A bit of news before the cases: a wave of new case studies landed in the AI Use Case Library this week. Jay Singh of Casper Studios added a batch — a fintech fraud build pushed to 99% precision, 400k+ voice-AI calls fielded for Netflix and Doritos, hedge-fund and PE-services workflows, and more — and Alex Lirtsman of CorralData added a run of healthcare and ecommerce data cases alongside the one below. 

This is the newsletter of Pluris, a network of pre-vetted applied AI experts and AI-native service organizations that specialize in helping businesses design and implement AI solutions that drive value.

Two free ways to use us. Browse the AI Use Case Library for real AI implementations from our network: what got built, what it cost, and what changed once it shipped, organized around the problem you're solving. Or submit an AI need through a Structured Expert Review: we'll anonymize it, match it to experts with relevant experience, and return 3–4 perspectives with pricing.

Two cases this week, both about the same thing: you can't fund, fix, or automate what you can't see. One is a DTC brand that couldn't tell which marketing was actually paying; the other a PE roll-up that didn't know what its 500 employees did all day. 

Alex Lirtsman and CorralData built a margin-aware marketing mix model that proved paid search returns 4x paid social, and caught a Meta reporting error before it could skew the budget. Felix Rösner and Marble captured how work actually happened inside the roll-up and found 40% of it was duplicated, handing back 200–250 hours a month before building a single automation.

Key takeaways from this week's cases:

My thoughts: This one is a good example of what happens when AI meets a real data foundation. CorralData did the unglamorous work first, getting marketing, ecommerce, and finance data into one place, and once that base existed a margin-aware model proving paid search returns 4x paid social became something you could build on. The part I'd sit with is that the model caught a Meta reporting error before it became the budget, and most teams would have shifted spend the wrong way and never known why. Get the data foundation right, and a long list of builds like this one opens up.

The setup. A multi-region DTC/CPG brand was spending across Meta, Google, Amazon DSP, and Microsoft, reading performance off in-platform reporting and last-click attribution. Those numbers credit whatever channel is easiest to track, not the spend that drives margin. Across the US, EU, Canada, and Australia, the team had no margin-aware view of which channels were genuinely paying.

What CorralData built. CorralData piped the brand's marketing, ecommerce, and finance data into one platform where every revenue dollar reads against SKU-level COGS, fulfillment, card fees, and shipping, currency-normalized across all four regions. On that base it built a region-aware, margin-aware marketing mix model that credits ad spend across the days it keeps working and fits separately per region. A daily Pre-Marketing Contribution Margin view and the AskCorral agent keep marketing, finance, and operations on the same numbers in time for in-season decisions.

The return, and where it came from. The model quantified each channel's true incremental return and showed paid search delivers 4x the return of paid social, giving the team a defensible basis for moving budget instead of trusting platform dashboards. Because the contribution view sits beside the platform-claimed and last-click numbers, a Meta account issue that was inflating in-platform reporting and shifting credit toward Google Brand surfaced right away, before anyone could read it as a real performance shift.

Best fit for: Multi-region DTC and CPG brands spending across many channels that need margin-aware, incrementality-based measurement rather than last-click attribution, with marketing and finance working from the same contribution-margin numbers.

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

[Headshot: Felix Rösner — none on file in Airtable; add in Beehiiv]

My thoughts: As the tools commoditize, what stays hard to copy is a true picture of how an organization actually works, and Marble gets that by capturing real activity instead of relying on interviews and shadowing. That's how it found 40% of the work in this roll-up was redundant. The approach: map the real workflow and quantify the waste first, automate second. That sequence is why their clients see ROI instead of stalled pilots.

The setup. A PE-backed roll-up had closed a number of acquisitions but had no real view of what its 500 employees did day to day. Legacy systems across the portfolio companies didn't talk to each other, quarterly reporting lagged, and integration planning ran on org charts and assumptions. Underneath all of it, 40% of the work was duplicated across teams that didn't know each other existed.

What Marble built. Marble started with a 50-person pilot, capturing lightweight screen and keystroke activity with PII redacted on the device before anything left it, hosted inside the client's own AWS environment. Its process-modeling system turned that raw activity into maps of how work actually happened, not how documentation claimed it did, then generated living SOPs that spread best practices across teams without manual knowledge transfer.

The return, and where it came from. The pilot surfaced that 40% of the work was redundant and handed the finding straight to leadership for action. Post-acquisition integration ran 80% faster because planning was built on real process maps instead of interviews, and 200–250 hours a month came back to the business — all before a single automation was built. The whole engagement came in under $25K.

Best fit for: PE firms managing post-acquisition integration for roll-ups, especially those inheriting legacy systems that don't talk to each other and need operational visibility before automating anything.

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

🎙️ This week's podcast: my conversation with Doyl Burkett, Managing Partner of Integrity Growth Partners ($800M+ growth equity). Doyl's view: "AI isn't going to break private equity. It's going to sort it." Every software and tech-enabled business lands in one of three buckets — AI-native, defensible-moat, or undifferentiated — and the gap between them only widens from here. What makes Doyl worth listening to is that he spends as much time on what AI does to his own firm as to his portfolio: he runs Integrity like an AI-native company on a system he calls CAPE, which scores 50 million companies for fit and writes outreach tuned on eight years of A/B tests, letting three junior people do what used to take fifteen to twenty. He even pointed that sourcing engine at an acquisition, finding the AI-native company a portfolio business needed to buy. The same sorting, he argues, is coming for the GPs who never formed an AI view. → Listen to the episode | Subscribe on Spotify

📚 New in the library this week. Beyond the two cases above, Jay Singh of Casper Studios added a batch of AI builds worth a look: a fintech platform pushed to 99% fraud precision, 400k+ voice-AI calls fielded for Netflix and Doritos, a hedge fund's pre-call prep cut to minutes, and a PE IT-services portfolio that cut SOW time 50–60%. → Browse the library

💬 Why I do this. This week, one of the two founders featured above told me investors and buyers had reached out after hearing him on the podcast. One even said the interview turned up in their Claude research on him. That's the whole point. I started Pluris to help companies move faster on AI and help the experts in our network grow their businesses, so I don't gatekeep: every expert is visible and every case study is public, even if it means a client reads how our experts break down a problem and hires one directly, around me. The work is free either way, and the exposure is the point. I wrote up the fuller version of why. 

Working on something where one of these patterns might fit? Reply to this email — I'll either make the intro to the right expert directly, or open a Structured Expert Review and come back with 3–4 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.