Price a product I describe, and defend the model you chose.
PICK · pricing an AI expense auditor, live, off one page of numbers
Ainsley Dunraven, who owns pricing at Redgate Systems, does this to every senior candidate now. Forty minutes into the interview, she puts the rehearsed questions away and hands over a single printed page: a product called FlagPoint, a real buyer's numbers, and a challenge to price it out loud, then defend the model against the one everyone in this category already uses. Ilona Faulkstone is the candidate in the room this time. What she doesn't know yet is that the page describes the exact mistake Ainsley made herself, three years ago.
The direct answer
Price FlagPoint on the volume of expense line items it actually scans each month, banded into tiers, not on how many employees are licensed to use it. A seat price can't see the one number that decides both Redgate's real cost and the buyer's real fraud risk: how many items the model can't clear on its own and hands to a person. Charge by seat, and the bill stays flat straight through a volume surge that quietly buries a six-person review team, and a buried review team is exactly where the real fraud gets through, not in the model.
Do this, in order
Price FlagPoint on expense line items scanned a month, in bands, not per employee seat.Why: it's the only number that tracks what actually drives cost on both sides, the compute and the human review it triggers.
Track the escalation rate as its own number, separate from raw volume.Why: a steady 15 percent rate on a growing item count is what actually eats a review team's hours, not the item count by itself.
Set a tier-crossing rule: flag a capacity conversation the moment a customer's real usage crosses into the next band for two straight months.Why: Winterbrook's invoice never moved during its surge. A rule like this would have, weeks before the backlog did any damage.
Never bill per item escalated to a human.Why: that pays FlagPoint's own model to flag more, and the buyer's audit team eats the extra hours, not Redgate.
Don't force item pricing onto a customer with genuinely flat, provable usage.Why: for an account whose volume hasn't moved in a year, a simple flat band is fine, and metering it more finely just adds friction with nothing to show for it.
Don't fix a capacity crunch by telling the audit team to work faster.Why: Winterbrook's team wasn't careless. They were reviewing a workload nobody's pricing had ever warned them was coming.
How to answer this, stage by stage
Nobody is grading whether Ilona can do the arithmetic. They're grading whether she reaches for the price everyone in this category already charges, out of habit, or actually checks what FlagPoint's own judgment costs someone, in real hours.
1
Read the page back before answering it
Say it like this
"Okay, let me make sure I've got this right. FlagPoint reads a receipt, checks it against travel policy, and marks each line item clean, needs a look, or high risk, before it ever reaches a manager. The buyer is nine hundred employees, one finance team of six doing the actual reviewing, about fourteen thousand line items a month, and sixty of those employees file forty each, every month, on their own."
Why this works
Repeating the numbers back buys real thinking time without one second of dead air, and it proves she's working from the page, not a script.
2
Say the plan out loud before naming a price
Say it like this
"I'm going to run PICK. Commit to a position first, say who feels each kind of pricing error, name which one is cheap and which one hides, then tell you what would change my mind."
Why this works
Naming the method signals a plan already in motion, not four ideas arriving in whatever order they occur to her.
3
Reframe what the page is actually testing
Say it like this
"This isn't really asking me to name a dollar figure. It's asking whether I'll reach for the price every expense tool has always charged, per employee, or check whether that number has anything to do with what FlagPoint actually costs to run."
Why this works
Stops the shallow answer that quotes a per-seat number confidently and never asks what a seat has to do with the product's real cost.
4
Commit, in one breath, before any reasoning
Say it like this
"Here's my position: price it on line items scanned, banded by monthly volume, something like thirty cents an item. Not per seat. At fourteen thousand items a month that's around forty two hundred dollars, close to what nine hundred seats at five dollars each would run anyway, so nobody's shocked on day one."
Why this works
This is the direct answer, said as a real number, before the interviewer has to wait for the reasoning to find out what she'd actually do.
5
Name who feels each kind of error, in real units
Say it like this
"Under a seat price, the sixty reps filing forty reports a month cost the same as employees who file two a year. Winterbrook pays for badges, not usage. Under item pricing, the bill tracks the fourteen thousand items they're actually running through it, something their own finance team can forecast against their own travel calendar."
Why this works
Naming both sides feeling something concrete is what keeps this from being a preference dressed up as an argument.
6
Name the asymmetry, with the number that makes it real
Say it like this
"A slow month under item pricing costs Winterbrook a little more than it should, visible on the invoice, easy to true up at renewal. A seat price that doesn't move when volume jumps sixty four percent is hidden and expensive. Their six-person team ran short about sixty hours a month for a whole quarter, and a batch of client-dinner charges got waved through during the crunch, twenty three hundred dollars, then again, nine thousand four hundred total, before a routine sample audit caught the pattern."
Why this works
This is the hardest move in PICK, and the one that actually decides the pick. If both errors cost about the same, there's no real tradeoff underneath the preference.
7
Say what would flip the pick
Say it like this
"I'd go back to a flat seat price only if Winterbrook's volume held within about ten percent of fourteen thousand for a full year, and only if Redgate could show, audited every quarter, that the escalation rate stays inside a fixed band no matter how volume moves. Without both of those, keep it on items scanned."
Why this works
Naming the exact evidence that would change her mind is what separates a confident pick from a stubborn one.
8
Close on the one line
Say it like this
"So: price it by the item, banded by volume, not by the badge. The moment you charge per seat, the invoice stops being able to see the one thing that actually drives risk here, how many items a person has to look at with their own eyes."
Why this works
Restates the position and the reason in one breath, so the interviewer leaves with the decision, not just the story behind it.
Let's learn
What happens the first time a price tag has to describe something a machine decides, one receipt at a time?
FlagPoint is the tool inside Redgate Systems that reads an expense report before a manager ever sees it: it reads the receipt, checks it against travel policy, per diem limits, cabin class, alcohol caps, and marks each line item clean, needs a second look, or high risk.
Five steps. The fourth one, the confidence score, is the step that quietly decides everyone's workload downstream.
Knowledge spark: what does a confidence threshold actually do?
FlagPoint doesn't just say "fraud" or "not fraud." It scores how sure it is, against a set of past reports someone has already checked by hand. Above a cut-off, the item clears itself. Below it, a person looks. Move that cut-off a little, and the number of items landing on a human's desk moves a lot.
Before FlagPoint, Winterbrook Freight's six-person finance team could only spot-check about one report in twelve out of the fourteen thousand filed every month, and wave the rest through. There wasn't a real alternative. Checking every line item by hand, even a light first pass, would have eaten close to three quarters of the team's entire month, leaving nothing for payroll audits or month-end close.
One in twelve, checked by eye. The other eleven, waved through because there was no other way to keep the month moving.
With FlagPoint running, the model clears about 85 percent of items on its own and escalates the rest, roughly 2,100 a month, to one of the six analysts, at about six minutes a look. That's around 210 hours a month, a chunk the team's usual schedule had room for, and for the first time close to all of the risky 15 percent got a real look instead of one in twelve of everything.
We didn't charge Winterbrook for what FlagPoint actually did. We charged them for how many people were allowed to open it.
Here's the turn. FlagPoint's own accuracy was never the problem, then or later. What mattered was a number nobody had priced at all: how many of those fourteen thousand items needed a person, and whether the six-person team still had room for that number once it changed. A seat price can't see that number. It only knows how many badges are on file.
Sixty employees sit in the corner that actually drives cost. A seat price draws one flat line straight through both clusters and charges them the same.
At its worst, the number moved and nobody's invoice said a word. Winterbrook opened a new regional hub, and travel filings for that unit climbed for a full quarter, pushing total volume from 14,000 to roughly 23,000 items a month. Escalations, holding at the same 15 percent, rose from 2,100 to about 3,450 a month, needing 345 hours of review against a team whose usual slack, once month-end close and payroll audits were accounted for, was only about 288. That's a shortfall of roughly 60 hours every month, for the whole quarter the hub took to ramp up.
The choice that mattered
Redgate priced FlagPoint's early rollout at Winterbrook flat, per seat, with no tie at all to how many items actually ran through the model. Nobody decided on purpose that a volume surge should be invisible on the invoice. It just was, because the invoice was never built to see it.
What I'd leave alone: FlagPoint's own classification quality. Its clear-versus-escalate split held at the same 85/15 the entire time, before the surge and during it. The fraud that got through didn't get through because the model got worse. It got through because the human layer behind the model quietly ran out of room, and nothing in the pricing had ever warned anyone that room would run out.
The lesson: a price that can't see the thing actually driving your customer's cost isn't neutral. It's a blind spot with an invoice attached, and it stays invisible for exactly as long as the number underneath it agrees to behave.
Now here is the same thing as a story
Read the short version above when you're in the room. Read this one when you want to feel why a bill that never changed could still let nine thousand dollars of client-dinner receipts through untouched.
Ainsley Dunraven has priced every product Redgate has shipped in six years, and there's one number she checks before she trusts any new one: what happens to this price the day the customer's usage stops behaving. She built that habit the hard way, on FlagPoint's very first enterprise rollout.
Winterbrook Freight signed on early, nine hundred employees, and Ainsley priced it the way every expense tool in the category always had: five dollars a seat, licensed by employee, nine hundred of them, forty five hundred dollars a month, flat. It was an easy number to say on a sales call and an easy one for Winterbrook's procurement team to approve. For the better part of a year, it was also a perfectly fine number. Volume sat around fourteen thousand items a month. The six-person finance team, led by Bronte Pennstock, cleared their queue most days with time to spare, and Bronte told anyone who'd listen that FlagPoint was the first tool in a decade that hadn't made her team's job harder.
It thinned in three quiet beats. Beat one: Winterbrook won a new freight contract and opened a regional hub, and travel filings from that unit started climbing, a few hundred more items a week, nothing anyone would call a spike. Beat two: a month in, escalations were running noticeably higher, but the team was still clearing them, just later in the day, and later felt like a scheduling problem, not a warning. Beat three: by the second month, Bronte's team was starting each morning already behind, and the backlog carried into the next day more often than it didn't.
The trigger wasn't a report landing on anyone's desk. It was a Friday when three different employees emailed Bronte, separately, asking where their reimbursements were. She pulled the queue and found 340 unreviewed escalations sitting in it, some of them nine days old.
No single bad week. A slow climb in volume, then a Friday when the backlog finally had a number attached to it.
Here's what Bronte did next, and it was the sensible thing every step of the way. With nine days of reimbursements sitting unopened and more arriving daily, she told her team to clear the backlog fast: anything FlagPoint had marked "needs a second look," rather than "high risk," got a quick skim and a stamp, reserving real scrutiny for the smaller high-risk pile. It was the only way to get nine employees their money back that week without falling further behind on the next nine.
The invoice never moved. It was the only number in the whole building that didn't.
Six weeks later, a routine quarterly sample audit, unrelated to any of this, pulled twenty random "needs a second look" approvals from that backlog sweep for a second opinion. One of them was a client-dinner receipt from a restaurant that, cross-checked against the client's own calendar, had no meeting on that date. It had been approved in the sweep, along with five more from the same employee across the same six weeks, totaling $9,400. FlagPoint had flagged every one of them correctly, at "needs a second look." Nobody had actually looked.
This is the whole story in one picture. One number held still on purpose. The other one didn't have that option.
The decision that opened the door traced back to the original pricing call, months before Winterbrook's contract even started. Ainsley had chosen per-seat pricing because it was simple to sell and simple for procurement to sign off on, and at the volume Winterbrook was running when the deal closed, it happened to land close to what item-based pricing would have charged anyway. Nobody in that early meeting asked what the number would do if volume ever moved. It didn't come up, because nothing about a flat monthly seat count seemed like the kind of thing that needed a second thought.
Run the rollout again with one change: FlagPoint priced by item scanned, banded by monthly volume, from day one. The bill at fourteen thousand items lands close to the old seat price, so nobody notices a difference at signing. But the moment volume crosses into the next band for two months running, Redgate's own systems flag it, automatically, well before any backlog forms. Winterbrook's finance lead gets a call about adding review capacity in month one of the hub opening, not a Friday full of angry emails in month three. The same client-dinner pattern still gets flagged by FlagPoint. This time, someone has the hours left to actually open it.
What Ainsley would tell herself, back in that early pricing meeting: a price that happens to be right at today's volume isn't the same thing as a price that's built to notice when volume changes. She'd priced the seat. She should have priced the thing the seat was standing in for.
PICK, run cold, off one page of someone else's numbers
Not a coin flip between "charge per person" and "charge per use." PICK is what forces a real answer to which mistake actually costs someone, instead of settling for whichever price is easiest to say out loud on the spot.
PPosition. Your pick, in one sentence, before any reasoning.
Item-volume pricing: about thirty cents per expense line item scanned, banded by monthly volume. Not a per-seat license across all 900 employees. At Winterbrook's normal 14,000 items a month, that lands around $4,200, close enough to a seat-based number that nobody's surprised at signing.
Say the pick first. An interviewer who has to wait through the reasoning to learn what you'd actually do has already marked this "it depends."
IImpact. Who feels each kind of error, and in what units?
Under a seat price, the sixty reps filing forty reports a month cost Winterbrook exactly what an employee filing two a year costs, so the price never signals which group is actually driving the model's workload. Under item pricing, Winterbrook's bill tracks their own real usage, forecastable against their travel calendar, and Redgate's revenue tracks the growth it's actually serving instead of leaving expansion on the table.
Name both sides feeling something real, or the impact step is just restating the position with different words.
CCost asymmetry. Which error is cheap and visible, which is hidden and expensive?
A slow month under item pricing costs Winterbrook a little more than it should, visible on the invoice, negotiated at renewal, nobody's fraud risk touched. A volume surge under seat pricing costs nothing that the bill can see, and then costs a six-person team about 60 hours a month it doesn't have, for a whole quarter, until a backlog forces a batch-approval sweep and $9,400 in receipts that were flagged correctly slip through anyway. Optimize against the second one. Nobody notices it by accident.
This is the hardest step, and the one that actually decides the pick. If both errors cost about the same, there's no real tradeoff here, just a preference.
One of these mistakes gets fixed at the next renewal. The other one gets fixed six weeks after a quarterly sample audit finds it.
KKill criteria. What evidence would flip the pick?
Two things would both have to hold before Redgate goes back to a flat seat price: Winterbrook's monthly volume staying within about 10 percent of 14,000 for a full year, and Redgate proving, audited quarterly, that FlagPoint's escalation rate holds inside a fixed band no matter how volume moves. Short of both, a seat price is a bet that usage never changes, and Winterbrook's own hub just proved it does.
Naming the number that would change your mind is what separates a confident pick from a stubborn one.
Cost, by the numbers: seat-based bill vs. item-volume bill, normal month and surge month
Seat-based, flat, 900 seatsItem-volume, normal monthItem-volume, surge month
At normal volume the two prices sit close together. The moment volume jumps 64 percent, the seat price doesn't move at all, and the gap is exactly the growth Redgate's own pricing failed to notice.
The kill line: review hours needed vs. the team's real slack, across the hub's ramp-up quarter
Inside the team's slackPast the team's slack
Review hours cross the team's real slack between month one and month two, a full month before Bronte's Friday backlog crisis, and about six weeks before the audit found the $9,400 pattern.
Three things worth stating directly, since this is where the real judgment sits. Redgate actually considered a second alternative to item-volume pricing and rejected it too: billing per item escalated, paying only for what a human actually had to review. It sounds fair, pay for the work that costs money, but it hands FlagPoint's own model a billing reason to loosen its confidence threshold and escalate more, at Winterbrook's expense in analyst hours, not Redgate's, the same reward-hacking risk any tool runs the moment its own judgment decides what gets billed. The AI-specific failure worth naming is a silent capacity collapse behind a threshold nobody was pricing: FlagPoint's classification never degraded, its 85/15 clear-escalate split held the entire time, and the guardrail isn't a better model, it's treating the escalation rate as its own leading number, watched separately from raw volume, with a tier-crossing rule that starts the capacity conversation before the backlog does. And the trade being accepted on purpose: FlagPoint's threshold is tuned to score a report in real time as it's submitted, not to run a slower, more exhaustive second pass, because a model that took a full minute per report would back up the queue on its own, before volume ever spiked. Redgate accepts a slightly higher escalation rate than a slower model might produce, in exchange for a tool that never becomes the bottleneck itself.
And if you want to be sure it really works, try it somewhere else
Same four letters, a pet insurer instead of a freight company, and this time the surge isn't a new office opening. It's a heatwave.
ClarionClaims is Meridyne Health AI's tool: it reads a veterinary insurance claim, checks the procedure codes and vet notes against the policy and the usual cost band for that treatment, and classifies it auto-approve, request more information, or flag for a human adjuster. Kioni Solarin owns pricing on it, for Alder Cove Pet Insurance's book of claims.
The decision Kioni would take back
Alder Cove's early contract priced ClarionClaims per adjuster seat, four of them, a flat number that had nothing to do with how many claims actually moved through the model in a given month.
Knowledge spark: what's upcoding, for a vet claim?
Billing a cheaper, routine procedure under the code for a more expensive one, a basic dental cleaning submitted as oral surgery, say. It's not always obvious from the claim alone, which is exactly why it needs a person, not a policy check, to catch.
Alder Cove runs about 9,000 claims a month normally, auto-approving 80 percent and escalating the rest, around 1,800 a month, to its four adjusters. A regional heatwave pushed claims to 15,000 a month for six weeks, mostly heatstroke and related emergency visits, and escalations climbed to about 3,000. The four-person team fell behind, and a request-more-information queue meant for follow-up got cleared in a rush the same way Winterbrook's did: skimmed, not read. A pattern of dental-cleaning claims billed as minor oral surgery, from a small cluster of clinics, went through for six weeks, about $61,000 total, before a routine reinsurer audit flagged it.
Same shape as Winterbrook's story, a different surge entirely. A seat price can't tell a heatwave from a quiet Tuesday.
Same rank, different lever: at Winterbrook the lever was a new regional hub. At Alder Cove it's weather, something no pricing meeting would ever think to model directly. Different trigger, same asymmetry: whatever a seat price can't see, it can't warn anyone about, whether the cause is a business decision or a heat index.
Swap the trigger and it still runs.
Speed: an interviewer caps you at ninety seconds. Skip straight to the position: price by the item scanned, banded by volume, because a seat count doesn't move when the thing driving real cost does.
Cost: no budget this quarter to build new metering. The cheapest version is a monthly manual reconciliation against logs you already have, not a new billing pipeline.
The model got better, for real: say FlagPoint's escalation rate drops from 15 percent to 6 percent overnight, a genuinely better model. Keep pricing by volume anyway. A better model lowers everyone's review-hour cost, it doesn't make a seat count correlate with usage.
Where people run it wrong.
They price on what's easy to count, seats, licenses, instead of what's actually expensive to run.
They watch a flat invoice and call it predictable, without checking whether the thing it's supposed to track can still move underneath it.
They fix a capacity scare by asking the review team to work faster, instead of fixing the price signal that hid the surge in the first place.
How to use it live. Ask what actually costs the vendor money before naming a price: "is this priced on what's easy to count, or what's expensive to run?" That question alone buys a breath, and it's usually the one the interviewer actually wants to hear asked.
Flashcards (tap any card to flip it)
1 · THE FRAMEWORK
What framework is this, and what's its one job?
Tap to flip
ANSWER
PICK: commit, then show the asymmetry. Built for tradeoff and pricing questions, especially the kind handed to you cold with no time to prepare.
2 · THE PERSON
Who is this answer about?
Tap to flip
ANSWER
Ainsley Dunraven, Head of Pricing at Redgate Systems, who owns FlagPoint's pricing and made the original seat-based call at Winterbrook Freight herself.
3 · THE ASYMMETRY
Which pricing error is cheap and visible, and which is hidden and expensive?
Tap to flip
ANSWER
Paying a little more in a slow month under item pricing is cheap and visible, fixed at renewal. A seat price that doesn't move during a volume surge is hidden, and costs a six-person team about 60 hours a month it doesn't have, until $9,400 in flagged receipts slip through anyway.
4 · THE POSITION
What did Ilona actually commit to, live, in the room?
Tap to flip
ANSWER
Item-volume pricing: about thirty cents per expense line item scanned, banded by monthly volume, roughly $4,200 a month at Winterbrook's normal 14,000 items. Not a per-seat license across 900 employees.
5 · THE OLD DECISION
What decision would Ainsley take back?
Tap to flip
ANSWER
Pricing FlagPoint's Winterbrook rollout flat, per seat, with no tie at all to how many items actually ran through the model each month.
6 · THE NUMBER
Fill in the blank: when volume grew from 14,000 to 23,000 items a month, escalations climbed from ___ to ___ a month, against a team whose real slack was only ___ hours.
Tap to flip
ANSWER
2,100 to 3,450 a month, against about 288 hours of real slack, a shortfall of roughly 60 hours every month for the whole quarter.
7 · THE REPLAY
Same hub, same surge, new pricing, what changes?
Tap to flip
ANSWER
The bill actually moves when volume crosses into the next tier for two straight months, which triggers a proactive capacity conversation in month one, not a Friday full of angry emails in month three.
8 · CROSS-PRODUCT TRANSFER
Section 4 answers this same question again for a different product. Which product, and what's the different lever?
Tap to flip
ANSWER
ClarionClaims, Meridyne Health AI's vet insurance claims auditor, at Alder Cove Pet Insurance. The lever there is a regional heatwave spiking claims, not a new office opening.
Check yourself Score: 0 / 0
True or false
1. True or false: the client-dinner receipts got through because FlagPoint's fraud classification got less accurate during the surge.
True
False
Show hint
Check "what I'd leave alone" in Let's learn.
Show answer
False. FlagPoint's clear-versus-escalate split held at 85/15 the entire time. The receipts were flagged correctly. The team simply didn't have the hours left to open them.
Multiple choice
2. Which pricing model did Ilona actually commit to for FlagPoint at Winterbrook?
A. Per employee seat, across all 900 licensed users
B. Per expense line item scanned, banded by monthly volume
C. Per item escalated to a human reviewer
D. A flat annual enterprise fee regardless of volume
Show hint
Check the direct answer, and the P step in the PICK recap.
Show answer
B. Item-volume pricing, roughly thirty cents a line item, banded by how many run through the model each month.
Fill in the blank
3. An escalation costs Winterbrook about $___, based on 6 minutes of review time at $42 an hour fully loaded.
Show hint
It's stated early in Let's learn, right where the interviewer's numbers are described.
Show answer
$4.20. Six minutes at $42 an hour, which is why 2,100 escalations a month adds up to real hours, not just a small line item.
Short answer, name the rejected alternatives
4. Ilona rejected two pricing models, not just one. Name both, and why each one fails here.
Show hint
Look at the paragraph right after the K step in the PICK recap.
Show answer
Model answer: Per-seat pricing, rejected because it charges the 60 heavy-filing sales reps the same as employees who barely use the tool, so the invoice never signals real usage. Per-escalation billing, rejected because it pays FlagPoint's own model to flag more, at Winterbrook's expense in analyst hours, not Redgate's.
Short answer, apply it yourself
5. Think of a subscription tool you use that could be priced either per seat or per how much you actually use it. Which pricing would hide the real driver of the vendor's cost, and how would you tell?
Show hint
Think about what actually costs the vendor money to run, compute, storage, human support, and whether seat count tracks that at all.
Show answer
Model answer: A cloud storage app priced per user seat hides how much data any one user actually stores. A team of five could store 20 gigabytes or 20 terabytes for the same price, so the invoice tells the vendor nothing about their real infrastructure cost. You'd tell by checking whether the price ever changes when your actual usage does.
Fill in the blank, work the number
6. If Winterbrook's volume had still grown from 14,000 to 23,000 items, but FlagPoint's escalation rate had dropped to 9 percent instead of holding at 15, would the kill criteria still call for dropping seat pricing?
Show hint
Recalculate escalations at 9 percent of 23,000, and compare to the team's 288-hour slack.
Show answer
Not on capacity grounds alone. 9 percent of 23,000 is about 2,070 escalations, close to the original 2,100, needing around 207 hours, comfortably inside the 288-hour slack. The capacity emergency wouldn't happen. But the volume growth itself is still unpriced under a seat model, so the forecasting case for item pricing still holds, even without a crisis attached to it.
Once the answer's out of your mouth
Why this works
Tests whether Ilona reaches for the price everyone in the category already charges, out of habit, or actually finds what FlagPoint's own judgment costs someone, in real hours, when usage stops behaving.
Follow-up traps
"Isn't item-based pricing just going to make Winterbrook's bill unpredictable and annoying?" Response: less unpredictable than a flat bill hiding a real capacity crisis. Banded tiers with a proactive notice at each crossing keep it forecastable while still tracking real cost.
"Why not just price per escalation, since that's literally the thing costing money?" Response: that hands FlagPoint's own model a billing reason to escalate more, and Winterbrook's team eats the extra hours, not Redgate. Keep the threshold's incentive clean of the invoice.
If pressed
The tier-crossing rule isn't just a bigger invoice showing up automatically. Redgate's contract requires a proactive capacity call with the customer's finance lead within 10 business days of a second straight month in a new band, before the next invoice even goes out, not after.
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.
