Code Is Cheap. Context Isn’t: Tau9 Labs’ Factory-Floor Playbook

Alex Gruebele & Chandler Gonzalez on how AI is unlocking a new era of custom software in manufacturing, cutting quoting from hours to seconds with builds measured in months, not years.

Author

Stuart Willson
December 05, 2025

This is the newsletter of Just Curious, the world’s leading network of applied AI experts. We connect operators and investors with proven partners to design, build, and deploy AI solutions that drive measurable growth, efficiency, and innovation.

You can submit your needs here, we’ll anonymize them, and return 5–15 expert perspectives with pricing from our network. Think of it like a “Mini RFP.” And it’s free.

Alex Gruebele & Chandler Gonzalez, Co-Founders, Tau9 Labs

Manufacturing rejects one-size-fits-all software. Plants vary, data is unruly, and the knowledge that makes things work lives in binders, emails, and the heads of operators. That’s why Alex Gruebele and Chandler Gonzalez at Tau9 Labs built their model around proximity: go onsite, watch the work, ask naïve questions, and build only what’s needed.

Or in their words: “We show up… map the workflow… and then build the smallest program that kills the problem.”

What came through when we spoke is a philosophy that sounds simple but is unusually rare in industrial AI: code is cheap now; knowing what to build is the hard part. When you understand the true bottleneck on day one, you can deploy something that stops the bleeding today, not after a year of remote scoping and ticket ping-pong.

Their recent quoting transformation for an injection molder illustrates the point: a platform built in about four months, designed so a buyer can upload a part file and check out in seven clicks. Because in quoting, the difference between a sale and a lost deal is often single-digit seconds.

As Alex put it, “If you respond first, your win rate jumps—often by 50%.”

Speed isn’t a feature. It’s strategy.

And stepping back, this conversation made me more bullish than ever on custom software (esp. inside industrial companies). When the constraint is context, not code, bespoke builds don’t just work better, they cost less, ship faster, and hit the P&L sooner. You’re not paying for features you don’t need; you’re paying to remove the precise bottleneck that’s costing you money.

What to expect:

  • Why “drop-ship” software development fails on the factory floor

  • The quoting build: 4 months to launch, 7 clicks from upload to pay

  • How cycle time fell from hours to seconds, and why win rates doubled

  • Where AI already pays in industrials: unstructured docs and semi-automations

  • Why 95% accuracy across 5 steps doesn’t survive real workflows

  • How on-site engineering shrinks timeline, cost, and failure rate

As you may have seen, we’re introducing our first-ever Just Curious member offer: a focused, two-week Data & AI Opportunity Strategy sprint from a leading Data & AI Consultancy. We’re making it available at 50% off. More on that toward the end of today’s issue.

We’re now on Spotify, Apple, etc.

We finally launched podcast featuring in-depth conversations with the operators building real, production-grade AI systems. The Tau9 Labs episode is now live on Spotify, Apple, and all major platforms.

Coming Up on Just Curious

  • Coming soon: The AI Lab, Watch Me AI, Alongside AI, Tenex, AI Momentum Partners

Watch the full interview with Tau9 Labs

Expert Q&A: Alex Gruebele & Chandler Gonzalez, Co-Founders, Tau9 Labs

Each week, we ask our applied AI experts five questions to surface their frameworks, lessons, and operator insights. Tau9’s answers follow, paired with Stu’s thoughts from our full conversation.

Tau9’s edge is simple: they go onsite, learn how the work actually happens, and build the smallest, sharpest piece of software that eliminates the bottleneck for good.

1. Describe who you are, what you do, who you do it for, and what makes your approach unique.

“Tau9 Labs build bespoke software and AI systems for industrial companies.

We’re experts in automation, AI, and manufacturing. But our real advantage is how we work. Most software shops work like drop-shipping: you send specs, they send code. That fails in manufacturing. The real world is messier than that: heterogeneous processes, unstructured data, tribal knowledge, and workflows that don’t fit off-the-shelf tools.

So we do the unfashionable thing – we show up. We sit with operators, maintenance, engineering, and leadership. We map the workflow, make sure the root causes are crystal clear, and then build the smallest program that kills the problem.

Doing it this way beats email-driven freelancing every time. Face-to-face, you spot the hard parts fast and decide what matters. You don’t get endless ticket ping-pong or weekly zooms about scope creep. You get a fix that works, measured in scrap avoided or uptime gained.

And we can do it fast. A solution that takes months to scope, a year to build and debug wastes millions in avoidable losses. When you understand the problem on day one, you can build and deploy something that stops the bleeding now. We’re scrappy problem solvers. We fly in, get hands-on, and build solutions that stick.”

Stu’s Thoughts:

In manufacturing, the bottleneck is almost never “the model.” It’s context. Processes diverge by machine, by shift, even by operator. And the tribal knowledge that determines whether a system works is almost never systematized. As Alex put it during our conversation, “The real process isn’t in a spec. It’s in the way people actually work on the floor.”

That’s why their approach—fly onsite, map the workflow, ask naïve questions—isn’t artisanal. It’s economically rational. A single walk-through can collapse weeks of remote assumptions. Alex told me about a client escalating a mysterious network issue across multiple teams; Tau9 solved it in four seconds by tracing a cable, something three remote calls had missed entirely. That’s the cost of abstraction.

A useful way to understand their method is through a simple lens: Context → Constraint → Kill Switch.

  1. Context: See the work as it happens, not as it’s described.

  2. Constraint: Identify the single step where time, scrap, or confusion accumulates.

  3. Kill Switch: Build the smallest tool that eliminates that step completely.

And notice what they measure: scrap avoided, uptime gained, response time reduced. Not “AI maturity,” not “feature completeness.” When the north star is operational impact, you naturally ship tools that are small, blunt, and immediately useful. In a world where code is cheap, the advantage shifts to teams who know exactly what to build, and what not to.

2. What problem are you most focused on solving right now, and how are you anticipating solving this problem with AI?

“Our projects in the past few months have centered around two areas. (1) workflow automation, and (2) finding savings from big data sets that have more than numbers in them.

The workflow automation stuff is getting better every day, but still limited to workflows that are a few steps (try compounding 95% accurate more than a few times haha!). But simple stuff that currently takes people 30–60 mins for one big step is often automatable. Aspects of programming equipment, data entry of course, responding to quotes, etc. And that can really add up if you have a team of 5–10 people doing that task.

The big data stuff isn’t really realtime. So what I’m loving here is – you give the customer the result rather than a tool to run daily. That result can be incredibly valuable: “we found a consistent failure in 6 of your plants running this SKU under these conditions. The component is failing in the field 18 months early costing you $3M a year in excess warranty coverages vs the SKU produced in other plants.”

What’s particularly interesting in doing this right now is the AI models aren’t good enough to be an app the customer can just dump a 100k PDFs into. The PDFs aren’t clean enough, it’s too much data to run in one go, etc. But – software guys (with a bit of industry-specific knowledge on their side) can understand the goal and write code as they go. By writing it to work on the specific problem, they can work around areas where AI is weak and would fail, while still leveraging AI for the bits it excels at (making projects economically feasible that weren’t before).”

Stu’s Thoughts:

What Tau9 is doing here is deceptively disciplined. Instead of chasing abstract “AI opportunities,” they look for high-frequency, high-friction steps, the ones that take 30–60 minutes and appear dozens of times a week. That’s where automation compounds. It’s also where teams routinely underestimate error rates. 

Alex joked about it, but he’s right: “Try compounding 95% accurate more than a few times.” Five linked steps at 95% each don’t get you something “mostly right.” They get you something operators stop trusting altogether.

Their second lane—unstructured data—is where the leverage gets even more interesting. Most manufacturers sit on decades of quality reports, technician notes, and warranty claims that no human could ever fully absorb. Tau9 doesn’t turn that into a tool. They turn it into one decisive sentence:

“Your component fails 18 months early in these six plants—$3M in annual warranty exposure.”

A finding like that immediately reshapes how a plant allocates time and budget.

And the way they get there is by treating AI as one ingredient, not the main dish. They wrap deterministic logic around model weaknesses and use AI only where it changes the economics. It’s a pragmatic hybrid pattern: AI for what it’s good at, software for what AI can’t yet do. For industrial companies, this is the real path to value; systems that work today, not someday when the models are perfect.

3. Where have you seen AI drive meaningful ROI the fastest — in weeks, not years?

“Where there are huge amounts of unstructured data that have yet to be used. One example is thousands of previous quotes that can be automatically read and used to accurately instant-quote future RFQs. Getting the bid out first increases the chance of closing the sale by 50%. Another example is mining a decade of quality and technician maintenance data. You can extract macro trends from details no human could discern – they can’t read and store that much text in their heads.”

Stu’s Thoughts:

Two patterns stand out here. First: quoting. The “art” of quoting dissolves once you map the logic. What matters is cycle time. 

Tau9’s instant-quote system cut process time from hours to seconds. And because buyers often award the deal to the first credible responder, the platform effectively doubled win rate.

Second: unstructured maintenance and quality data. Most manufacturers have a decade of problems hiding in plain sight: text no human can ever fully consume. AI can. When Tau9 turns that into a single actionable insight (what’s failing, where, under what conditions), the ROI is immediate.

If a workflow affects revenue or warranty exposure—not just back-office efficiency—it belongs at the front of the automation queue.

4. What’s the most common mistake companies make when they try to “do AI”?

“Hah – yeah I’d say trying to “do AI” is a bad start. Trying to push a solution for the sake of innovation, often at problems that are easy to solve but maybe not that valuable. Or look valuable and easy to solve, but fail often at runtime.

It’s more about having some specific problems in mind. Then you call in problem solvers like us. And we can tell you how solvable it is (with AI or more traditional software). And often times it isn’t solvable with AI today (or at least not at a good ROI). But it might be in 4 months.

Capabilities and AI models are changing fast. We’ve gone back to prospective customers 6 months later and told them “ok now we can do this for ya.”

Though wanting to “do AI” isn’t all bad. Even if you don’t know exactly what you want to AI, you can call in experts and they can talk to you and your staff. Look at where you’re losing money or not hitting growth goals. And collaboratively figure out if AI can solve those. We’re especially fond of this approach because we build custom software. So it’s often some mix of AI and deterministic code that works best. Not trying to push “my AI agent can automate your whole X team overnight!” which rarely works..”

Stu’s Thoughts:

Alex’s warning about companies trying to “do AI” should feel uncomfortably familiar to most operators. Too many teams start with an imagined solution instead of an actual problem. They pick a workflow that seems automatable, or flashy, or strategic, and only later realize it’s brittle, low-value, or impossible to maintain.

Tau9 approaches the problem backward in the best way: start with the pain, not the model. 

In our conversation, Alex told me about a client pushing hard for an AI-driven drawing-analysis tool. It looked innovative, but once they dug in onsite, the real goal wasn’t image understanding at all, it was freeing engineers to do higher-value diagnostic work. Reframing the problem unlocked a simpler solution that shipped five times faster and returned value immediately.

This is the throughline in their philosophy: use AI when it delivers ROI, not because it’s AI. And be willing to wait. As Alex said, “We’ve gone back to customers six months later and told them, ‘Okay, now we can do this for you.’” They treat AI capacity as a moving frontier, not a fixed toolset.

For leaders, the takeaway is simple and practical: Don’t ask ‘Where can we apply AI?’ 

Ask ‘Where are we losing money or time?’ 

Start there, and the right technology becomes obvious.

5. What’s one AI initiative you’ve seen consistently move EBITDA within 1–2 quarters?

“We’ve pretty much done one-offs for people that all do quite different things, so it’s tough to say anything about “consistency” from personal experience. What I can say is it’s much faster to write code now, so moving EBITDA in 1–2 quarters is extremely doable. So far most of our projects have taken ~3–4 months to launch. That’s the kind of stuff that can automate say a third of the workflow of a team of 10 though, so it definitely moves the needle.”

Stu’s Thoughts:

You can tell a lot about a team by how they talk about time. Tau9 doesn’t promise “transformations”; they promise 3–4 month cycles that eliminate a meaningful slice of labor or cycle time. And in manufacturing, that’s enough to move EBITDA inside a single quarter.

Their quoting build is a great example. In the first month alone, the system generated roughly $40K of incremental reenue, before any marketing push, before any broad rollout. By week six, the client expanded scope because the system wasn’t a demo or a prototype; it was already changing operator behavior.

Alex’s point that “code is faster to write now” understates what’s really happening. The cost of iteration has collapsed. With AI-accelerated development, small teams can deliver value at a cadence that matches PE timelines: weeks for proof points, quarters for EBITDA impact.

Tau9 scopes projects based on when value must appear, not when the tech will be perfect. If a workflow can’t return impact in 1–2 quarters, they don’t start there. That discipline is why their systems don’t stall in POC purgatory. Yhey’re engineered from day one to become part of the business.

Key takeaways

  1. Prioritize time-to-quote. If revenue depends on speed, target a sub-60-second RFQ cycle and measure win-rate delta when you're first.

  2. Mine your “paper” plant. Use AI to read 5–10 years of maintenance, quality, and warranty data; deliver findings that change decisions, not dashboards.

  3. Shorten automation chains. Limit AI sequences to 3–4 steps or add human gates. Never multiply 95% accuracy across 8 steps and call it production.

Member Offer: Data & AI Opportunity Strategy (50% Off)

Are you a private equity or private credit leader trying to determine where data and AI can actually drive meaningful leverage? In sourcing, underwriting, portfolio oversight, or fund operations?

We’re launching the first offer from the Just Curious Network — a two-week engagement delivered by a leading Data & AI consulting firm that works exclusively with private equity and private credit. The goal: help your team identify the highest-value, lowest-friction ways to apply data and AI, and develop a plan you can execute in months, not quarters.

What you’ll gain:

  • A prioritized roadmap of 3–5 high-impact AI + data use cases tied directly to your workflows

  • A buy-vs-build analysis showing what should be built internally vs. sourced from vendors

  • A phased 3–6 month implementation plan with clear sequencing, dependencies, and org enablers

  • A final presentation + written report for partners, CIOs, and portco leadership

Valued at $15,000 — offered for $7,500.

If you'd like details, just reply “Strategy.”

Connecting with Tau9