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 Pluris AI Use Case Library — over 140 published cases now, with about 190 more in the queue (329 in the dataset and growing) — 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 an Expert Review: we'll anonymize it, match it to experts with relevant experience, and return 3–4 perspectives with pricing.
Two cases this week that solve the same problem: the analyst queue. In both, the people who could answer the question were the bottleneck, and the fix was to make the answer self-serve without making it wrong. Jay Singh of Casper Studios did it for a hedge fund's research desk; Aanikh Kler of Lazer Technologies did it for a Fortune 500 CPG's business teams.
We also added a run of new cases this past week, from Lazer, CorralData, DevDash Labs, Quantif, and others. If you want to stand out in a noisy market, the best way to do it is through the proof of your work.
Key takeaways from this week's cases:
Ship in small modules. Casper split the hedge fund's workflow into three tools that each shipped and earned their keep on their own, taking analyst feedback after every release, so no single system had to be right on day one.
Build to how the client already works. The fund runs on "singles and doubles over moonshots." Casper matched the build to that philosophy and liked it enough to adopt the phrase across its own work.
Trust comes from governed context. A model pointed at the warehouse can sound right. Lazer's versioning and approval layer, controlling exactly which definitions the agent can use, is what got it to 100% correct answers in the P&L pilot.
Ship the governance layer alongside the tool. Lazer built an AI Hub for approving tools, managing access, and auditing usage, so the enterprise governs every AI deployment that follows from one place.
The queue was the bottleneck. Both companies already had the data. What they lacked was a way for anyone to ask it a question and get an answer they could trust, so the work stopped funneling through a handful of scarce analysts.
One thing I'm seeing more of: readers clicking straight through to talk to the expert behind a case. That's the point. If one of the builds below fits what you're working on, hit the link and I'll make the intro.
How One Hedge Fund Cut Pre-Call Prep From 60 to 5 Min
Jay Singh, Founder & CEO at Casper Studios
Expert | Jay Singh, Founder & CEO at Casper Studios |
Industry | Financial Services |
Build type | Knowledge Management & Search (RAG) + Data Synthesis & Reporting + Natural Language Processing |
Result | 80% Reduction in Pre-Call Research Time · 50% Reduction in Post-Call Analysis Time · Thousands of Analyst Hours Reclaimed Annually |
Implementation Time | 2–4 months |
Cost | Not disclosed |
Full case |
A $1 billion hedge fund covering about 200 public names had 15 analysts, and before every corporate access call each one spent 30 to 60 minutes on a single equity: pulling the 10-Q, the earnings transcript, the internal model, and old meeting notes, then turning all of it into a list of questions. Multiply that by hundreds of names and the prep alone was most of the job.
Casper Studios mapped the workflow before touching a model, then shipped three small modules separately rather than one system. The first writes the pre-call questions from the filings, transcripts, and the models living in Excel, SharePoint, and OneNote. The second reads the call transcript afterward for sentiment and signal. The third flags whether anything an executive said should move a buy, sell, or hold assumption in the fund's model.
Pre-call prep dropped from 30–60 minutes to about 5. Post-call review roughly halved. Across 15 analysts and hundreds of equities, that adds up to thousands of analyst hours back every year, and analysts showing up to calls better prepared.
What I like here is the philosophy behind it. The fund lives by "singles and doubles over moonshots," and Casper built to match it: three modules that each earned their keep, shipped and refined with analyst feedback, instead of one tool that had to be perfect at launch. Casper liked the framing enough to adopt it across its own engagements.
Best fit for: Hedge funds, asset managers, and PE firms where analysts lose disproportionate time to manual pre-meeting research and post-meeting synthesis that can't scale with headcount, especially teams already using AI individually but not yet wired into shared, workflow-embedded systems.
→ Want to talk to Jay? Reply to this email and I'll make the intro. Or request one directly on the case study page.
A Fortune 500 CPG Team Hit 100% Accuracy in Its P&L Pilot
Aanikh Kler, Head of AI at Lazer Technologies
Expert | Aanikh Kler, Head of AI at Lazer Technologies |
Industry | Consumer Goods & CPG |
Build type | Conversational AI (Chatbot / Agent) + Data Synthesis & Reporting + Decision Support & Scoring |
Result | 100% Query Accuracy in the P&L Pilot · Tickets Eliminated From the BI Path |
Cost | Not disclosed |
Full case |
At a Fortune 500 consumer goods company, people in finance, marketing, and operations couldn't get an answer out of their own data without filing a ticket and waiting in the analyst queue. Several earlier attempts at a natural-language tool had already fallen short.
The demo version of this is easy. A model that talks to the data warehouse and sounds right is a weekend project. One that is right every time is not, and a confident wrong answer is worse than no tool at all, because it sends people back to the queue for good.
So Lazer built the control layer first: a system to version, approve, and evaluate the context the agent is allowed to use, so every answer comes from approved definitions and nothing else. Then they pointed it at governed Snowflake data, with Retool running the application and agent layer.
In the P&L pilot, users got correct answers on every query. The company's data lead said people were "getting all correct answers for everything," and named the versioning work as the reason it beat the earlier attempts. Phase 2 was approved on the strength of it.
Lazer also built the layer around the agent: an enterprise AI Hub for approving tools, managing access requests, and auditing usage, so leadership has one control plane for every AI deployment that comes next. The 100% got the attention. The governed context underneath it is the reason the number held, and it's why people kept using the tool instead of drifting back to the queue.
Best fit for: Large enterprises where business teams depend on analyst queues for data answers, and leadership needs centralized governance over a growing portfolio of AI tools.
→ Want to talk to Aanikh? Reply to this email and I'll make the intro. Or request one directly on the case study page.
Also worth your time this week
🎙️ Latest podcast: Jimmy Bijlani on starting with the P&L, not the tool. Jimmy told me about a PE-backed software company that was sure it was ahead on AI, until his team went under the hood and found developers copy-pasting code by hand and a support bot answering almost nothing. The fix was one boring internal workflow, built on software the company already owned, worth about $700K over three years. → Website | Spotify
💬 A couple of other posts worth a look:
Anthony Pompliano praised the young founders building AI consultancies on "you need AI, pay us a boatload, we'll do it for you." I don't see it that way. If a firm wants a big check up front, you're talking to the wrong people. Start with one contained project, prove it out, and scale from there. → Read it
A CFO-advisory co-founder told me this week that all ~1,000 of his finance clients are getting AI crammed at them from above. My thoughts on why the office of the CFO is where domain expertise outweighs the model, and why AI adoption is really a capital-allocation decision. → Read it
🗣️ From me: OpenAI published a certified-partner locator this week, with Pluris network participants Artium and Eliza among the firms on it. I wrote up why a spot on that list tells you a firm cleared a bar, and how to find one that's actually solved your specific problem. → Read it
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 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.
