NextLink Labs

Your company makes dozens of decisions a week. By Friday, it has forgotten most of them. Not the outcomes. The outcomes show up eventually — in the invoice, the shipped feature, the mess. What gets forgotten is the decision itself. What was decided, who decided it, what was promised, and why. You have watched this happen. A vendor gets chosen on a Tuesday call, and three weeks later the same debate reopens from scratch because nobody wrote down the reasoning. A client mentions something important, and it lives in one person's notebook — if it lives anywhere. A new hire asks why things work this way, and the honest answer is that the people who knew are gone and the reason left with them. Here is the strange part: companies are meticulous about their least important records and careless with their most important ones. Every email is stored. Every document has version history. Every ticket has an audit trail. But the place where real decisions actually happen — people talking to each other — is the one medium with no record at all. The truth of your company happens out loud. And out loud evaporates. For most of business history, there was no fixing this. That constraint is gone. Jordan Saunders wrote up what changed, what it looks like in practice, and the test you can run in your next leadership meeting that costs nothing. 🔗 Link in the comments. #AI #InstitutionalMemory #DigitalTransformation #Leadership #OperationalExcellence #AIStrategy #EngineeringLeadership

Most custom manufacturers have bought quoting software. Almost all were back in Excel within a year. The software is not the problem. Your best estimator is the pricing system, and it sits in their head. Here's what these tools get wrong, and what replaces the spreadsheet. So start with the uncomfortable question. How old is your head estimator? Whatever they know about how to price your work is the single largest undocumented asset in the business. Quoting software is not badly built. It is built for a different company. It assumes the product already exists, with a catalog, a set of options, and rules about what can go with what. In a custom shop the product does not exist until the quote invents it. Every request for quote is a small act of design. There is no catalog entry to configure, because the thing has never been made before. So watch what an estimator actually does with a new request. They go looking for the jobs that resemble this one. Not identical. Similar enough to be useful. Then they adjust. That job ran over on labor. This customer always changes the spec twice. We lost money on the last one in this alloy. This one carries a different margin because of who is asking. None of that fits in a configurator. It fits in a spreadsheet, which is exactly why the spreadsheet keeps winning. The tool has to hold judgment, and a catalog cannot. The reason this is worth revisiting now is narrow and specific. The hard part was never the arithmetic. It was retrieval. Finding the past jobs that rhyme with this one, when they were named differently and documented badly. That problem is now tractable. Pull the comparable jobs. Pull what they actually cost, not what they were quoted at. Assemble a draft. Hand it to the estimator with the reasoning attached. The estimator stays in the chair. They correct it, override it, and throw it out when it is wrong. Nothing goes out without the estimator's signature. Three honest conditions before anyone starts. Your history is scattered across PDFs and email and has to be organized. You have to start capturing actual costs against estimates. And early on it will pull the wrong comparable sometimes. At NextLink Labs, we built exactly this for manufacturing and field services teams, so treat that as our bias stated up front. The prize is not that the quote which took two weeks goes out in two days. It is that pricing judgment stops being one person's memory. What happens to your pricing the day your estimator retires? If you do not like the answer, that is the project. Subscribe to the newsletter: https://lnkd.in/efpcmnTk

Every company runs on two systems. The one they bought. And the one their people built around it in spreadsheets. The ERP holds the records. The CRM holds the contacts. And around both sits the real operating system of the business — the pricing workbook, the scheduling sheet, the export someone runs every Monday, the Access database in engineering that nobody in IT wants to talk about. That layer is where the actual process lives. Leadership usually treats it as a discipline problem. Why won't people just use the system? Wrong diagnosis. The workaround layer exists because it has to. Your ERP was built for your industry. Your company is not your industry. The big platforms cover the 80% every business in your space shares. Your people rebuilt the missing 20% in the only tool that would bend — and now your differentiation runs on spreadsheets, re-keyed by hand, reconciled monthly, known fully by one person. The answer isn't to customize the ERP core. And it isn't to force vanilla adoption either. It's a third option almost nobody talks about. Jordan Saunders wrote up what that looks like — and why your people already wrote the requirements. 🔗 Link in the comments. #DigitalTransformation #ERP #CustomSoftware #Operations #EngineeringLeadership #Manufacturing #OperationalExcellence

Most custom manufacturers have tried quoting software at some point. Most of them were back in Excel within a year. The software didn't fail because the team resisted change. It failed because CPQ software is built on one assumption that doesn't hold in a custom shop: that your product already exists. There's a catalog. There are options. Configure, price, quote. But in a custom shop, the product doesn't exist until the quote invents it. Every RFQ is a small act of design. There's no price book to look it up in — because the thing has never been priced before. So the estimator isn't configuring. They're reading a drawing, remembering the job from 2022 that was almost like this one but in stainless, recalling that it blew through its machining estimate because of one tolerance callout — and adjusting. That's not catalog work. That's judgment running on top of history. And that's exactly why Excel wins every time. The spreadsheet bends to whatever the estimator knows. The enterprise tool demanded the shop fit its model. The estimator won. Of course they did. But here's the real cost of that win: all of that pricing knowledge lives in one or two heads and a folder of files only they can navigate. Ask yourself how old your head estimator is. Jordan Saunders wrote up why CPQ fails custom shops — and what actually works when every quote is built from scratch. 🔗 Link in the comments. #Manufacturing #CustomManufacturing #CPQ #AIStrategy #DigitalTransformation #EngineeringLeadership #JobShop

Watch what your most experienced people actually do all day. Most of it is not judgment. It is retrieval. The estimator digging through old jobs to price a new one. The senior engineer answering the same architecture question for the fourth time this quarter. The account lead rebuilding a proposal from three past SOWs. The controller assembling month-end numbers from the same five systems every time. The support lead searching last year's tickets because she's the only one who remembers the fix. These are your best people. You pay them for their judgment. They spend most of their week finding things so they can spend a fraction of it judging. The retrieval is invisible — it doesn't show up on a timesheet. It just shows up as your senior people being busy, your turnaround times being long, and everything important waiting on one person's memory. That last part is the real exposure. When the knowledge lives in someone's head and a folder structure only they understand — you don't own it. They do. This is where AI should go first in your business. Not automation. Not agents. Retrieval. Jordan Saunders wrote up why — and what it actually looks like to build it right. 🔗 Link in the comments. #AI #AIStrategy #OperationalExcellence #EngineeringLeadership #DigitalTransformation #FutureOfWork #CTO

Your developers are running coding agents. Your ops, finance, and marketing teams are pointing agents at their own tools and data. Almost none of it is governed. The debate about whether to use agents is over. You are already using them. The only open question is whether you govern them — or just tell yourself you do. Here's what makes this hard: the thing that makes an agent useful is the exact thing that should worry you. Autonomy is the point. You want it reaching into systems and acting on its own — that's where the leverage is. But an agent authenticated as one of your developers, sitting on a laptop, holding production credentials, free to call whatever tool it likes — is quietly the least-governed compute in your entire company. Nobody set it up that way on purpose. It happened one useful tool at a time. Most AI governance doesn't fix this. It's theater — a policy PDF that agents don't read, or an API gateway that catches the easy stuff and misses everything happening at runtime. Jordan Saunders wrote up what actually works — and why most organizations are governing the wrong layer entirely. If agents are already inside your company, this is the most important thing you'll read about them this week. 🔗 Link in the comments. #AI #AIGovernance #AIStrategy #DevSecOps #EngineeringLeadership #CTO #Cybersecurity

There are two clocks running in AI right now. Clock 1: Cost per token keeps falling. Every lab undercuts the last announcement. The chart only goes one direction. Clock 2: What your company actually spends on inference every month. For most teams shipping real AI features — that number is climbing. Both are true. And most organizations aren't having the same conversation about both. Engineering reads Clock 1 and says the economics are improving. Finance reads Clock 2 and asks why the line item doubled. They're not wrong. They're just not talking about the same thing. The real issue is that cheaper tokens don't mean smaller bills. Cheaper tokens mean you do more with them — the same way cheaper cloud compute never once lowered anyone's AWS spend. Jordan Saunders wrote up why the old capacity planning models break completely for agentic AI — and what engineering and finance should actually be doing together about it. If your AI spend surprised someone in the last budget cycle, this one's worth a read. 🔗 Link in the comments.

Your feed is full of two kinds of AI content. Agents that supposedly run entire companies while the founder sleeps. And screenshots of chatbots telling customers to glue cheese to pizza. One is marketing for AI. The other is marketing against it. Neither tells you anything about what AI actually looks like when it ships. We run a consultancy that builds software for mid-market companies. We spend our weeks inside real codebases and real P&Ls. The AI that survives contact with production looks nothing like either side of your feed. It's narrow. It's boring. And it's quietly doing real work. → A model reads inbound support tickets and routes them with a confidence score. Anything below the threshold goes to a human. → A pipeline turns meeting transcripts into structured CRM updates that a person approves in one click. → A system drafts the first pass of a proposal from past project data. An account lead edits it. No autonomy theater. One or two things, wrapped in ordinary software, with a human at every point that matters. That gap between demo AI and production AI is not a small one. It's where most of the budget dies. Jordan Saunders wrote up what that gap actually looks like — and what it takes to cross it. 🔗 Link in the comments.

Your feed is full of two kinds of AI content. Agents that supposedly run entire companies while the founder sleeps. And screenshots of chatbots telling customers to glue cheese to pizza. One is marketing for AI. The other is marketing against it. Neither tells you anything about what AI actually looks like when it ships. We run a consultancy that builds software for mid-market companies. We spend our weeks inside real codebases and real P&Ls. The AI that survives contact with production looks nothing like either side of your feed. It's narrow. It's boring. And it's quietly doing real work. → A model reads inbound support tickets and routes them with a confidence score. Anything below the threshold goes to a human. → A pipeline turns meeting transcripts into structured CRM updates that a person approves in one click. → A system drafts the first pass of a proposal from past project data. An account lead edits it. No autonomy theater. One or two things, wrapped in ordinary software, with a human at every point that matters. That gap between demo AI and production AI is not a small one. It's where most of the budget dies. Jordan Saunders wrote up what that gap actually looks like — and what it takes to cross it. 🔗 Link in the comments.

Buying Databricks is the easy part. Getting to value fast is where the real work begins. NextLink Labs is officially a Databricks Bronze Partner — and that's exactly the problem we solve. We help teams build governed, production-grade lakehouse platforms in weeks, not quarters — with Unity Catalog, CI/CD, and DevSecOps discipline from day one. Planning a Databricks build this year? Let's talk. #Databricks #DataEngineering #Lakehouse #DataPlatform #DevSecOps #DataPlatform #PlatformEngineering