Public relations

How Claude Code replaced a $20,000-a-year platform

Sam Polstein builds competitor dashboards and answer engine analytics with no coding experience, replacing reporting that once cost $20,000 a year.

A capability that cost roughly $20,000 a year early in Sam Polstein's career is now something he builds himself. Polstein, who runs communications and content for deep tech startups at HAUS, uses Claude Code to build client dashboards, competitor tracking and answer engine optimization analytics with, in his words, virtually no coding experience.

He is not a developer, and he is explicit that what he makes is not software in any serious sense. That is the point.

What does he build with it?

Three kinds of internal tooling, all reporting rather than product:

Custom dashboards for clients to track their competitors
Competitor tracking as an ongoing feed rather than a commissioned study
Analytics for GEO and answer engine optimization

His own summary of what changed, given to Artificial Reality, the AI buyer-evidence podcast hosted by Galina Fendikevich:

"You're able to kind of come up with your own products ... helpful analytics in terms of geo and answer engine optimization, and building dashboards for clients to keep track of their competitors — this kind of reporting that used to be either very expensive and a huge part of agency overhead, or didn't exist."

Read the last clause carefully. Some of this was expensive. Some of it was not available at any price.

What did it replace?

A category, not a subscription. The comparison Polstein reaches for is the social listening platform agencies chased when he started:

"At the start of my career ... the big powerful tools of the day, the Crimson Hexagons of the world ... they were charging 20K a year for access. And it's just nuts to now have — you can build a custom solution with me having virtually no coding experience. With Claude Code it's nuts."

Crimson Hexagon is not in his stack today. It appears in the account purely as the cost benchmark that makes the current moment legible: $20,000 a year for access to capability he now approximates with a bespoke dashboard.

Where does this stop working?

At the edge of simple. Polstein draws the boundary himself: this is not enterprise software development, it is "more about making a pretty simple solution, like building a dashboard." He also notes where the practical ceiling sits — dashboard work is exactly where his Claude usage limits, which "definitely come into play pretty quick," start to bite.

So the honest version of this story is not that an agency replaced a software vendor. It is that an agency replaced a reporting line item, up to the point where the thing being built stops being simple.

What does this mean for the software you buy?

Artificial Reality's analysis, not Polstein's: this is a story about the collapse of the mid-tier software market. Tools that were valuable mainly because they were expensive to build are now cheap to build badly and good enough. What survives is either infrastructure genuinely hard to replicate, or a product doing something a competent user cannot assemble alone.

The test that separates them is visible in this same account. Media list building is the one job Polstein cannot build his way out of. He says so plainly, the incumbent platforms he uses — Muck Rack and Cision — are still "somewhat out of date," and prompting a general-purpose model is worse. He can generate a competitor dashboard on a Tuesday and still cannot answer whether a named reporter will cover a story.

What should you do this week?

Go through your recurring software spend and sort every line into one of two piles: reporting, and judgment. Reporting is anything that collects things you could describe precisely — mentions, rankings, competitor pages, citation counts. Judgment is anything that requires knowing something not written down anywhere.

The reporting pile is where a tool like Claude Code applies first, and where Polstein started. Build one of them yourself before your renewal date and see whether the output clears the bar the invoice was buying. If it does, you have learned the price of that line was never about the capability.

Keep paying for the judgment pile. Polstein still pays two PR platforms for journalist data he calls out of date, because the alternative is worse — which tells you exactly how much a product doing something you genuinely cannot do yourself is worth. Which of your invoices is really just a report?