The loudest story about AI is that its only job is to cut yours. Goldman Sachs put 300 million jobs in the automation blast radius, and when a company announces layoffs now, AI is the reason that comes to mind. There's something real in the fear. But the frame is doing damage right now: it anchors every AI effort on efficiency, which is one tool in a much bigger box.

The record we keep tells a different story. Across 110 deployments in the Pluris AI Use Case Library, 51% document revenue growth rather than cost reduction. The companies winning with AI aren't cutting heads. They're doing more with the teams they already have. That's what proof of work is for: a record specific enough that a stranger can check what got built, what it cost, and what changed.

Both cases below are the revenue version of that record. Ciridae helped a restoration firm recover hundreds to thousands more on every proposal while handing its CEO and COO their time back. Remix Partners gave a 70-year-old manufacturer the capacity to finally hit a 48-hour quote target it kept missing, without hiring.

Key takeaways from this week's cases:

My thoughts: When we sat down, Jack (Weissenberger) put it plainly: after two hours of comparing line items by hand, a person's mind goes numb, and AI doesn't. Ciridae ran its pipeline side by side against the leaders doing the work manually and came out ahead, which is proof of work in miniature. You can check the record instead of taking the claim on faith.

The setup. A home restoration firm's CEO and COO were each spending three hours on every proposal, comparing contractor estimates against insurance adjustments line by line to find discrepancies worth negotiating. That review ate 20 to 30% of their combined time. Deals that needed several rounds lost momentum, and sometimes lost the client. The bottleneck was paperwork sitting on the desks of the two people least able to afford it.

What Ciridae built. Ciridae spent time on-site first, encoding how the CEO decided what was worth challenging into structured rules the system could apply at scale. The pipeline ingests both documents at once, classifies every line item, applies those rules to flag where the contractor's scope and the insurer's adjustment diverge, and produces an annotated comparison for a human to review before negotiating. It runs on the Ciridae platform across Claude, OpenAI, and Gemini. Discovery and rule-encoding came first; the pipeline itself shipped in 48 hours, inside a months-long engagement.

Why the knowledge capture came first. The implementation is only as good as the judgment behind it, so Ciridae bottled the judgment before writing the automation. The system drafts the comparison; a human still runs the negotiation. In use, it caught discrepancies more accurately than the manual process about 10% of the time and recovered hundreds to thousands of dollars more per proposal. The bigger lever is the bottleneck itself: with slow negotiations no longer stalling deals and 30% of leadership time back, Ciridae projects the work could double the firm's enterprise value.

Best fit for: Home restoration firms, contractors, and insurance-adjacent service companies where reconciling proposals against adjustments is a recurring, manual executive burden.

→ Want to talk to Jack? Reply to this email and I'll make the intro. Or request one directly on the case study page.

My thoughts: Justin put this week's thesis better than I did. The owner's goal, in his words, is to divorce revenue growth from headcount growth: never net-hire another person in the office, and still double or triple revenue in a few years. That's what "do more with the team you have" looks like when an operator actually means it.

The setup. A California custom manufacturer with 40,000 SKUs and 70 years of history hit the ceiling of what its internal team could build. Quoting was manual end to end: engineers printed specs from a legacy system, re-keyed the data into a second system, advanced a bill of materials by hand, and chased quotes through back-and-forth with customers. The company's own 48-hour RFQ target was almost never met.

What Remix Partners built. Remix ran its GenAI Kickstart discovery, interviewing engineers, operations leads, and the executive team to map the quoting workflow and find where it actually broke. The first unlock was unglamorous: the legacy system produced proprietary output no other system could read, so Remix converted it to JSON. From there, over 3.5 weeks, they built a Claude Code workbench that ingests an incoming RFQ, compares it against all 40,000 prior parts, drafts a bill of materials, converts it into a price quote, and takes a first pass at the production drawings. Custom validation flags specification conflicts before they reach the floor.

The order of operations they got right. The JSON conversion came before anything AI touched the workflow. Most teams reach for the model first and find the data won't move; Remix solved the plumbing, then let the workbench run on top of it. Engineers now review a draft that is roughly 90% complete instead of starting from a blank page, which frees the senior people for the complex designs where the company competes. Projected result: 60–70% less engineering time per order.

Best fit for: Operators (CEOs, COOs, CFOs) at companies up to roughly 500 employees, across professional services, manufacturing, life sciences, healthcare, and VC/PE, who know AI matters and want a partner that ships working systems.

→ Want to talk to Justin? Reply to this email and I'll make the intro. Or request one directly on the case study page.

🎙️ On the podcast, both of this week's firms in their own words:

💬 This week's post: why the "AI just cuts jobs" frame costs you twice, and the 51% of the Pluris AI Use Case Library pointed at revenue instead. → Read it

Working on something where one of these patterns might fit? Reply to this email, and I'll either make the intro to the right expert directly or open an Expert Review and come back with 4-5 expert perspectives and pricing in a few days. Both are free, and I read every reply.

— Stu

Proof of Work: Applied AI Case Studies. The newsletter of Pluris.