Aqmen AI

What a week! Last week was one of those intense but incredibly rewarding ones for us at Aqmen AI. Together with Felix Beccar Varela, we spent several days in London meeting with current and future consulting partners, as well as others helping us shape our product, from PE investors to tech gurus, gathering invaluable feedback to keep refining our solution. By Friday, we were running on fumes… but feeling deeply fulfilled. As we approach Thanksgiving (a tradition not native to us Argentinians but one I’ve come to appreciate), I can’t help but feel immensely grateful. For my co-founders, our team, and all those who believe in what we’re building: The backbone for strategic decisions that drive resource allocation 📸 Below, Felix teaching a room full of sharp consultants how to use Aqmen’s platform. Shaping, together, the future of Commercial Due Diligence. If you’re involved in CDDs and haven’t seen what we’re building yet, reach out. And if you’re in London over the next two weeks, let’s meet, I’d love to show you Aqmen in action.

We’ve just refreshed our website at Aqmen AI – clearer, faster, and more focused on what really matters: Lio, our AI-powered solution built for consultants by consultants. The new site shows how Lio helps you: 🌳 Build market structures and driver trees in minutes 📊 Generate zero-defect Excel models programmatically 👨💻 Adjust assumptions and make changes to models instantly as new data comes in (aka someone decided that 'Small Companies' are now 1-10 instead of 1-20 employees ... 🙈). We’ve also added new product videos so you can see Lio in action, and launched a blog where we’ll share frameworks, market-sizing guides, and case studies from real projects. Take a look and let us know what you think! Miguel Casares Santiago Segarra #AqmenAI #Lio #Consulting #PrivateEquity #CommercialDueDiligence https://www.aqmen.ai/

🚀 The future of AI for work isn’t general — it’s purpose-built and precise. After my last post, several people asked me to compare Lio by Aqmen AI with Excel’s native AI assistants. So I ran the exact same prompt again, this time adding Shortcut AI to the previous comparison with ChatGPT and ChatGPT Agent Mode from OpenAI. Then I asked ChatGPT itself to score the Excel models produced by each tool. 📊 Result: unchanged. Lio by Aqmen AI drastically outperforms everything else. See image below. (Last time Lio scored 9/10; this time 10/10 on the same Excel — showing that LLMs are stochastic. This randomness is exactly why you don’t want an AI agent writing your spreadsheet cell by cell.) 💡 Why does Lio win? Because Lio never writes the Excel at all. 🧩 The logic is captured first through clear entities like dimensions, segments, drivers, and expressions. ✅ Once the user approves that logic, the Excel is generated programmatically, not guessed by an LLM. That means: ❌ No hallucinations ❌ No ambiguity ❌ No formula errors ➡️ Just deterministic precision. 🧠 How to do it right (for any structured workflow): 1️⃣ Codify the workflow → break it into clear, reusable entities (objects, variables, relationships, rules). Domain expertise is key here! 2️⃣ Human-in-the-loop → experts validate the logic before execution. 3️⃣ Programmatic generation → once approved, outputs (Excel, reports, dashboards) are produced deterministically from those definitions. This recipe works anywhere precision matters (finance, engineering, biology, operations), not just market sizing. 🍝 Example: Dry Pasta in Brazil AI might suggest a relationship like ➡️ Q = Total population × Penetration × Annual consumption Once the user validates this logic, the mapping of that formula into the Excel model becomes fully programmatic. Every instance is generated automatically, referencing the correct cells across hundreds or thousands of rows. You don’t want an AI agent rewriting that equation cell-by-cell. That’s inefficient, error-prone, and FUNDAMENTALLY the wrong use of intelligence. 🎯 Takeaway In structured, high-stakes workflows, vertical AI beats general AI every time. If you care about accuracy, auditability, and speed, specialization wins. Miguel Casares Felix Beccar Varela

🎉 Lio’s birthday 🎉 (and a look back at where it all started) Even though Aqmen AI may still be under a year old, Lio’s story began before that, when Santiago, Felix and Miguel sat down to design what our solution should look like… and, in doing so, gave birth to Lio. Less than a year later, that vision has grown into a platform helping consulting teams around the world work faster, smarter, and enjoy their work more. Lio was designed to make consultants’ lives easier. Automating the grunt work, supercharging deliverables, and letting people focus on what really matters: thinking, insight, and impact (the very reason they joined consulting to begin with). He’s a dog because he’s a consultant’s best friend… though we’re starting to think there’s a little G.O.A.T. in him too. 😉 Happy Birthday Lio, we love you! 💙

Really enjoyed reading the "State of AI in Business 2025" report (link below) and wanted to share some thoughts on what we’re seeing with our pilots at Aqmen AI on what works today: 1️⃣ Focus on a problem & map it to a workflow – the report calls this “deep understanding of our workflow”. When you know the pain (and find a new way to solve it), everything becomes easier. 2️⃣ Design partners FTW – this isn’t an out-of-the-box shopping experience. Customers have different problems, and solutions work best when vendors and customers partner to prioritise and solve them together. 3️⃣ The bar is high – pilot users are used to AI that “just works” in their personal lives. Feels like a Blackberry vs iPhone moment (👴). The difference at work is *context* ('context is king'? 🤔) and how well AI solutions are able to ingest it. Vendors that deeply learn enterprise context will be in a much better position to win. 4️⃣ Play the long game – “success vs failure” in pilots is too binary. This is a journey, and work will look radically different in 5–10 years (remember BlueJeans for videoconferences? 🤣). Pilots will “fail,” and that’s ok – you iterate, jump on the next problem, and keep going. The gains show up over time, not overnight. Working on these pilots with consultancies, I see first-hand how our world of work is changing – especially on problems I used to face in my first 7 years doing due diligence. I honestly wish I had back then what I get to use today. What excites me most is how it lets us rise above the grunt work that used to eat up countless hours. This doesn’t mean jobs disappear tomorrow or that we’ll suddenly all work three-day weeks. But it does mean *now* is the moment to lean in, become an “AI-native” professional, and adapt your role into something new – maybe even into jobs that don’t exist yet. Those who embrace it early will be first in line when those roles emerge, internally or externally. cc Miguel Casares Santiago Segarra

Last Friday we published “A Guide to Market Sizing” on our website: 🔗 https://lnkd.in/dvkRJHA5 It’s our take on best practices for market sizing. A complete guide showing how top consultants approach the exercise, with steps, definitions, worked examples, and common pitfalls. In our conversations with consulting professionals, one theme comes up again and again: - 9 out of 10 due diligences and, - 4 out of 5 strategy projects include a market sizing. And yet, those leading the workstream often say the same thing: 👉 “Market sizing is an art.” I used to agree. But today, I believe it should be treated as much more of a science (without losing the craft and personalisation that make each case unique). Why standardise market sizing? 📚 Easier to teach and learn 🔍 Auditable and transparent 📊 Flexible to run scenarios 🤝 Seamless to pick up a colleague’s work (no more consulting nightmares) 🗂 Structured to codify knowledge That’s what this guide is about: making market sizing clear, structured, and repeatable. Aqmen AI’s platform, with Lio's guidance, takes market sizing to another level. But that’s a post for another day 🙂(Or reach out if you can't wait to see Lio in action.)

I'm finishing the week on a high note. We had a product release this week, and I woke up yesterday morning to the email below. "I spent 10 hrs on Aqmen last night. Freaking blew me away. Created 3 versions of a global market model for [*redacted*]. It will take a little getting used to with how much thinking I can outsource. But man did it produce something amazing. The most useful thing I found was the drivers sensitivity analysis. Never tweaked a model to this much confidence before, despite having no clue"

🤖 With GPT-5, Gemini 2.5 Pro and other foundation models… do we still need vertical AI? 👉 My view: absolutely yes. And market sizing is a perfect example. On the surface, it looks simple: 💡 “What’s the size of the pasta market in Brazil?” Ask a model and get a number back. But in practice, market sizing is not just number crunching. It’s a workflow that blends: 🔹 Structuring the problem into a value driver tree 🔹 Pulling data from messy sources (census tables, filings, associations) 🔹 Reconciling inconsistencies and outdated stats 🔹 Building a transparent, auditable model (often in Excel) 🔹 Adding narrative and confidence behind the estimates ✨ LLMs are great for some of this (data gathering, drafting narrative). ⚠️ But they fall short on others like keeping consistency across 100+ branches of a value driver tree, enforcing structure, or generating audit-ready outputs. That’s where Lio by Aqmen AI comes in: ✅ Scaffolding the workflow ✅ Error-proofing calculations ✅ Integrating trusted data sources ✅ Delivering repeatable models that consultants and strategists can actually use This is the same reason vertical AI thrives in law, finance, and biotech: foundation models are powerful, but too unstructured to replace domain workflows. 🚀 I don’t see Lio as “an AI agent that answers market sizing questions” but rather as “an AI agent that builds structured, repeatable, and auditable market models in minutes instead of days.” Visit https://aqmen.ai/ to learn more. Miguel Casares Felix Beccar Varela