies that adopted X grew 23% faster.” Unstated: the sample includes only companies that survived adopting X. Companies that adopted X and folded are not in the data. This is classic survivorship bias and it inflates the apparent efficacy of whatever was adopted.

The audit: For any “companies that did X grew Y” claim, ask whether the reference class includes companies that did X and failed. If not, the number is inflated by an unknown factor.

Red flag 5: The passive voice on risks

Risks in consultant decks are almost always written in passive voice. “Costs may be incurred.” “Timelines may slip.” Compare that to active voice in the recommendation section: “We will capture $12M in savings by Q3.”

The asymmetry is intentional. Active voice makes the recommendation feel inevitable; passive voice makes the risks feel abstract. The audit is to rewrite every risk bullet in active voice with a named owner: “Our ops team will lose 40 hours per week during migration.” That rewrite usually changes the emotional weight of the risks by an order of magnitude.

Red flag 6: The missing counter-recommendation

Every real decision has an opposing recommendation. If the report argues for Option A, there is a defensible argument for Option B that was worth making. A report that doesn’t contain a serious treatment of the strongest counter-recommendation is a sales document, not an analysis.

The audit: Read the report and try to steelman the opposite recommendation. If you can write a 3-paragraph argument for the opposite position in 10 minutes, the report is incomplete.

Red flag 7: Case studies from unrelated industries

A classic trick: “This is how Netflix handled it.” Your company is a B2B logistics firm. Netflix’s business model, capital structure, and customer behavior have almost nothing in common with yours. The case study is there for narrative ornament, not evidence.

The audit: For any case study cited as precedent, ask: do the financial dynamics of the cited company match ours (similar gross margin, similar customer acquisition profile, similar capital intensity)? If not, the case is illustrative at best, misleading at worst.

Red flag 8: Plausible-range arithmetic

“We estimate the opportunity at $50–$200M.” That’s a 4× range. Ranges that wide mean the underlying model is undercooked. A confident estimate would be tighter. A range this wide is the analytical equivalent of “a number between 0 and 1 million,” formatted to sound rigorous.

The audit: For any wide-range estimate, ask for the assumption sensitivity. What has to be true for $50M? What has to be true for $200M? Usually the answer reveals that the high end is a stretch scenario and the low end is the realistic case.

Red flag 9: The org-chart redesign

An org chart redesign buried in a strategy deck is a reliable indicator that the real work of the deck is political. The strategy is cover for a power shift. The redesign isn’t necessarily wrong, but it is rarely derived from the strategic analysis in the preceding slides.

The audit: Ask whether the same strategy could be executed under the current org. If yes, the redesign is a separate recommendation and should be debated on its own merits. If no, the deck should be explicit about which org changes are load-bearing and why.

Red flag 10: Unspecified “should”

“We should invest in AI.” This is a sentence, but it’s not a recommendation. Missing: who invests how much over what time horizon in which specific AI capabilities against which measurable goals. A real recommendation answers those five questions. A rhetorical recommendation doesn’t.

The audit: For any “we should X” sentence, ask whether the who / how-much / by-when / for-what / measured-how is answered anywhere in the deck. If not, the recommendation is unfinished work disguised as strategy.

Running the audit with Synthesis

Here’s a prompt that produces a structured audit of any consultant deck. Paste the deck’s text (or the key recommendations section) into Synthesis with the following prompt:

Audit this consultant report for the ten red flags common to strategy decks: (1) the three-options rhetorical structure, (2) cherry-picked single-figure benchmarks, (3) implementation hand-waves, (4) survivorship bias in case data, (5) passive-voice risk framing, (6) missing counter-recommendations, (7) case studies from unrelated industries, (8) implausibly wide arithmetic ranges, (9) org-chart changes unrelated to strategy, and (10) “should” statements without specific who/how-much/by-when. For each red flag that applies, cite the specific passage and explain the analytical gap. End with a ranked list of the three most important questions to ask the consultants before acting on their recommendations.

What you get back is a 500–800 word audit that a skeptical board member would take an hour to write. The orchestrator runs it in about 45 seconds because the five models’ different priors naturally surface the rhetorical patterns from different angles.

This is a pattern worth internalizing more broadly: consultant reports are a domain where multi-model diversity beats single-model rigor. One model will agree too much with the report’s framing. Five models with different priors will catch more of what the report is doing.

What to do with the audit output

A good audit is input to a conversation, not a replacement for judgment. After you run the audit:

1. Identify the one or two most important gaps. Not all red flags are equal. A missing counter-recommendation is worse than unrelated case studies.
2. Send the gaps to the consultants before the presentation meeting. This sounds aggressive; it’s actually collaborative. The good consultants will welcome sharp questions. The ones who bristle are the ones whose analysis won’t hold up.
3. Use the gaps to drive the meeting agenda. Don’t re-present the deck. Spend 20 of the 60 minutes on the three weakest parts of the analysis.
4. Document the gaps and the answers. In six months you will want to know which parts of the recommendation were robust and which were hand-waved. The audit becomes the scaffolding for that review.

The asymmetry to remember

Consultants are usually smart, often right, and reliably underpriced against the value they deliver. This post is not a dismissal. It is a reminder that you are the only person in the room whose fiduciary duty is to your own decision. Consultants have to close the engagement. You have to live with the outcome.

Reading their reports critically is not a sign of disrespect. It’s the work that the engagement was meant to produce inputs for. The consultants delivered the analysis; the decision is still yours, and the quality of the decision is determined by how hard you push on the analysis before you act.

The broader point

Any document written to persuade — consultant decks, vendor proposals, internal strategy memos — has the same rhetorical patterns. The ten red flags above are a starting kit. With a tool that can apply them at scale, the audit becomes routine: the document gets paused for a minute while Synthesis pressure-tests it, and the output of the audit becomes the agenda for the meeting.

This is a practical use of multi-model AI that most teams haven’t adopted yet. It’s cheaper than hiring a second consultancy to check the first one, faster than doing it manually, and — because it uses diverse priors — more thorough than any single reader can be.

Pressure-test the document. Then decide.