Luca AI

Hassaan Mirza recently shared how to go through Exploratory Business Analysis with Luca ⬇️ This, along with other AI insights, are all located in our community: https://lnkd.in/g3hJgGyj Some of the most valuable questions you can ask about a business are also the broadest: How are we actually doing? Where should I be looking? Before drilling into any one metric, it pays to get the lay of the land before you commit to a depth-first investigation. That's exactly what Luca is built for. In this walkthrough I'll take a single high-level question — how has my revenue evolved over the last three years? — and let Luca carry it all the way from raw data to a shareable one-page report. Step 1: Start broad I don't want to begin with a hypothesis. I want a quick, honest overview of how revenue has moved over the past three years. Here's the prompt I give Luca: "Extract my revenue figures from the last 3 financial years, then perform a centered 9-month rolling average to smooth out quarterly trends so we get a clean overview of the data. Plot it on a timeseries graph and export the underlying data to a CSV." ⚠️See the output in picture #1 before reading further This is deliberately high level. It's a breadth-first search over my business health — a way to see the whole shape of things before deciding which areas are worth drilling into. The takeaway is immediate: the revenue trend is heading the wrong way. Step 2: Put numbers to it A chart tells me the direction; it doesn't give me the language I need for an end-of-quarter retrospective. For that I want precise, metric-driven figures. So I follow up with: "The trend is clearly negative over these three years. Extract the year-over-year metrics so I can put a number on it." ⚠️ See output in picture #2 before reading further These YoY figures put the situation into sharp perspective, which is exactly the kind of numerical point of view I can stand behind in a review. Step 3: Package it for an audience Now I want something I can actually share. A chart and a few numbers live in my head; a clean one-pager gets other people to engage with the findings and take them seriously. So I ask Luca to pull everything together and to go looking for supporting context of its own. ⚠️ See output in picture #3 before reading further Why this works What I like about this flow is that it mirrors how good analysis actually happens. You start wide, get a feel for the shape of the data, then progressively sharpen. This can take you from a smoothed trend line, to hard YoY numbers, to a document built for an audience. No SQL, no spreadsheet wrangling, no exporting between five different tools. Next time you're staring at a vague "how are we doing?", try handing it to Luca Eric Bidinger and letting it run breadth-first. You might be surprised how quickly the picture comes. Click here to check out Luca: https://ask-luca.com/

Check out a recent community post from Eric Bidinger ⬇️ The best AI users do not outsource thinking Using AI with deliberate intention may be what separates quality work from slop. As companies increasingly mandate employees use AI to boost productivity, emerging research suggests the technology can backfire when used poorly. In some cases, AI is creating extra work as employees teach themselves how to use it and spend additional time correcting inaccurate or incomplete outputs. At the same time, some workers are becoming overreliant on AI to complete tasks, raising concerns about deskilling and the erosion of critical thinking and human judgment. It's difficult to pinpoint the threshold that constitutes responsible AI use. But if companies insist employees incorporate AI into their work, they must be equally invested in helping them use it well. That means developing training that strengthens the human skills that make AI most effective. Check the comments to learn how we leverage AI 👇

Your AI model isn’t the problem, it’s actually your data. What sometimes gets lost in AI conversations is that the LLM itself is the same for everyone. A massive airline with an eight-figure AI budget and a thousand engineers has access to the same underlying models as a small e-commerce brand. There's no secret, more powerful version of GPT or Claude that only enterprises get. So if everyone has access to the same intelligence, why do some businesses get real value from AI while others get expensive noise? The difference isn't having a better model but the foundation underneath it. Business data is messy by default. This isn’t because anyone is careless, but because running a company means focusing on customers and product, not on keeping every system perfectly labeled and reconciled. Shopify says one thing, your accounting platform says another, your ad platform says a third. None of that is unusual, it's just what happens when you're busy running a business instead of maintaining a database. The problem is what happens when you point AI at that mess. It doesn't hesitate or flag the inconsistency. It gives you a confident, well-written answer anyway and you have no way of knowing if it's the right answer or a fluent guess dressed up as one. That's the part nobody's selling you about the real work associated with AI. It's the unglamorous process of cleaning the data, reconciling the systems, and giving the AI a foundation it can actually reason from. Skip that step, and a smarter model just means a more convincing wrong answer. The moat was never the intelligence. It's what you build underneath it. Follow for more practical insights on AI and E-commerce. Eric Bidinger

We're live in a few hours. 🔴 AMA Session is today, 11:30 AM EDT/ 4:30 BST Eric and the team will be answering real questions on AI, data, automation, and the stuff that actually moves the needle in e-commerce. If you've got a question sitting in your head surrounding fragmented data, marketing that's not converting, or an AI tool you're not sure is worth it… this is the spot to bring it! Click the link below to join ⬇️

Prompt of the week from Eric Bidinger Here's one he’s been running on Luca this week ⬇️ The prompt: "What the single biggest lever I can pull to increase my profit in the next 30 days." Why it works: Simple Effective Tailored to your business Identifies a quick win and shows you how to implement it Screenshot of the output: (scroll to see)

We're two days away from our AMA Session and we want your questions in the room! Our Topic: Why more data doesn't automatically mean more clarity. Before we lock the agenda in, we want to hear from you. What's the question you'd actually want answered around AI and data? The thing you've been stuck on, the decision you keep putting off because the numbers aren't telling you anything clear. Drop it in the comments below ⬇️ We're not building this session around what we assume people need, but what they actually need!

Alex Owen shared his thoughts on voice-to-text vs. typing out prompts ⬇️ Voice-to-text has been one of my biggest AI unlocks. Not just because it's faster (or because I’m a crazy). But because of what it does to the quality of what I actually get back. When you type a prompt, you automatically compress it. You strip it down to the shortest version that still makes sense. When you talk it out instead, you naturally include the stuff you'd have edited out. This includes details like the backstory, the constraint you're working around, the detail that felt like too much to type but actually matters. And context is the whole game with AI. What you put in is what you get out. A stripped-down prompt gets a stripped-down answer. A prompt with real context, specifically your actual numbers and situation, gets something you can actually use. So now when I'm working with AI tools, I just talk it out. Ramble a little. Give the messy version. It's faster to produce, and every time, it's a better input. Try it the next time you'd type a two-line prompt, talk it out instead and see how much more makes it in. See below to know where to click!

Live Q&A Session on AI trends + Data & Automation: hosted by Luca AI Co-Founder & CEO Eric Bidinger Most e-commerce leaders lack clarity to make decisions. That's the problem we're tackling in our next private AMA session. If you've ever stared at five different dashboards and still couldn't answer a simple question about your business, we can help (because we’ve been through that before!). We'll cover: → Why more data doesn't automatically mean better decisions → What "clarity" actually looks like in practice → Real examples from businesses that fixed this Who: Any founder or builder with an interest in AI & a burning desire to learn + grow Where: Anywhere you have connection When: 4:30 - 5:30 pm BST (in the morning for folks in the U.S.) Drop your questions in the comments. We'll answer the best ones live. Link below ⬇️

Buying an AI tool is not the same as implementing AI Most businesses have already done the first part. They've got a subscription, a login, maybe a few people on the team poking around with it. But adoption isn't the same as implementation. And the gap between the two is where most of the value gets lost. Here's what's changing: the businesses seeing real results aren't the ones with the most tools. They're the ones who understand that there is unglamorous work first to do, like cleaning data, giving clear context, a system that actually understands their business instead of guessing at it. Why it matters: an AI tool sitting on top of messy, disconnected data doesn't make the mess smarter. It just gives you a faster, more confident-sounding wrong answer. What to actually do about it: before adding another tool, ask what it would need to know about your business to give you a genuinely useful answer and whether it currently has access to that. Check out the link in the comments to join our community - made up of other founders & the Luca team, including Eric Bidinger. ⬇️

Prompt of the Week from Usman J. 👇 Here's one he's been running on Luca ⬇️ The Prompt: "Act as an e-commerce marketing strategist. Analyze my Meta Ads data and suggest 3 high-impact creative variations for Q3." Why it works: You're giving Luca 3 things it loves: 1. A role → "e-commerce marketing strategist" sets the lens 2. Real data → your actual Meta Ads numbers, not a hypothetical 3. A clear ask → exactly 3 variations, not "some ideas" When it comes to prompting with Luca, you get out what you put in. This one boxes Luca in just enough to get sharp, usable output. What to look for in output like this: → 3 genuinely different angles, not the same idea reworded → Each one tied back to something in your data (a top performer, a drop-off, an audience) → Specific enough to actually brief a designer, not "try new creative" fluff Screenshot below 👇 Want more tips like this? Join the E-commerce Leaders community 👉 link in comments. Eric Bidinger