CaseAdvancedModel Fluency & the AI PM Role / Managing stakeholder expectations and AI hype / #4

Your board asks why competitors ship AI features faster. Prepare your answer.

ORDER · what Alba Thorndike actually tells Otterburn's board about Tidewatch

Otterburn Analytics builds Coldwire, a tool that watches supplier signals, port data, and customs filings, and flags a supplier that might fail to deliver, usually within about a day of a real disruption. A rival called Tidewatch ships its own alerts in under twenty minutes. After two competitive deals mentioned Tidewatch's speed on the way out the door, the board asked product manager Alba Thorndike a fair question in the quarterly meeting: why does a competitor ship AI features faster? She has until Friday to decide what's actually worth telling them.

The direct answer
Tell the board the truth, backed by a number checked firsthand: Coldwire ships slower because every alert clears a corroboration check Tidewatch skips, and in a week of testing Tidewatch's own product, three out of every four of its alerts turned out to be false. Never promise to match their speed on a set timeline, that's a claim nobody can take back once it's wrong. Name the one place worth actually speeding up instead.
Do this, in order
  1. Say the real tradeoff plainly: Coldwire ships slower because every alert clears a check Tidewatch skips.Why: this is the actual answer to the board's question, not a defense of the roadmap.
  2. Back it with a number from a hands-on test, not a claim about company values.Why: a board that hears "we're being careful" with no evidence behind it will assume it's an excuse.
  3. Never promise to match their speed on a fixed timeline.Why: that promise is the hardest thing in this whole answer to walk back if it turns out to be wrong.
  4. Check that "faster" is really the right thing to compare before agreeing to it.Why: an alert that arrives early and wrong isn't actually ahead of one that arrives right.
  5. Name the one place worth actually speeding up.Why: shows judgment instead of a blanket defense of the current pace.
  6. Commit to re-testing the rival's product on a real schedule.Why: their false-alarm rate could change, and the honest answer should change with it.

How to answer this, stage by stage

Nobody is grading whether Alba can sound confident for five minutes. They're grading whether the thing she says can survive someone on the board actually going and checking it.

1
Pin the question to one board, one product, one afternoon
Say it like this
"Let's make this real. I run product for Coldwire at Otterburn Analytics, we flag supplier disruptions for procurement teams. My board asked me straight, in the quarterly meeting, why a rival called Tidewatch ships alerts faster than we do."
Why this works
Naming the real product and the real room stops the answer from turning into a general speech about competition.
2
Say out loud what the question is really testing
Say it like this
"This isn't really asking me to explain our engineering speed. It's asking whether I'll get defensive, or whether I'll tell them something true and checkable about what 'faster' is actually buying Tidewatch's customers."
Why this works
Reframing it early stops the answer from turning into a roadmap promise nobody asked for.
3
Name your method before you use it
Say it like this
"I'd run this as ORDER. Outcome, what the board actually needs from me. Reversibility, which claim is hardest to take back once I've said it. Dependency, what has to be true for speed to even be the right thing to compare. Evidence, what I can check cheaply before I open my mouth. Rank, the actual answer, defended."
Why this works
Two seconds of structure signals a method, not a mood, before the pressure of the room sets in.
4
Give the call, committed, before any evidence
Say it like this
"Here's the answer. We ship slower because every alert clears a check Tidewatch skips. Their speed likely comes with a much higher false-alarm rate. I've checked, and here's what I found."
Why this works
This is the direct answer, said plainly, before a single number arrives to back it up.
5
Show the number you went and got yourself
Say it like this
"I ran Tidewatch's own trial for a week against the same fifty accounts we already watch. It sent sixty-one alerts. Sixteen were real. Coldwire, watching the same accounts that same week, sent nineteen. Seventeen were real."
Why this works
A number from a hands-on test beats a claim about company values every time it's said in a room.
6
Say which option you turned down, and why
Say it like this
"I thought about promising the board we'd match their speed by next quarter. I'm not doing that. That's a promise I can't take back if it's the wrong call, and right now I don't have evidence it's the right one."
Why this works
Naming a rejected option shows real judgment, not just the first idea that occurred to her.
7
End on what you'd check again, and when
Say it like this
"I'll run this same test on Tidewatch again next quarter. Their product changes, and if their false-alarm rate ever comes down, my answer changes too."
Why this works
Ends on something the board can actually hold her to later, not just a confident closing line.

Let's learn

Coldwire is a tool that watches supplier signals, port data, customs filings, and financial filings, and tells a procurement team when a supplier might fail to deliver.

Before anything like it existed, a procurement team tracked this by hand: a shared spreadsheet, a Google alert, and whoever remembered to check the news that week. It took a real person about half a day every week, and it still missed things that broke overnight, a factory fire, a border closure, a supplier gone quiet.

Coldwire changed that. It checks every signal against a second source and against a labeled history of past false alarms before it tells anyone anything.

What does "clears the gate" mean? Before an alert goes out, it's checked against a second, independent source and against a list of past signals of that same kind that turned out to be nothing. If it doesn't clear that bar, it doesn't ship yet. It waits, or it never ships at all.

That gate is why a team hears about a real problem a median of about a day after it happens, not half a day of manual searching a week later. Out of every 100 alerts Coldwire sends, about 92 turn out to be real.

Hand sketched comparison diagram titled Same signal, two different gates. Left panel, a funnel icon labeled Tidewatch, caption: no second source checked, alert sent in under 20 minutes. Right panel, a gauge icon labeled Coldwire, caption: second source checked first, alert sent in about 29 hours.
Same disruption, same morning. One tool checks first and speaks second. The other speaks first.

Then the board asked a fair question. Why does Tidewatch, a well-funded rival, ship its alerts in under twenty minutes?

Here is the turn. The extra hours Coldwire takes are not the real story. The real story is what happens to a team that gets alerts fast and wrong. A procurement lead piloting Tidewatch got sixty-one alerts in one week. She called suppliers, checked with logistics, chased down every single one. Sixteen were real. By the fourth week, she stopped checking them one at a time. She started skimming subject lines and moving on. That is exactly when a real port closure alert sat in her inbox for nine days before anyone noticed, because it looked like every other alert she had already learned to ignore.

Alerts sent vs. alerts that were real, one week, same fifty accounts
60 30 0 19 17 Coldwire 61 16 Tidewatch
Alerts sentConfirmed real
Coldwire sent fewer alerts, and nearly all of them held up. Tidewatch sent three times as many, and most of them didn't.
We are not slow. Tidewatch is fast and wrong three times out of four, and that costs more than the extra hours ever did.

At its worst, a tool that ships fast and wrong does not save anyone time. It trains a person to stop reading it, and the one time it's actually right, nobody is listening anymore.

The choice I would take back For a while, Otterburn answered every "why is a rival faster" question with a defense of the roadmap, built from press coverage and a landing page, never from actually trying the other tool. I would go get the evidence first, every single time, before I ever open my mouth in a board meeting.

What I would leave alone: the corroboration check on high-cost signal types, a supplier's real bankruptcy risk, a port closure, stays exactly as slow as it needs to be. Getting one of those wrong costs a factory a shutdown, not an afternoon, so speed is never the thing worth trading there.

The lesson: a board asking why a rival is faster is not asking you to match them. It's asking whether you actually know what their speed is buying their customers. Answer that, and the question stops being a threat.

Now here is the same thing as a story

The short version above is what you'd actually say in the room. Read this one for the six weeks nobody at Otterburn was watching Tidewatch's own customers get burned.

Merricat Isbell has run procurement for Ilkeston Outfitters for eleven years. She can read a shipping delay before customs even calls, mostly because she still keeps her own list of which factories run behind after a holiday.

Ilkeston brought in Tidewatch that spring, a six-week pilot before anyone signed anything. For the first two weeks, it felt like magic. A signal would fire, sometimes minutes after Merricat's own contact had barely heard the news, and she checked every single one. She called the supplier. She called logistics. Every call came back the same way: yes, that's real, good catch.

By week three, she'd stopped calling on the small stuff. By week four, she was skimming subject lines over coffee and deleting most of them without opening a single one. Nobody told her to. She just noticed she was spending two hours a day chasing alerts, and only about one in four of them ever turned into anything real.

Hand sketched two panel comparison titled How Merricat's mornings changed in four weeks. Left panel, a person icon labeled Week 1, caption: Merricat checks every alert with a phone call. Right panel, a person icon labeled Week 4, caption: Merricat skims the subject line and moves on.
Nobody decided to stop checking. It happened the way a habit always thins, a little at a time.

Nothing dramatic happened in week five. That part matters. No single bad morning, no one email that broke her trust all at once. It built up the way a habit always does, until checking stopped being the default at all.

Then, in week six, a real alert came through: a strike at a port her biggest supplier ships through. It sat in her inbox for nine days, indistinguishable from the forty other alerts she'd already learned to skip that same week, until a factory contact mentioned it on an unrelated call. By then, a container that should have rerouted was sitting behind a picket line, and Ilkeston's warehouse was two weeks from an empty shelf.

Merricat didn't blame herself, and she shouldn't have. She'd have been a fool to keep calling on every alert when three out of four turned out to be nothing. That's the sensible thing to do with a tool that cries wolf. The tool trained her to stop listening, and then, once, it needed her to be listening.

She called Otterburn's sales team the next morning, not angry, just done. She told the whole story in about four minutes, the way you'd tell a friend about a bad week, and asked if Coldwire actually worked any differently.

That call reached Alba by Monday. She didn't write a competitive one-pager off it, the way Otterburn used to. Two years earlier, in a meeting about a different rival, she'd done exactly that: pulled slides together from press coverage and a landing page, called it a competitive response, and never once opened the other tool herself. It read well. It was also wrong about two of its three claims, and someone caught it before it went out. She never wanted to sit in that room again.

So this time, on Tuesday, she signed up for Tidewatch's own trial with a card that wasn't Otterburn's. She pointed it at the same fifty supplier accounts Coldwire already watched. By Thursday evening, she had a full week of alerts logged next to what had actually happened to each supplier. Tidewatch sent sixty-one. Sixteen were real. Coldwire, watching the same accounts that same week, sent nineteen. Seventeen were real.

We didn't build something slower. We built something that has to be right before it's allowed to be fast.

I want to say the problem is that Tidewatch is a bad product. It isn't, not exactly. It's a product built to be believed the moment it speaks. Coldwire is built to be checked once and trusted after. Those are two different bets about what a procurement team actually needs from an alert, and only one of them survives a supplier that fudges a filing, a satellite feed with a gap in it, or a customs form that's three days late itself.

Here's what I'd have told my past self, the one writing that slide deck two years earlier: a board asking why a rival is faster isn't asking you to race them. It's asking whether you actually know what their speed is costing their customers. On Friday morning, Alba walked into the board meeting with that answer already in hand. Not a promise. A number.

The five letters that keep an honest answer from turning into a promise

PICK would fit if this were one tradeoff with a clean line down the middle. Here there are four real ways to answer the board, and the job is ranking them by what's hardest to undo. That's ORDER's job.

Hand sketched labeled parts diagram titled The five letters, before you open your mouth. A center document icon labeled Board Answer, with five callouts around it: Outcome, what they need. Reversibility, hardest to unsay. Dependency, what has to be true. Evidence, cheap to check first. Rank, the real answer.
Five questions to run through before a single word reaches the board.
OOutcome. What the answer actually has to protect.
All four things Alba could say are competing for the same job: does the board walk out with an honest, checkable read on the real tradeoff, or does it walk out with a defensive dodge dressed up as confidence. Not "sound composed for five minutes." A board that trusts what she tells them next time, whether the news is good or not.
Name the outcome before ranking anything. Skip this and any answer she gives sounds like a matter of personality, not judgment.
RReversibility. Which claim is hardest to walk back.
Promising the board "we'll match Tidewatch's speed by next quarter" is cheap to say and expensive to undo. Once it's on record, missing it reads as a broken promise, even if the honest technical answer is that matching that pace means dropping a check currently catching three out of four of Tidewatch's own false alarms for their customers. Naming the real tradeoff honestly costs nothing to update later if the evidence changes.
This is why order matters, not preference. One choice is a sentence she can revise next quarter. The other is a commitment the board will hold her to whether or not it was ever the right call.
Hand sketched two panel comparison titled Which claim is hardest to walk back. Left panel, a box icon labeled Promise to match their speed, caption: bolted shut, hard to take back once said to a board. Right panel, a scale icon labeled Name the real tradeoff, caption: swings both ways, easy to update as the evidence changes.
Reversibility isn't a reason to avoid the harder truth. It's a reason to be sure of it first.
DDependency. What has to already be true.
For "they ship faster" to actually be the right comparison, their speed has to come with alerts worth trusting. That's the premise worth checking before agreeing to it. If Tidewatch's alerts are fast and mostly wrong, speed isn't really the thing being compared at all. Accuracy is, and the board is asking the wrong question without knowing it yet.
This is why Alba couldn't answer from a press release. The dependency only gets checked by actually watching what the other tool does, on real accounts, for real days.
Hand sketched flow diagram titled What has to happen before the board meeting. Five boxes connected by arrows, left to right, the second box highlighted: Hear Merricat's story. Run the trial herself. Log alerts against real events. Decide what to say. Brief the board.
Every box after the second one only works once the second one is actually done.
EEvidence. What's cheap to check before answering.
Before saying anything to the board, Alba ran Tidewatch's own trial for a week against the same fifty accounts Coldwire already watches, and logged every alert against what actually happened. Sixty-one alerts, sixteen real. Her own product, the same accounts, the same week: nineteen alerts, seventeen real.
That's the number that actually settled this, not a guess about whether a rival "probably cuts corners."
False alarms Merricat had to chase, week by week
10 5 0 Stops checking each one, week 4 Real alert sits unread, week 6 Week 1 Week 4 Week 6
False alarms, weeklyThe week checking stopped
By week four the line hadn't spiked, it had just climbed steadily. That's what made it easy to miss, and easy to stop watching.
RRank. The actual answer, defended.
Tell the board plainly: Coldwire ships slower because every alert clears a corroboration check Tidewatch skips, and in a week of testing, three out of every four of Tidewatch's alerts were false. No promise to match their timeline. The one thing worth actually speeding up is the check on low-cost signal types, a routine weather delay, a minor customs hold, where being wrong costs an afternoon, not a factory shutdown.
If this rank would be identical no matter what the outcome in the O step was, say "make Otterburn look composed on a slide," it was picked by habit, not judgment. Swap the outcome to "protect the board's trust in what Alba says, whether the news is good or bad," and the rank still holds, because an honest number costs nothing to defend later.
Hand sketched timeline titled The week Alba went and checked. Four milestones along a horizontal line, the fourth one highlighted: Monday, hears Merricat's story secondhand. Tuesday, signs up for Tidewatch's own trial. Thursday, logs 61 alerts against real outcomes. Friday, gives the board the ranked answer.
Four days between hearing a secondhand story and standing in front of the board with a number instead of a guess.

Three things worth stating directly, since the real judgment sits here. The alternative worth naming and rejecting is promising the board a fixed timeline to match Tidewatch's speed, which sounds like the most reassuring answer in the room. It loses because the evidence check already answered the real question: the rival isn't just faster, it's wrong about three out of every four alerts it sends, and matching that pace would mean dropping the same corroboration check that's currently the whole product. The AI-specific failure worth naming is alert fatigue from an ungated model output: a system that skips corroboration and confidence calibration trains the very people watching it to stop reading it, which is exactly what happened to Merricat. The guardrail is what Coldwire already does, checking every signal against a second source and a labeled history of false alarms before it ever reaches a person. And the trade-off is accepted on purpose: real hours of latency, a median of about a day, in exchange for an alert a procurement team can act on without checking it themselves first.

And if you want to be sure it really works, try it somewhere else

Same five letters, a claims desk instead of a supply chain, and the sanitized deck wears the exact same shape.

Threave runs Falconport, a tool that flags insurance claims likely to be fraudulent before an adjuster ever opens the file. A rival tool, Quickclaim, flags a claim within minutes of submission, with no cross-check against claim history or a second data source, while Falconport takes closer to a day because it checks each flagged claim against a labeled set of past confirmed fraud before it ever reaches an adjuster's queue. Threave's board, after a broker mentioned Quickclaim's speed on a renewal call, asked product manager Ottavio Brandywine the exact same question: why does a competitor ship AI features faster?

Hand sketched quadrant chart titled Sorting Threave's options for the same board question. X axis, cost to walk back, from cheap to hard. Y axis, how well checked, from guessed to tested. Promise faster flagging sits hard to walk back and guessed. Name the real tradeoff sits cheap to walk back and well tested. Quietly lower the bar sits moderate cost and guessed. Say nothing new sits cheap and guessed.
The one option worth keeping sits in the top left. The other three are cheap to say and cheap to be wrong about.

Same steps, mapped onto Threave. Outcome: protect whether a real fraud case actually gets caught before it pays out, not whether the answer sounds confident. Reversibility: a monthly, hands-on check of Quickclaim's real accuracy is easy to adjust later; promising the board a faster flagging queue by a set date is not, once claims staff start planning around it. Dependency: the flag only means something if Falconport is checked against real confirmed fraud, not a clean training set built from obvious cases. Evidence: Ottavio ran a two-week side-by-side on the same claim volume and found Quickclaim flagged claims in about four minutes with a confirmed-fraud rate near one in five, while Falconport's alerts landed within a day with a confirmed rate above eight in ten. Rank: name the tradeoff honestly, hold the line on the corroboration check, no promised timeline to match Quickclaim's speed, and commit to a real recheck of the rival's product every quarter, since a rival's own numbers can genuinely improve.

Swap the trigger and it still runs.
Speed: an interviewer caps you at ninety seconds. Skip straight to the rank: name the real tradeoff first, everything else is support.
Cost: there's no budget to run a full hands-on trial before the meeting. Say so plainly, and bring whatever's cheap and real instead, three public reviews naming the same problem, rather than pretending a guess is a number.
The model got better, for real: say the rival's next release actually closes the gap. Say that too, plainly, and change the rank. The method doesn't say never switch, it says decide from real evidence either way.

Where people run it wrong.
They answer the board's question by defending the roadmap instead of checking the premise.
They promise a timeline to sound decisive, before they have any evidence that timeline is the right one.
They let "we care about quality" stand in for a number, and a board can always tell the difference.

How to use it live. Before answering, ask out loud: "have I actually checked what their speed is costing their customers, or am I about to defend a habit?" If the honest answer is that you haven't checked, say you'll check, then go do it before your next answer.

Flashcards (tap any card to flip it)

1 · THE FRAMEWORK
What framework fits deciding what to actually tell a board?
Tap to flip
ANSWER
ORDER: outcome, reversibility, dependency, evidence, rank. Built for ranking real choices by what's hardest to undo, and picking what a board actually needs to hear is exactly that kind of ranking.
2 · THE CAST
Who holds each role in this story, and where do they work?
Tap to flip
ANSWER
Alba Thorndike is product manager for Coldwire at Otterburn Analytics. Faustine Corbenic chairs the board that asked the question. Merricat Isbell runs procurement at Ilkeston Outfitters, the customer who piloted Tidewatch first.
3 · THE OUTCOME
What does the board actually need from this answer?
Tap to flip
ANSWER
An honest, checkable read on the real tradeoff, not a defensive excuse and not an empty reassurance about caring more.
4 · REVERSIBILITY
Which claim is hardest for Alba to walk back once said to the board?
Tap to flip
ANSWER
A promise to match Tidewatch's speed on a set timeline. Naming the real tradeoff honestly is far easier to stand behind later, since it can update as the evidence does.
5 · THE OLD HABIT
What did Otterburn used to do instead of checking a rival's claims firsthand?
Tap to flip
ANSWER
Answered "why are they faster" questions from press coverage and a landing page, never from a hands-on test. Alba broke that habit by signing up for the rival's own trial herself.
6 · THE NUMBER
Fill in the blank: in one week, Tidewatch sent ___ alerts and ___ were real, while Coldwire sent ___ alerts and ___ were real.
Tap to flip
ANSWER
Tidewatch: 61 sent, 16 real. Coldwire: 19 sent, 17 real. Same fifty accounts, same week.
7 · THE RANK
State the final answer Alba gives the board, in one line.
Tap to flip
ANSWER
We ship slower because every alert clears a check they skip. Here's the hands-on number. No promise to match their timeline. The one thing worth speeding up is low-cost signal types.
8 · CROSS-PRODUCT TRANSFER
Section 4 answers the same board question for a different product. Which one, and who asks it?
Tap to flip
ANSWER
Falconport, Threave's insurance-claims fraud tool. Product manager Ottavio Brandywine runs the same ORDER method against a rival called Quickclaim.

Check yourself Score: 0 / 0

True or false
1. True or false: Alba's plan is to promise the board that Coldwire will match Tidewatch's alert speed by next quarter.
  • True
  • False
Show hint
Check the Reversibility step and stage 6 of the walkthrough.
Show answer
False. She rejects that promise on purpose, because it's the hardest claim in the whole answer to walk back if the evidence turns out to say otherwise.
Multiple choice
2. Why does Alba say Coldwire ships slower, in her own words to the board?
  • A. Otterburn's engineers work fewer hours than Tidewatch's.
  • B. Every alert clears a corroboration check Tidewatch skips.
  • C. The board asked her to slow the product down last quarter.
  • D. Coldwire doesn't have enough paying customers yet.
Show hint
Check stage 4 of the walkthrough and the direct answer.
Show answer
B. The speed gap comes from a real design choice, checking a second source before an alert ships, not from a resourcing problem or a board directive.
Fill in the blank
3. During Alba's one-week trial, Tidewatch sent ___ alerts, and ___ of them turned out to be real.
Show hint
Check the Evidence step in the framework recap.
Show answer
61 sent, 16 real. Roughly three out of every four alerts Tidewatch sent that week were false.
Short answer, name the rejected option
4. What option did Alba consider and turn down before the board meeting, and why?
Show hint
Check stage 6 of the walkthrough and the Reversibility step.
Show answer
Model answer: Promising the board a fixed timeline to match Tidewatch's speed. She turned it down because it's a promise she can't take back if it's wrong, and at the time she had no evidence that matching their pace was actually the right call.
Short answer, apply it yourself
5. Think of a product or service you use that seems to move faster than a rival's. What would you actually want to check before assuming faster means better?
Show hint
Check the Evidence step, what Alba actually went and got before saying anything.
Show answer
Model answer: Check the rival's real output quality yourself, cheaply, a trial account, a hands-on test, a handful of public reviews naming the same problem twice, rather than assuming speed alone means the faster option is actually better.
Short answer, work the number
6. If Tidewatch's false-alarm rate dropped from about 74% to something close to Coldwire's 8%, should Alba's answer to the board change? Why or why not?
Show hint
Check the Dependency step, what has to be true for speed to be the right thing to compare.
Show answer
Model answer: yes. The Dependency step's premise would flip. If Tidewatch got genuinely accurate as well as fast, "why are they faster" stops being a quality tradeoff and becomes a real question about whether Coldwire's gate is too slow. That's exactly why she committed to checking again every quarter, instead of treating this answer as settled forever.
Before you close the answer
Why this works
Tests whether a candidate will name an uncomfortable tradeoff honestly in front of leadership, or reach for a promise that sounds confident in the room and turns into a liability the moment someone actually checks it.
Follow-up traps
"Why not just speed up the whole eval gate to match them?" Response: because that gate is what's catching three out of four of Tidewatch's false alarms for their own customers. Speeding all of it up trades a real, measured cost for a comfort answer with no evidence behind it.

"Isn't this just an excuse for being slow?" Response: no, because the number came from a hands-on test on the same accounts, not a claim about company values, and the one place actually worth speeding up is named directly: low-cost signal types where a wrong alert costs an afternoon, not a factory shutdown.
If pressed
The corroboration check isn't one flat rule. It's a per-signal-type confidence bar, calibrated against how often that exact kind of signal has turned out to be a false alarm historically, which is why Coldwire can afford to move fast on some signal types and stay deliberately slow on others, instead of applying one delay to everything.
From U2xAI Academy

From answering questions to owning outcomes.

A live workshop where you ship a working AI agent, defend a launch decision, and walk away with a portfolio recruiters can't wave off, not just more questions to study.

  • A live AI agent you actually shipped
  • A launch decision you can defend under pressure
  • An interview-ready portfolio, not more flashcards
Know more