> ## Content Index
> Fetch the complete content index at: https://www.vysora.co/llms.txt
> Use this file to discover other available public pages before exploring further.

# We Sold Judgment. Founders Got the Plan. AI Amplified Both.
- URL: https://www.vysora.co/we-sold-judgment-founders-got-the-plan-ai-amplified-both/
- Published: 2026-11-10T07:30:00.000Z
- Updated: 2026-11-10T07:30:11.000Z
- Description: In sales, our “no” questioned the founder’s plan. In delivery, it protected ours. As AI pushes agencies to sell judgment, that difference becomes the business model.
- Author: Enes Lisovac
- Tags: AI Downstream, Agencies, Founders, Product judgment, Product development

"Enes, where's the judgment I bought?"

A client asked me that halfway through an engagement.

During sales, I'd won her trust by telling her what not to build. I challenged her assumptions, questioned her business model and product priorities, and recommended spending less, even when the agency could have earned more. That conversation closed the deal.

Then the contract was signed, and the delivery team took over. They were capable people, doing good work. Here's how she described what she was getting:

"No, that's outside the scope"

"No, you already approved that design"

"No, that would introduce security risks"

Different reasons. The same answer.

Her question named something I'd seen for years. In ["We Sold Projects. Founders Needed a Loop"](https://www.vysora.co/we-sold-projects-founders-needed-a-loop-ai-changed-who-owns-it), I wrote about how delivery turns learning into change requests. 

In ["We Sold More. Founders Needed Less"](https://www.vysora.co/we-sold-more-founders-needed-less-ai-changed-the-math), I wrote about the economics that keep agencies selling effort. This piece is about the way out many agencies are now reaching for: 

*selling judgment instead.*

In my experience, that bet fails twice. And AI makes both failures faster.

---

## Two Kinds of No

Before the contract, my "no" meant: I don't think you should spend your time and money on that.

During delivery, "no" meant: we can't change what we agreed on, unless you pay.

One questions whether a decision is right. The other protects a decision already made.

Both can be valid. Scope protects a budget the founder agreed to. An approved design keeps a team from redoing the same work every week. A security objection can be the most important thing a team ever says to a founder. A delivery team that said yes to everything would serve her worse.

But the two do different jobs. The first “no” applies judgment to the founder’s decisions: whether something is worth building, what should come first, how to build it without spending more than the stage justifies... The second applies judgment to the agency’s commitment: does this fit what we agreed, the scope, the timeline, the budget, and the margin we priced in?

The first protects the founder’s runway. The second protects the plan, and the agency’s time and money along with it. And that plan is often written before any user touched the product.

You can usually tell them apart by what comes after. 

The first kind of "no" comes with a question or an alternative: what are you actually trying to learn, and is there a faster and cheaper way to learn it? The second ends the conversation. Even a valid security concern can be either. "That would introduce security risks" closes the topic. "That would introduce security risks, so here's a safer way to get what you need" keeps judgment in the room.

She'd bought the first kind. What she got, more and more, was the second.

---

## Why Judgment Stayed in the Sales Room

I saw this pattern across the agencies I worked for. We earned trust by challenging the initial plan, then organized delivery around executing the revised one.

The people in delivery weren't the problem. The work was set up that way.

The judgment that got us hired lived in one or two people's heads. It worked when they were involved. But their time paid off most before the contract, so that's where it went. Once a deal closed, they moved on to the next one.

Delivery inherited the plan, not the reasoning behind it. A proposal records what was decided. It rarely records what was rejected, why, and what evidence would change the decision. So the team receives a scope without the judgment that shaped it. And the only safe way to treat a scope you don't fully understand is to protect it.

The teams were also built for execution. Strong engineers, good designers, a delivery manager to keep it on track. Some had real product judgment, but that's rarely what they were hired or staffed for. Senior product people are expensive, and the margin model put them where they paid off most: in sales, and spread across several accounts at once.

And when someone in delivery did see that the plan was wrong, they rarely had the authority to change it. Changing direction meant a change request, a conversation with the account manager, sometimes with sales or leadership, and a commercial negotiation with the founder. The person who saw the problem was almost never the person who could decide. Most people learned that raising it created work for everyone, themselves included.

Then delivery is measured on exactly that: on time, on budget, as specified. In the previous article, I wrote that nobody gets a bonus for the phase a founder didn't buy. Nobody in delivery gets one for reopening a decision the founder already approved, either. Questioning the plan can even look like questioning the sale, or the client.

So the judgment didn't disappear. Where it lived in one or two heads, it stayed before the contract, where it helped win the deal and cost the agency nothing. Where it showed up in delivery, it had no authority to act. Either way, it didn't scale.

---

## The New Bet

Now AI is absorbing more of the execution work, and agencies are making a new bet: 

*we'll sell judgment.*

It's a reasonable move. If AI makes effort cheap, value moves to what effort never captured: knowing how to validate, what to build, what to cut, what will break... what the founder hasn't seen yet. Effort was primarily how agencies got paid.

But the bet assumes two things. 

First, that the judgment exists across the team, not just in one or two people. 

Second, that it survives the contract: that it's still exercised when it's inconvenient, after the plan is approved and the budget and timeline are committed..

The old model never had to test either assumption, because nobody sold judgment directly. It came bundled with the effort, strongest at the start and thinning out as the engagement went on. Selling it on its own means it has to show up in every phase, including the ones where it costs the agency something.

I wanted to see what AI does to both assumptions.

---

## What Happened When I Tested It

Over four months, I tested that bet with five delivery teams. What happened depended less on the tools than on who was in the room.

With experienced product people running the work, AI gave them more ideas to kill. One team generated 12 approaches in a couple of hours, then spent days arguing over which ones were worth pursuing. They ruled out 10\. Without AI, they said they’d have had two approaches at best, and far less worth arguing about.

That's what judgment looks like when options get cheap. Generating alternatives costs almost nothing, so more of the time goes into rejecting them. The experienced teams didn't ask whether an option was good. They asked whether it was right for this product and its customers, at this stage: what it would teach them, who would actually use it, how to validate it before building it, what it would cost to undo, what would happen if they didn't build it at all... Most options AI generates are good in general. Very few are right for now. The experienced teams could tell the difference, and AI gave them more material to apply it to.

Where less experienced people ran it, exploration went the other way. They judged each option on its own merits, and on its own merits almost every option looked good: a competitor feature to match, an integration competitors don't have, an AI feature nobody else offers yet, a second user role, an edge case to cover... Each made sense in isolation, so each became a reason to add. Then they'd combine them, a few, sometimes many, and go all in. Little got killed. Most of it got combined.

It's easy to see why. Without a sense of what stage the product was in, and what it most needed to learn next, adding felt safer than cutting, because nobody gets blamed for covering an edge case. And combining options now cost almost nothing, so the old brake, "we can't afford to build all of this", stopped working.

Then the build phase started, and all five teams ended up in the same place. 

Even the teams that had argued hardest about what to build started defending what they’d chosen. Every "no" became more convincing. Every existing decision became easier to justify. Every questionable plan became faster to execute.

Here, too, AI did what it was asked. 

Asked whether a change fits the scope, it finds the reasons it doesn't. 

Asked to defend a decision already made, it builds the case. 

As I wrote in the [first article](https://www.vysora.co/we-sold-projects-founders-needed-a-loop-ai-changed-who-owns-it) of this series, it tends to agree with the hypothesis you bring to it. In delivery, the hypothesis is the plan.

---

## The Bet Fails Twice

AI amplifies what's already there: 

*the quality of the thinking, and the incentives of the phase.* 

Early on, it amplifies the thinking. Later, it protects the plan.

So the bet fails twice: once quietly, once loudly.

The quiet failure happens before the build. Less experience turns exploration into scope, and only a few people can tell. The founder sees a team that explored widely and a plan that covers everything. It looks like more judgment, not less. The cost shows up months later. Money is the visible part: a runway that paid for everything. The bigger cost is learning. When a first version does ten things, it's harder to tell which one users actually came for, or which one drove them away. And when the answer finally arrives, there's far more to change, and far less money and time left to change it.

The loud failure happens during delivery. Even good judgment gets absorbed into defending the plan, and everybody can tell. The founder feels it every time a question gets a "no" before it gets a hearing. That's when someone asks: 

*where's the judgment I bought?*

Neither failure started with AI. Both were there before it. AI makes them faster and more convincing: more options to combine before the build, better arguments for the plan during it.

An agency selling judgment has to survive both. In the first, experience is scarce. In the second, the incentives win. Announcing "we sell judgment now" fixes neither.

---

## What to Ask Before Paying for Judgment

For founders, the practical question is how to tell judgment from a pitch for it. 

A few questions make the difference visible, and most are worth asking before you sign:

- Who from the sales conversation will still be working on your product, and for how many hours a week? "Available when needed" usually means not available.
- Who on the delivery team has the authority to say “we shouldn’t build this anymore”, and make it stick? And when they do, does the contract treat the cut as a decision or as a change request?
- Can they show you a time a team changed direction mid-delivery because of user evidence, and what happened to scope and price? If every example is a change request, you know which "no" you'll get.
- When the team says no, what does it offer instead? Judgment comes with a question or an alternative. Protection points at the contract.
- How many options did they consider, and why did they reject the ones they did? A plan with no rejected options isn't thorough. Nobody judged it.

For agencies, the questions are harder, because the honest answers cost something. I'll leave them as questions:

- If your two best product people left tomorrow, which clients would notice first: the ones you're selling to, or the ones you're delivering for? Whichever it is, that's where your judgment lives.
- Look at your last five proposals. How many record what you advised the client not to build, and why? If none do, delivery starts without the reasoning that won the deal.
- When did someone in delivery last stop work because the evidence changed, and what happened next: recognition, a change request, or a difficult conversation? If nobody can remember, the team has already learned the answer.
- Take one "no" your team gave a client last month. Did it protect the client's runway, or your margin? AI will make both sound equally convincing, so if you can't tell, neither can the client.
- In your next discovery, count the options generated and the options killed. If the plan kept most of it, AI amplified the exploration, not the judgment.

These are the questions an agency has to answer before “we sell judgment” means anything. And answering them is only the start. 

Selling judgment for real changes how an agency prices, contracts, staffs, rewards, and measures success. It isn’t a new offer on the old model. It’s a different business. How each agency gets there depends on where it starts, and that’s where the real work is.

---

## The First No and the Second

Her question wasn't a complaint to manage. It was a test of the business model.

Many agencies would fail it without anyone doing anything wrong.

She wasn’t asking us to say yes to everything. She was asking us to keep exercising the judgment that earned her trust after the contract was signed.

She bought the first "no". Delivery gave her the second. Everything between those two is how the agency is built: what it gets paid for, who stays in the room, who can change the plan, what gets rewarded, what a plan is allowed to become once real users touch it...

AI will make both kinds of “no” cheaper to produce and easier to defend. Which one a founder hears after signing is decided long before anyone writes a prompt.

See AI Clearly. Think Beyond the Hype.

Subscribe 

Email sent! Check your inbox to complete your signup. 

No spam. Unsubscribe anytime.