Explain how to compete when a model provider could ship your feature natively.

Every evening, before Parcelview built anything, a listing agent typed her own property descriptions from memory and a stack of photos. Parcelview is a real estate platform. ListLine is the part of it that drafts a listing description from those photos, checks the wording against fair housing rules, and pulls three nearby comps to suggest a price. Devon Okafor is the product manager who has to decide what ListLine is for, now that a major model provider's own consumer app can draft a decent listing caption for free.

How to answer this, stage by stage

Nobody is grading whether you know a model provider might ship your feature. They're grading whether you can name, out loud, which part of your product survives that day and which part doesn't.

Let's learn

Here is what happens when a company builds its pricing around the one feature that was always going to get cheaper to copy.

ListLine drafts a description from a stack of listing photos, checks the wording against fair housing rules, and pulls three nearby comps to suggest a price. Parcelview sold it as a standalone add-on for 49 dollars a month, and 1,400 of the platform's 3,600 active agent seats paid for it, about 39 out of every 100 seats, bringing in roughly 68,600 dollars a month on that line alone.

This is the whole evening ListLine touches. None of it needed a big model to begin with.

Here's the turn: a major model provider added "write this as a real estate listing" to its free consumer app. The caption it produced wasn't better than ListLine's. It was free, and it was good enough. Within eight weeks, monthly renewal among agents on the standalone add-on fell from 91 percent to 64 percent, a 27-point drop. The extra churn wasn't really about writing quality. It was about agents comparing the two tools on the one thing that's easy to compare side by side, and judging the whole product by the part it was about to lose.

At its worst, chasing the free tool on caption quality alone doesn't just cost engineering time nobody has. It teaches agents that the whole product is worth exactly what the caption box is worth, which is nothing, because a caption box is now free everywhere.

What I would leave alone: the model generating the first draft sentence doesn't need to change. A more expensive model wouldn't move renewal a single point; the agents who are staying are staying for the fair housing check and the comps, not a better verb choice.

The lesson: if your price tag sits on the one feature that's cheapest for someone else to copy, you didn't build a moat. You built a demo, and you priced it like a product.

Now here is the same thing as a story

The short version above is what you'd say defending this call to Devon's own VP. Read this one for how it actually looked from a phone in someone's blazer pocket.

The phone is cracked in one corner, and Camila Duarte keeps it in her blazer pocket between showings. She has sold houses for six years, and she can walk a buyer through a kitchen and know, before they say a word, whether they're picturing themselves there.

ListLine arrived in her account in the spring. For most of a year it was the best five minutes of her evening. She'd upload eight photos from a walkthrough, get a draft back in ten seconds, tweak two lines to sound like her, and post it. The comp pricing suggestion was usually close enough to save her a call to her broker. She paid the 49 dollars without thinking about it, the way you pay for parking.

One of these she notices and fixes in five minutes. The other one Parcelview doesn't notice for weeks.

Then, at a broker open house, a colleague she'd known for years asked her, half joking, "You're still paying for that thing? My assistant just uses the free one in her phone now, it's basically the same." Camila laughed it off. Then she opened the free tool that night, fed it the same eight photos, and read the caption twice. It wasn't as good. It wasn't bad either. It took four seconds and cost nothing.

She kept her ListLine subscription two more months, mostly for the fair housing check, which had caught a phrase in one of her drafts that would have gotten her broker a complaint letter. But she started writing captions herself again on the easy listings, the ones where nothing tricky needed checking, and only opening ListLine for the harder ones. Her renewal, when it came up, was a five-second decision either way. For hundreds of agents like her who'd never used the compliance check, it wasn't.

The caption sits in the one corner a free tool was always going to reach first.

Here is the decision Parcelview would take back. When ListLine launched, the team priced the caption generator as its own line item, separate from the compliance check, because the caption was the feature agents noticed first and would pay for on its own. That was a fine call with no free competitor in sight. It stopped being fine the day a free version of the exact thing being sold showed up inside a tool every agent already had.

Only one of these four parts was ever going to be free somewhere else.

Devon's team repriced ListLine as one bundle, no separate caption tier, the compliance check and the comp price built into the same seat cost. Camila's renewal came up again three months later. This time it wasn't a five-second decision about a caption. It was a two-minute conversation about whether she trusted the price suggestion more than her own gut, and she did, because it had been right on the last four houses. She renewed. The caption, free everywhere now, didn't come up at all.

PICK, the position that survives a free copycatNot a features race against a bigger model budget. PICK is what stops the fight from ever being about the caption.

The caption fell down the leftmost branch. The comps and the compliance check never did.

The recap, one line per letter: position is bundling the caption into the workflow instead of selling it alone, impact is naming that a dull caption is cheap while a churned seat is hidden and expensive, cost asymmetry is optimizing against the hidden one, and kill criteria is a renewal floor that forces a repricing conversation instead of another quarter of waiting.

And if you want to be sure it really works, try it somewhere elseSame four letters, a field-service company's furnace-diagnosis tool instead of a real estate platform. A different old decision breaks this one.

Ductra sells software to HVAC technicians, and one feature lets a technician photograph a furnace's error panel and get a plain-language diagnosis of what's likely wrong. A major model provider's photo-analysis feature can now do a rough version of that same read for free. Mapped onto PICK: position is bundling the photo diagnosis into the parts-ordering and warranty-lookup workflow, never selling "AI reads your furnace" as its own tier. Impact is naming that a technician second-guessing one photo costs five minutes; a technician who stops trusting the whole app and orders the wrong part costs a truck roll and an angry customer. Cost asymmetry says optimize against the truck roll, not the photo read. Kill criteria is watching whether technicians still open the parts-ordering screen even after they've started using a free photo tool for the first-guess diagnosis. The old decision here isn't a pricing call, it's a default setting: Ductra's app defaulted to showing the photo diagnosis first and the parts list second, which made sense when the diagnosis was the hardest part to build and wrong once a free tool made that same diagnosis unremarkable.

Same shape of picture, a different industry. The left side is always the part that gets free first.

Swap the trigger and it still runs.
Speed: an interviewer caps you at sixty seconds. Say "bundle the part that needs our own rules and data, never sell the part a free tool already does," and stop.
Cost: there's no budget to rebuild the pricing page before the next renewal cycle. Say so honestly, and start by moving the compliance check into the free tier as a hook, holding the comps behind the paid bundle where the real value sits.
The model gets better, for real: if the free tool's caption quality keeps improving, that's not a threat to the bundle, it's free marketing for how normal AI-written captions have become, which makes the compliance check look more essential, not less.

Where people run it wrong.
They price the flashiest feature on its own, because it demos the best, not because it's the hardest to copy.
They wait for a full quarter of renewal data before admitting the standalone tier is bleeding, instead of setting a kill line in advance.
They spend engineering time trying to out-write a free model instead of watching what agents actually keep paying for.

How to use it live. The moment someone describes a feature a model provider could ship natively, ask yourself: if I stripped this down to only the part that needs our own data or our own rules, what's left? Build the pitch around what's left.

Every point on this line is the same eight weeks, told as a decision instead of a slow surprise.

Flashcards (tap any card to flip it)

Check yourself Score: 0 / 0