The Position · Edition 6

AI Can Inform Your Next Price — It Shouldn't Set It

17 July 2026

Every roadmap and pricing tool on the market is racing toward the same pitch: "agentic." Not dashboards anymore — agents. Systems that don't just show you what's happening, but tell you what to do next, and increasingly, do it for you.

Some of that is genuinely useful. Most of it is a category error dressed up as progress.

Here's the position: AI that suggests a roadmap priority or a pricing move — backed by a real body of evidence, and able to show exactly where that evidence came from — is one of the most valuable things competitive intelligence can do. An AI that automates that same decision, or hides the reasoning behind a confidence score, isn't an upgrade. It's a liability wearing an upgrade's marketing.

The pitch that's quietly changing

Two years ago, CI and product tools competed on visibility: better dashboards, faster alerts, cleaner battlecards. The human still made every call.

The pitch now is different. "Autonomous agents." "Set-and-forget pricing." "AI that manages your roadmap." The language has shifted from informing the decision to making it — and the shift usually happens quietly, in a feature description, without anyone stopping to ask whether the system earned that authority.

It rarely has.

Suggestion and automation are not the same feature with different marketing

A suggestion is an input to a human decision. Automation removes the human from that decision. Confusing the two isn't a semantic nitpick — it's the entire risk profile of the product.

A roadmap suggestion that's wrong costs you a conversation: someone reviews it, sees the reasoning doesn't hold up, and discards it. A roadmap commitment that's wrong — auto-populated into a planning tool, communicated to a customer, quietly built on for two sprints — costs you engineering time, a broken promise, and a much harder conversation later.

Pricing is the same, with sharper edges. A pricing suggestion that's wrong gets caught in review. A pricing change that ships automatically to a live customer base is a support ticket, a churn risk, and — for anyone in a regulated or contract-bound segment — potentially a compliance problem, before a human ever sees it.

Roadmap and pricing decisions carry second-order consequences — customer trust, competitive response, internal morale — that no model is positioned to weigh alone. That's not a limitation of today's models specifically. It's a structural reason these particular decisions should stay behind a human, permanently.

The evidence bar most "AI suggestions" skip

Assume, for a moment, that a tool keeps a human in the loop. That's necessary, but it isn't sufficient. The next question is whether the suggestion is backed by enough evidence to be worth a human's attention at all.

A suggestion built on three sales anecdotes and a hunch is not meaningfully different from the roadmap-by-loudest-voice problem we've written about before — it's just wearing an AI badge now. The badge doesn't add rigour. It just adds false confidence.

A suggestion is only worth surfacing when the pattern behind it clears a real threshold: a feature gap cited across a meaningful share of competitive losses, not one prospect's passing comment. A pricing gap showing up consistently across live competitor moves, not a single scraped page. If the evidence doesn't clear that bar, the correct output isn't a suggestion dressed up as one — it's "not enough data yet." That answer is less exciting. It's also honest, and honesty is the entire value proposition of an intelligence layer.

Show your work, or it's not a suggestion — it's a black box

The final condition is the one most agentic tools skip entirely: the suggestion has to show where it came from.

"We recommend raising your mid-tier price by 8%" is not useful on its own. It's an assertion with a number attached. "We recommend raising your mid-tier price by 8%, based on three competitor moves in the last quarter that widened the gap beneath your current price point — here are the specific pricing pages and dates" is a suggestion a pricing lead can actually evaluate, challenge, and either act on or reject with a reason.

The difference is whether the person receiving the suggestion can see the evidence and interrogate the reasoning, or whether they're being asked to trust a score. A confidence percentage with no visible source is not transparency — it's a black box with a number stapled to the front. If a suggestion can't point to the specific deals, signals, or pricing moves behind it, it hasn't earned the right to be acted on.

How this shapes what we build

This is the standard we hold our own agentic suggestions to.

When [Pro-Position.AI](

That's slower than "autonomous." It's also the only version of this that a product or pricing lead should actually trust.

The framework to steal

Before you trust — or ship — any agentic suggestion feature, run three questions:

Is this suggesting or executing? → If it changes a customer-facing price or a roadmap commitment without a human clicking approve, it's automation wearing a suggestion's name.
Is there enough evidence behind it? → If nobody can say how many deals, signals, or data points support it, the honest output is "not enough data yet" — not a confident-sounding guess.
Can it show its work? → If the person receiving the suggestion can't see the specific evidence behind it, it's a black box handing out opinions, not an assistant helping you decide.

If a tool fails any of these three, the "agentic" label is doing marketing work, not product work.

Where this bites first: any team that lets a roadmap or pricing tool cross from suggesting into deciding — without an evidence gate or visible reasoning — will find out the hard way, usually months later, once a quietly auto-shipped price change or an unreviewed roadmap commitment has already reached a real customer. By then it isn't a product bug. It's a trust problem, both with the customer and with whoever signed off on the tool internally.

First published as an edition of The Position, Pro-Position.AI's newsletter on competitive intelligence, product strategy and go-to-market.

Keep exploring: