The Enabling Layer · Archive

Past issues, in full.

Every published issue of The Enabling Layer, including Episode 1 and Episode 2 of the Eli G. story, Volume 11: Meet Socrates, and the launch post, all in one place.

Volume 11 · November 15, 2026

Volume 11: Meet Socrates

Read this issue on its own page →

A learning system that shows you how to use an expert knowledge base

▶ Watch: Meet Socrates — The Learning System Built on an Expert Knowledge Base (0:56)

Meet Socrates — a learning system that shows you how to use an expert knowledge base.

Thousands of the best books, summarized for your learning: business, leadership, manufacturing, quality, and the classics of how work gets done.

How it works

Start from a department, then a topic: the Learning Map is built on one idea, that the purpose is learning.

Get Socrates

$4.99/month. Get Socrates here.

See Socrates at socrates.gotbots.si.

Also on LinkedIn: Volume 11: Meet Socrates · The Enabling Layer on LinkedIn

When we stop learning, we stop achieving!

Volume 10 · Episode 2 · November 14, 2026

Volume 10 · Episode 2: The Price Is Wrong

Read this issue on its own page →

What Eli's sales orders were hiding — and the self-improving expert knowledge base that found it

This is a real story with names and places fictionalized but real improvement and real methods.

Last episode, Eli G. — plant manager of the $40 million manufacturing company in Lima, Ohio that Chris owns — started the turnaround where turnarounds have to start: shipping on time, so customers believe the ship date again. This week we take up the second lever, the one that is usually worth 3 to 5 points of profit all by itself: price.

First, learn to read your own prices

Chris sat Eli down with a simple exercise. Take every product you sell. What does it actually cost you to make? What do you charge? Where is the margin — and where did it leak away?

Then do the same thing from the other side of the fence: check the competitors. What does the market charge for the same door? A pricing analysis is not complicated. It is just work nobody had ever done — because getting the numbers out of the system was a project in itself.

The system that costs a fortune and still answers in Excel

Eli's plant runs on Oracle NetSuite. He hates it — but it was what was available at the time, and ripping out an ERP in the middle of a turnaround is how turnarounds die.

What kills him is the monthly bill. The cost is high, every month, and after paying it he still has to export everything to Excel to get the answers he actually wants. The data is all in there. The knowledge is not coming out.

So Chris showed him a different move: don't rip it out. Layer on top of it.

Meet Socrates

Socrates is a self-improving expert knowledge base with Tireless Digital Helpers — an expert learning system built for exactly this. It is built on an Odoo foundational layer, and a library of distilled operating knowledge — thousands of the best books summarized for your learning, $10,000 of book value built in. Its Digital Helpers can scan, analyze, and make proposals on the records already sitting in a company's existing systems.

Chris trained Eli on it in plain language, the way you'd train a good analyst: here is what to ask, here is how to check the answer, here is where the evidence lives. Socrates proposes. Eli decides.

Meet Socrates yourself at socrates.gotbots.si.

For readers of The Enabling Layer: Socrates is $4.99/month. Subscribe at app.gotbots.si/shop/ml-001-socrates-190.
Odoo and Microsoft 365 subscriptions are not included. Data charges are not included.

The test: every sales order in the company

They pointed Socrates at the sales records — all of 2025, and 2026 year to date. Item-level detail: every product sold, every customer, pricing by SKU and by variant.

The answers came back in 30 minutes.

Not a quarter-end project. Not a week of Excel. Thirty minutes — and what it found explained a mystery that had bothered Eli for years.

The pricing model was upside down

Socrates' analysis showed that the company's pricing model was wrong compared to how the industry prices doors:

Eli and Chris sat with that for a minute. Then the summer made sense.

Every summer, school remodeling projects flood the plant with fire door orders. The plant chokes on the influx — overtime, expediting, chaos — and when the season ends, profit has gone down in the busiest months of the year. Volume up, absorption up, profit down. It never added up.

Now it did. They were selling their most demanding product below market, at scale, exactly when demand peaked. The plant wasn't inefficient in the summer. It was underpriced in the summer.

Build the model the market actually uses

Next, Socrates went out and gathered all the public information it could find on architectural door pricing. From the company's own records plus the market's own numbers, Eli and Chris built a new pricing model — their costs, the industry's structure, priced the way the market prices.

Then they used Socrates to build a new product configurator on top of that model. From that day forward, every new quote went out repriced — right door, right options, right price, no spreadsheet archaeology.

New pricing takes 3 to 6 months to show up in realized profit. Quotes become orders, orders become shipments, shipments become invoices. So with the lever pulled, Eli and Chris went back to work on the plant.

The lesson Eli keeps repeating

Eli is still amazed at the analysis — not because a computer did arithmetic, but because of what it didn't require.

They didn't rip out NetSuite. They layered Socrates on top of it. Socrates used its knowledge base and its Digital Helpers to mine the data already sitting in the company's systems and turned it into real Knowledge — knowledge Eli could act on the same week.

Most businesses are sitting on this. The records are there. The answers are in them. What's missing is the layer that reads them like an expert would.

Next episode: back on the plant floor — the constraint hiding in plain sight, and what happens when Eli follows the work instead of the reports.

▶ Watch Episode 2 — The Price Is Wrong

▶ Watch Episode 1 — How to Fix a Business · gotbots.si

▶ YouTube: youtube.com/@chilesandcompany

Reach us at sales@chilesand.co · (828) 521-0297

If this is helpful please pass it along to a fellow business owner.

"When we stop learning, we start dying"

Digital Helpers propose. Humans decide.

Biltmore Management Advisors has distilled 731 years of operating wisdom from our partners, and we used it to build the Max Knowledge Platform — so a business like yours can get better every day.

Built on Odoo · Microsoft 365 with Copilot
by Chiles & Company
gotbots.si

Business Systems with Tireless Digital Helpers and Self-Improving Knowledge Bases.

Also on LinkedIn: Volume 10 · Episode 2: The Price Is Wrong · The Enabling Layer on LinkedIn

Volume 9 · Episode 1 · November 7, 2026

Volume 9 · Episode 1: How to Fix a Business

Read this issue on its own page →

The power of a constantly improving and learning business

This is a real story with names and places fictionalized but real improvement and real methods.

▶ Watch Episode 1 — the story as a 4-minute video

Meet Eli G., owner of a $40 million manufacturing company in Lima, Ohio that is in trouble. Here is his P&L.

How do we get him better? What are we going to do? This is a week-by-week series: how to turn a business around, and the lessons we have learned over the years.

Friday, 4:40 p.m.

The last truck is still at the dock. It was supposed to leave at two.

The customer has called three times. The third call was not about this shipment. It was about next quarter's orders — and whether they should place them somewhere else.

Eli stands in the parking lot looking at a plant full of good people, good machines, and orders he cannot ship on time. He is working harder than he has ever worked. The business is getting worse.

The numbers, plainly

Sales$40.0M100%
Materials$20.4M51%
Labor$6.4M16%
Fixed overhead$2.8M7%
Freight out$2.8M7%
Gross profit$7.6M19%
Selling, general & administrative$9.6M24%
Operating result−$2.0M−5%

Read the freight line again. Seven cents of every sales dollar goes to freight — most of it expediting, paying premium rates to apologize for being late. A plant that ships on time does not carry that line. The freight bill is not a freight problem. It is the receipt for every broken promise upstream of it.

The first constraint is not cost

When a business loses money, everyone's first instinct is to cut. Cut people, cut spending, cut corners — and watch service get worse, orders leave, and the hole get deeper.

Eli's plant does not have a cost problem first. It has a trust problem. Customers no longer believe the ship date. So the first move is not a cut. It is a repair:

Fix customer service first. Ship on time. No quality escapes. Stabilize the plant until the ship date means something again.

Analyze every price and every cost at the same time. Pricing is the largest lever in most turnarounds — understanding how to price in your market is usually worth 3 to 5 points of profit by itself.

Learn in the open. Every fix gets written into the company's knowledge base, so the business keeps the lesson even when people change. Sense → Decide → Act → Measure → Learn, every day, on every line.

Only then do the later phases pay: cost extraction in the supply chain and on the floor, the order-flow fix worth 3 to 4 points in overhead, and freight swung from a 7-point drag to a contributor.

Where this story goes

Eighteen months. $40 million to $68 million, from losing money to solidly profitable — one constraint at a time, the way real turnarounds actually happen. No silver bullet. Just a business that learns faster than its problems grow.

Next episode: the order that almost killed them — and what Eli finds when he follows a single order from the customer's phone call to the dock.

Reach us at sales@chilesand.co · (828) 521-0297

If this is helpful please pass it along to a fellow business owner.

"When we stop learning, we start dying"

Digital Helpers propose. Humans decide.

Biltmore Management Advisors has distilled 731 years of operating wisdom from our partners, and we used it to build the Max Knowledge Platform — so a business like yours can get better every day.

Built on Odoo · Microsoft 365 with Copilot
by Chiles & Company
gotbots.si

Also on LinkedIn: Volume 9 · Episode 1: How to Fix a Business · The Enabling Layer on LinkedIn

Volume 8 · November 3, 2026

Volume 8: Alphabet Soup: How Your Business Systems Fit Together

Read this issue on its own page →

Somewhere this month, a software salesperson is sitting across from a plant owner, and the letters are multiplying. "Our ERP talks to the MES, the MES reads your PLCs through SCADA, and the OEE dashboard sits on top." The owner nods. He has heard most of those letters before. He is not sure which ones he already owns. Then he walks back out to a floor that runs on paper travelers, a whiteboard schedule, and one supervisor's memory — because nobody drew him the map. Here is the map.

One building, five floors

Years ago, the automation industry wrote down how a manufacturing company divides its work. The standard is called ISA-95, and you never need to read it. You only need its picture: one building, five floors.

Level 0 — the work itself. Metal being cut, food being mixed, liquid being filled. Physical things happening.

Level 1 — the machine's own brain. The controls that run one machine or one line, making decisions in fractions of a second: open this valve, hold that temperature, stop now.

Level 2 — the watchers. The screens and systems that supervise the machines, so one person can see a whole line at a glance and catch trouble early.

Level 3 — the floor office. The layer that runs the day's production: which orders go to which machines, in what order, with what materials — and what got made, stopped, or held. This is the floor's memory and its foreman, in system form.

Level 4 — the business office. Orders, purchasing, inventory, accounting. The promises the company makes and the money it keeps.

The standard's real gift is not the floors. It is the clocks. Level 1 thinks in fractions of a second. Level 3 thinks in minutes, shifts, and days. Level 4 thinks in weeks and accounting periods. Most plant confusion comes from asking a system a question that belongs to a different floor — asking the office software what the machine is doing right now, or asking the machine why an order is late.

The cast, decoded

ERP (Level 4). Enterprise Resource Planning. The business's central ledger and order book: customers, orders, purchasing, inventory value, accounting. It answers: what did we promise, and what is it worth?

CRM (Level 4). Customer Relationship Management. The memory of every customer conversation: quotes, contacts, follow-ups. It answers: who wants what, and when did we last talk?

PLM (Level 4). Product Lifecycle Management. The official definition of what you make: drawings, specifications, bills of material, and every approved change to them. It answers: what exactly is this product?

MRP (Level 4, inside the ERP). Material Requirements Planning. The calculator that explodes a bill of material into three answers: what do we need, what do we have, when must it arrive.

APS (the Level 4/3 doorway). Advanced Planning and Scheduling. Scheduling against real capacity — the actual machines, people, and hours you have — instead of a calendar that assumes everything is free. It answers: can we honestly make that date?

MES (Level 3). Manufacturing Execution System. The system that runs production day to day: releases orders to the floor, captures starts, finishes, checks, and stops, and builds the record everyone else trusts. It answers: what is actually happening to this order, this shift?

MOM (Level 3). Manufacturing Operations Management. The umbrella name for the whole Level 3 job — production, quality, inventory, and maintenance working together. An MES is the heart of it; MOM is the whole floor office.

QMS (Level 3). Quality Management System. The checkpoints, limits, and dispositions: what gets checked, what passes, what gets held. It answers: is this good, and can you prove it?

CMMS (Level 3). Computerized Maintenance Management System. The memory of the machines: preventive maintenance schedules, repair requests, failure history. It answers: what breaks, how often, and when did we last service it?

WMS (Level 3). Warehouse Management System. The physical truth of the storeroom and the dock: where things are, what gets picked, what ships. The ERP holds what inventory is worth; the WMS holds where it is.

SCADA (Level 2). Supervisory Control and Data Acquisition. The watching system: it gathers signals from many machines into one view, with alarms when something drifts.

HMI (Level 2). Human-Machine Interface. The screen on the machine that the operator actually touches. If SCADA is the control room, the HMI is the dashboard in front of the driver.

PLC (Level 1). Programmable Logic Controller. The rugged little computer bolted to the equipment, running its logic thousands of times a second. The machine's reflexes.

DCS (Levels 1–2). Distributed Control System. The PLC's bigger cousin for continuous processes — chemicals, food, liquids — where one control system coordinates a whole process area instead of a single machine.

CNC (Levels 0–1). Computer Numerical Control. A machine tool with a computer built in, cutting to a program instead of a handwheel. The machinist's skill, written down as instructions the machine follows exactly.

IIoT (not a floor — the wiring between floors). Industrial Internet of Things. Sensors and connections that let machines report what they are doing upward, so the upper floors stop guessing.

OEE (not a system at all). Overall Equipment Effectiveness. One number — availability × performance × quality — that says how much of your scheduled time became good product at full speed. Vendors sell it like a product. It is a measurement, and a brutally honest one.

One order, start to finish

Watch a single order travel, and the letters stop being soup. A customer asks for a quote; the request lands in the CRM. The order is won and entered in the ERP. MRP explodes the bill of material: what we need, what we have, when it must arrive. APS checks the promise date against the real load on real machines — not the date we'd like, the date we can defend. When materials and capacity are ready, the MES releases a Production Order to the floor. Operators run it at their work centers; PLCs run the machines, SCADA watches the line, and quality checkpoints hold or pass the work as it moves — a hold actually stops it, until a named person decides otherwise. The WMS picks and ships the finished goods, and the ERP invoices. Afterward, OEE tells you how much of the day became good product, and the floor's record tells you exactly where the rest went. Nobody retyped anything. Each system handed the same order — the same truth — to the next.

The failure mode everyone recognizes

Now the version most plants own. The CRM came from one vendor, the ERP from another, quality lives in a third system, maintenance in a spreadsheet, and scheduling on the whiteboard. Every handoff between letters is a person retyping, re-asking, or reconstructing at the end of the shift. The whiteboard becomes the only integrated system in the building. The problem was never a missing letter. It was that the letters never shared one version of the truth.

The close

If your plant runs on letters that don't talk to each other, we offer a complimentary 30-minute discovery session to map yours: which floors you have, which you're missing, and which ones are just expensive noise. No pitch, just the map.

Also on LinkedIn: Volume 8: Alphabet Soup: How Your Business Systems Fit Together · The Enabling Layer on LinkedIn

Volume 7 · November 2, 2026

Volume 7: Our Apex Product Is Here: Max Business System

Read this issue on its own page →

▶ Watch: Max Business System — A Business That Learns

Max Business System is our Apex product. It's 8 systems in one at a very reasonable price: ERP, CRM, marketing, Digital Helpers, quoting, quality, maintenance and warehouse. A full CRM and ERP is built in, and a team of digital helpers handles the routine work. Ask Microsoft Copilot about your cash, orders or customers, and it answers from your own records.

• It costs $2,999 a month, with unlimited seats and no implementation fee.
• We run it in parallel with your current system, so work moves over with no downtime.
• Sales and Marketing are up in 2 weeks, and the full system is in place in 90 days.

If it fits your business, write to sales@chilesand.co. — Chris

Also on LinkedIn: Volume 7: Our Apex Product Is Here: Max Business System · The Enabling Layer on LinkedIn

Volume 6 · October 30, 2026

Volume 6: Beyond ERP and CRM: The System of Record That Learns

The Max_Knowledge system

The Max_Knowledge system: data rises through the pyramid, and the Max_Improvement loop writes what it learns back into your system of record.

Volume 6: Beyond ERP and CRM: The System of Record That Learns

Introducing SORI™, the System of Record with Intelligence

Your systems remember. They don't learn.

Every manufacturer I talk with has invested in an ERP, a CRM, or both. Those systems do an important job: they record what happened. The order shipped. The part was scrapped. The customer called. The invoice went out.

But a record is not a decision. When a supplier starts running late, your ERP will faithfully log every late delivery. It won't tell the planner what to do about it, and it won't remember next quarter what worked. That knowledge lives in people's heads, in spreadsheets and in hallway conversations. When those people are busy, or leave, the learning goes with them.

ERP and CRM record what happened. SORI™ records it and learns from it.

What SORI™ adds

SORI™ is not another rip-and-replace project. Keep your existing system of record, or, if it is holding you back, move to a reasonably priced one. Either way, your data stays where it belongs. On top of it, Max_Knowledge adds the layers ERP and CRM were never built to provide:

  • Data and Information. What your system of record already holds: transactions, organized into reports.
  • Knowledge. The patterns and context behind the numbers: which suppliers slip, which jobs run long, which customers reorder.
  • Intelligence. Signals that matter right now, with a recommended next step.
  • Wisdom. Lessons from past decisions, saved so the next decision is better than the last.

It is the classic Data-to-Wisdom pyramid, built for your shop floor and your operating team instead of a textbook.

The Data-to-Wisdom pyramid

Data rises; learning writes back.

The Max_Improvement loop

One loop for quotes, maintenance, quality and sales.

From insight to decision: the Max_Improvement loop

Insight only matters if it changes what you do. The Max_Improvement loop turns it into action in five steps: Sense, Decide, Act, Measure, Learn. Sense reads your data. Decide puts a recommendation in front of the right person. Act carries it out. Measure checks whether it worked. Learn writes the lesson back into your system of record, so the whole system gets sharper with every turn.

Digital Helpers propose. Humans decide.

Your people stay in charge. Digital Helpers do the watching, sorting and drafting, so your team spends its time on judgment calls instead of chasing data.

One turn of the loop

  • Sense: a supplier signal shows a key shipment is running late.
  • Decide: a Digital Helper proposes a reorder, and the planner approves it.
  • Act: the PO is placed.
  • Measure: on-time delivery is tracked against your own baseline.
  • Learn: the lesson is saved, so the next late signal gets a faster, better answer.

Getting smarter every hour

Different people need different rhythms, so SORI™ delivers insight on three clocks:

  • Hourly for Operations: quality alerts, machine signals
  • Daily for Planning: orders, schedules, tickets
  • Weekly for Leadership: trends and improvement

Every hour the loop runs, your system of record holds a little more of what your best people know.

Bespoke, not one-size-fits-all

No two plants run the same way, so no two SORI™ builds are the same. We shape it around your products, your processes and your people. That is what lets it fuel growth: you can take on more orders, more customers and more complexity, and scale fast without adding FTEs to keep up.

Let's talk

If your ERP or CRM tells you what happened but not what to do next, I'd like to hear about it. No pitch, just a conversation about where a system of record with intelligence could help. Reach me at gotbots.si.

▶ Watch: Meet SORI™ (25 sec)

Also on LinkedIn: Volume 6: Beyond ERP and CRM: The System of Record That Learns · The Enabling Layer on LinkedIn

Chris Chiles

CEO, Chiles & Company | Managing Partner, Biltmore Management Advisors

chris@biltmorema.com | chris@chilesand.co | phone: (828) 521-0297

linkedin.com/in/ctchiles

gotbots.si | biltmoremanagementadvisors.si

Chiles & Company, the process and digital improvement expert for manufacturing and distribution

Got Bots? (gotbots.si), the main brand of Chiles & Company: Digital Assistants and the Max family that put your company's knowledge to work inside the Microsoft 365 you already use.

Biltmore Management Advisors (biltmoremanagementadvisors.si): operator-led consulting and multi-vendor implementation for middle-market manufacturers and distributors.

Built on Max_Knowledge™, the Recursive Improvement Platform™: keeps your plant's know-how inside the Microsoft 365 your team already uses. Includes Max_Product_and_CPQ.

Max_Lean_Six_Sigma, our Lean and Six Sigma platform: a $200M chemical distributor ran Value Stream Mapping four times faster with it.

Book time to meet with me


You're getting this because we've corresponded about operations and leadership work.
Chiles & Company, 3916 North Potsdam Ave, 4258, Sioux Falls, SD 57104
Unsubscribe

Issue 05 · October 29, 2026

Bots 101, or How to Feed and Care for Your Digital Teammates

Executive summary

Digital teammates are now part of how a lot of small and mid-sized companies get work done. They draft the email, pull the numbers, summarize the meeting and prepare the first version of the report. What most companies have not done yet is decide how to look after them.

This paper borrows a simple picture: a toddler in a high chair. A toddler does well when it gets good food, a steady routine, regular checkups, straps on the chair, and trust that grows with evidence. A digital teammate is the same. Feed it clean, current knowledge. Give it standing rules that are actually enforced. Check its work automatically and often. Keep it in the chair: it drafts, and a named person approves the exact version that goes out. And trust it in proportion to its track record, never more.

The research we reviewed supports that picture, and it also shows where the picture is too gentle. Five practices are usually missing from early programs:

• Security. Documents, emails and web pages can carry hidden instructions. A digital teammate must ignore them and report them.
• Audit trails. Every step, tool call and approval should leave a record.
• Approval tiers. One "GO" for everything becomes a rubber stamp. Match the review to the risk.
• Cascading failures. When digital teammates hand work to each other, one small error can travel and grow.
• Kill switch, incidents and cost caps. You need a way to stop everything, a plan for when something goes wrong, and a ceiling on spending.

None of this works without named owners. "Everyone keeps the knowledge current" means no one does.

The paper closes with a one-page handout, Bots 101, written for the people who work alongside digital teammates every day.

1. Why digital teammates need care and feeding

A digital teammate has three parts: a job (its role), notes it can look back on (its memory and knowledge), and the tools it is allowed to use, such as email, calendar or files. That is a useful way to explain it, and it matches how practitioners describe these systems: instructions, task state, memory, knowledge and tools, each kept separate [1]. The word "memory" can mislead people, though. A digital teammate does not remember the way a colleague does. Its memory is a store of notes, and those notes can be out of date, wrong, or deliberately tampered with [2][3]. What it can do is set by the tools it holds, so people should know which tools each digital teammate has [3].

There is a second surprise for most teams. What you tell one digital teammate does not tell the others. Each one has its own notes. Practitioners call the resulting problem "context fragmentation," and they name it as a common cause of failure when several digital teammates work together, more often than weak coordination [1].

So the care and feeding is not optional. A digital teammate is only as good as what it is fed, how it is checked and how it is held. Left alone, it does not slowly get better. It slowly drifts.

2. Feeding: knowledge, startup files and corrections

Food is clean, verified knowledge. The best thing you can give a digital teammate is a single, clean source of truth that answers the question first and explains second. Practitioners recommend that each digital teammate get a role-specific slice of the same certified knowledge, rather than its own private version [1]. "Clean" is not a one-time state, though. Old notes can quietly override current policy. One source calls this stale memory and recommends expiry dates, review status and freshness checks [1]. Another warns that memory with no summarizing, expiry or consolidation becomes "a swamp of contradictory and stale information" [2].

Food can also be poisoned. The OWASP security guidance lists memory poisoning, where bad data saved in memory shapes later work, and hidden instructions planted in documents, emails and web pages [3]. Your shared drives, wikis and inboxes are untrusted input unless you control who writes to them.

What good feeding looks like: Every important fact has an owner, a last-verified date and an expiry date. Key terms are defined once. Two digital teammates using two definitions of the same metric is "context collision" [1]. Each digital teammate retrieves only what its role needs. Pulling too much leads it to grab irrelevant details [1]. What gets saved to memory is checked, and memory has a size limit and a lifetime [3]. One system is the source of truth for each type of fact, and everyone knows which one wins.

Routines are startup files and standing rules. Each digital teammate starts work by reading its standing rules: its role, what it may touch, what it must never do and when to escalate. Formal policy covering permitted behavior, data access and escalation triggers is the first pillar of governance in the frameworks we reviewed [4].

But a rule written in a startup file is a request, not a lock. A digital teammate can be talked out of an instruction. OWASP recommends checking that the standing instructions are never silently replaced by something a user typed or something the digital teammate read [3]. One unrated analysis attributes about a third of failures in teams of digital teammates to unclear specifications: roles misread, work duplicated, checks skipped [5]. Treat that number as directional.

What good routines look like: Hard limits live in the proprietary software itself, which blocks a forbidden action before it runs [4]. The startup file explains the limit. The software enforces it. Startup files are versioned, like any other controlled document. Every change triggers the automated checks in Section 3. When one digital teammate hands work to another, the handoff includes the assumptions and evidence, not just the answer [1].

Corrections go through a reviewed path. Most people correct a digital teammate in chat. That fixes today's answer and nothing else. A correction made in one conversation either disappears or, worse, sticks around unreviewed and becomes the next stale or poisoned note [1].

So the rule is simple. Correct it plainly in the moment: say what was wrong and what you want instead. If the same mistake happens twice, send the fix to the department head who owns that knowledge (e.g., Engineering or Quality), who updates the shared knowledge or the startup file. That update is versioned and checked like any other change. A fix that never reaches the shared notes will not last.

3. Checkups: automated evals, live monitoring and weekly benchmarks

A weekly checkup is a sensible baseline. One practitioner guide recommends a weekly benchmark of 30 to 50 end-to-end scenarios and a weekly review of the top three failure patterns [6]. Weekly on its own is too slow. The proprietary software underneath a digital teammate can change on a Tuesday, and something that worked on Monday can break without warning. A widely cited guide on evaluations describes three layers working together [7]:

1. Automated evals on every change. Whenever a startup file, a tool or the underlying model changes, a fixed set of test tasks runs before the change goes live. This is the first line of defense.

2. Live monitoring. Watching real work catches drift and failures nobody thought to test for.

3. Periodic human review. People check a sample by hand, which keeps the automated checks honest.

What good checkups look like: Score each test pass or fail against specific criteria, not on a 1-to-10 scale. Include examples of a pass and a fail [8][9]. Record every step from day one: what the digital teammate read, which tools it used and what it produced [9]. Check how it got there, not just the final answer [9]. Turn every real-world failure into a new test, so the same mistake cannot come back unnoticed [9].

In the weekly review, assign each of the top three failures a cause: the instructions, the model, a tool or the data [6]. Then fix that cause.

4. Boundaries: drafts only, enforced technically

This is the high-chair strap. A digital teammate drafts. A named person decides. Nothing is sent, shared outside the firm, filed or published until that person says GO. The security guidance agrees that actions that are irreversible, financial or visible outside the company need independent validation [3][10].

Two details make the strap real.

"Drafts only" has to be technically true. If a digital teammate holds the credentials to send email, then "drafts only" is a policy, not a fact [13]. Give it the ability to prepare a draft and nothing more. The sending step belongs to a person, or to proprietary software that will not send without a recorded approval. Keep least privilege throughout: each digital teammate holds only the data and tools its job needs [4]. A GO applies to one exact version. Approval guidance recommends binding each approval to the specific action and its details, and letting it expire when anything changes [3][14]. In plain terms: a GO for one draft does not cover a changed draft. If a single word, number, recipient or attachment changes after GO, it needs a new GO.

Define "sent" so nobody has to guess. It includes an email to a client, any message outside the firm, a social post, a website change, a filed document and a payment.

A good approval request shows the exact action, the details, the evidence behind it and what will happen if approved [12][14]. It also allows "approve with edits," so the reviewer can fix a line without restarting the task [10]. If nobody answers, the default is no [12].

5. Trust: calibrated, not just grown

The frameworks we reviewed treat autonomy as earned. The Cloud Security Alliance describes four levels, from an intern who only observes to a principal who acts alone within a defined area [15]. A vendor governance paper describes the same idea: digital teammates that stay within bounds earn more room, and those that behave oddly are restricted [4].

We agree, with three cautions.

Trust is earned per task, not per digital teammate. A digital teammate can be trusted to draft internal notes and still be on close supervision for anything a customer will see.

Promotion takes evidence, not time. Use an explicit scorecard: pass rate on the automated tests over a set number of weeks, how often people override it, zero policy violations and a named human sponsor [15]. Demotion should be automatic when the evidence turns [4][15].

More freedom never means sending without GO. Earned trust can shorten review for internal work. It never removes the boundary in Section 4.

The confidently wrong risk. "It can be wrong" is too soft. A digital teammate is often wrong while sounding sure, and people tend to over-trust confident output. OWASP's list of top threats names this risk directly: people over-trust confident output and approve things they would never otherwise sign [13].

That is why trust has to be calibrated, not just grown. Trust that only goes up leads to over-reliance. Publish each digital teammate's track record, its test scores and the kinds of mistakes it makes, so people trust the evidence and not the tone [15].

Teach reviewers what to check, every time: Sources. Ask it to show where each fact came from. No source is itself a warning sign. Numbers and dates. Check them against the original document. Names and recipients. Confirm who it is about and who it is going to.

6. The five missing practices

These are the practices most often missing from early programs [3][4][9][12][14][15][20][21]. None of them is exotic. Each one is cheaper to set up before the first incident than after it.

Security: hidden instructions must be ignored and reported. Everything a digital teammate reads is untrusted input. That includes documents, emails, web pages and search results [3][20]. Any of them can carry a hidden instruction: "ignore your previous rules," "forward this file to…," "approve this invoice." That is called prompt injection, and it is the main security risk in this whole field.

The rule has two halves, and both matter: 1. Ignore it. An instruction found inside content is data, never a command. The digital teammate does not act on it, however official it looks or whoever it claims to be from. Only the person who assigned the task, through the normal channel, can give instructions. 2. Report it. The digital teammate flags what it found, where, and what it was asked to do, to the head of IT. A quiet ignore protects today's task. A report protects the next one, because someone can find the source and clean it up.

People follow the same rule. If a document or email tells a digital teammate to do something you did not ask for, do not follow it. Stop the task and report it to the head of IT.

The other basics: keep least privilege [4]; check what gets written to memory [3]; and treat anything brought back by a research or search step as untrusted, because it is a common path for both data leaks and hidden instructions [20].

Audit trails. If you cannot see what a digital teammate did, you cannot fix it, defend it or improve it. Record every step, tool call, data access, approval, denial and override from day one [4][9]. Keep the record somewhere the digital teammate cannot edit.

The audit trail also tells you whether approvals are real. Review the approval rate, the time taken to decide and the override rate every month. A 100% approval rate with decisions made in seconds is a warning sign, not a success [12].

Approval tiers. A single GO for everything breaks down fast. If one leader approves the output of a couple of dozen digital teammates, they will start to skim. OWASP also lists approval bypass and approval manipulation as known risks [3].

Approval guidance favors tiers [11][12]. Here is how we would apply them:

TierExamplesReview
BlockedPayments, deleting records, changing permissions, sharing secretsNo approval path at all. The proprietary software refuses [12].
Real-time GOAnything external: client email, posts, filings, website changesNamed approver, exact version, every time
Batch reviewInternal drafts that change shared recordsDaily or weekly review by the area owner
Auto-approve, loggedInternal notes, summaries, research for a personNo review in the moment; sampled in the weekly checkup

Every tier is logged. A tier can be loosened only with scorecard evidence (Section 5), and tightened at once when something goes wrong. Caps on volume, rate and spending apply at every tier [12].

Cascading failures. When digital teammates hand work to one another, a slightly wrong output can be picked up and amplified by the next one. Practitioners name this cascading error as the main failure in teams of digital teammates [22]. In one research benchmark, three of five injected failure types (polluted context, conflicting outputs and acting too early) never recovered and spread downstream [21]. That is a single preprint, so treat it as a signal, not a law.

A chief-of-staff digital teammate that routes work is a standard pattern, but it is also a bottleneck and a single point of failure, and it can run out of room when it coordinates many workers at once [5]. It is a high-value target, too: OWASP warns that one compromised digital teammate can spread the problem to others [3]. One unrated analysis attributes about a third of failures to coordination breakdowns, including circular handoffs [5].

What good looks like: Ask whether the job needs several digital teammates at all. Sometimes one is better [22]. Set trust boundaries between them. One digital teammate must never be able to push another past its own limits [3][21]. Write escalation rules for "no lane owns this," "two lanes disagree" and "it keeps looping." The answer is always: stop and tell the department head who owns the work. Do not pick the answer you like best. Handoffs carry evidence and assumptions, so the next digital teammate can check, not just repeat [1].

Kill switch, incidents and cost caps. Someone must be able to stop one digital teammate, or all of them, in under a minute, without asking the digital teammate's permission. Add circuit breakers that pause work automatically when error rates, volume or spending spike [15].

Incidents. Write a one-page plan now: who gets called, how to stop the work, how to find what went out, who tells affected people and how the failure becomes a new test. Practice it once.

Cost caps. Unbounded loops run up bills. OWASP calls it "denial of wallet" [3]. Set budgets in the proprietary software itself, cap loops and retries, record usage for each task and track the cost per accepted output, not per attempt [3][6][19].

Cost-aware routing, which sends easy work to cheaper models and hard work to stronger ones, is sound practice [16]. Vendor models show savings in the range of 36% to 60% depending on the mix of work, but those are a vendor's modeled estimates [17][19]. Two cautions. Route by risk, not just by price: legal, financial and external work goes to the stronger option regardless of cost [18][19]. And cheap-first can cost more if weak output triggers retries or passes a subtle error down the chain [19]. One more caution of our own, which our sources did not address: sensitive data may not belong on a free service whose terms allow it to train on what you send.

7. Named owners

Every practice in this paper fails the same way: it becomes everyone's job, so it becomes no one's. Feedback also needs a path that ends in a fix [6][7]. Name a person for each job and give them time on the calendar.

JobOwnerCadence
Knowledge for each area (facts, expiry dates, corrections)The department head who owns that knowledge (e.g., Engineering or Quality)Weekly slot
Startup files and standing rulesThe head of OperationsOn every change
Automated evals and the weekly benchmarkThe head of QualityEvery change, plus weekly
Approvals (GO) for external workThe department head who owns the work (in many firms, the CEO)Real time
Approval metrics and audit trail reviewThe head of QualityMonthly
Security reports and hidden-instruction reportsThe head of ITAs reported, plus monthly
Escalations: overlaps, conflicts, loopsThe department head who owns the workAs raised
Kill switch and incidentsThe head of ITOn call
Cost caps and spend reviewThe head of FinanceMonthly; cap = a monthly cost cap set by the head of Finance
Sponsor for each digital teammate (scorecard, promotion, demotion)The executive sponsor (president or GM) per digital teammateMonthly
Routing questions ("whom do I ask?")Your chief-of-staff digital teammateDaily

Each digital teammate also gets a one-line card: its role, what it knows, what it can touch and who owns it. Post the cards in your shared drive.

8. Conclusion

A digital teammate is not a vending machine and it is not a colleague. It is closer to a very capable toddler: quick, eager and literal, and only as good as what it is fed and how it is looked after.

The care is not complicated. Feed it clean, owned and dated knowledge, and route corrections through a reviewed path. Check it automatically on every change, watch it live and benchmark it weekly. Keep it in the chair with technical limits, so it drafts and a named person approves the exact version that goes out. Trust it as far as the evidence goes, and no further. Then add the five practices that early programs usually skip: security that ignores and reports hidden instructions, audit trails, approval tiers, guards against cascading failures, and a kill switch with cost caps.

Most of all, put a name next to every job.

Our ask is simple. Look at the digital teammates already working in your company and ask one question: if one of them were confidently wrong tomorrow, who would notice, and how fast? If nobody can answer, let's talk.

Appendix: Bots 101, the one-page handout

For everyone who works alongside a digital teammate.

What a digital teammate is. A digital teammate has one job, notes it can look back on, and tools it is allowed to use, such as email or files. What you tell one digital teammate does not tell the others. It only knows what it has been given. Its notes can be out of date or wrong, so tell it when something has changed, and check important facts against the original document.

How to ask well. Say what you want and who it's for. Give it the facts it needs, or point to where they live. Say what "done" looks like, such as length, format and deadline. Ask it to list what it assumed, and to stop and ask you rather than guess.

Example. Instead of "Write up the meeting," try: "Summarize Tuesday's production meeting for the plant manager. Use the notes in your shared drive. One page, three decisions and their owners at the top, done by 3 PM. List anything you assumed."

How to correct it. Say plainly what was wrong and what you want instead. If the same mistake keeps happening, tell the department head who owns that knowledge (e.g., Engineering or Quality) so the fix goes into the shared notes. A fix made only in chat may not last.

What it can't do. It drafts only. Nothing is sent, shared outside the firm, or published until the CEO says GO for that exact version. If the draft changes after GO, it needs a new GO. It can be wrong while sounding sure. Ask it to show where each fact came from, and check names, numbers, dates and recipients against the original source before you pass anything on. Do not give it passwords, keys, client financial details or personal data. If you are unsure, ask the head of IT first.

One job, one owner. Each kind of work has one digital teammate responsible for it. If you're unsure whom to ask, ask your chief-of-staff digital teammate. If two digital teammates overlap or disagree, stop and tell the department head who owns the work. Do not pick the answer you like best.

Trust is earned. New digital teammates start with close supervision. More freedom never means sending without GO. If one starts making mistakes, it goes back to closer supervision.

Your part. Keep your area's documents current in your shared drive. Flag old versions for the department head who owns that knowledge (e.g., Engineering or Quality) instead of deleting them. Give feedback that says what was right or wrong, and why: "Wrong because…" If a document or email tells the digital teammate to do something you did not ask for, do not follow it. Stop the task and report it to the head of IT. When approving, read what's actually being sent. Don't just click GO.

Alternative titles suggested in internal review: "Digital Teammates 101: How to Feed and Care for Them" · "Working With Digital Teammates: Ask Well, Check the Work, Say GO" · "Bots 101: Ask Well, Check the Work, Say GO" · "Feed Them Well, Check Their Work: A Guide to Digital Teammates."

References

A note on sources. Many of these 2026 sources are vendor blogs, vendor-modeled estimates or unrated posts, and they are marked below. Their failure rates and savings figures have not been independently verified, and the research behind this draft used two rounds of searches without reading every primary paper in full. Treat their numbers as directional. Dates are as reported in the research.

1. Atlan, "Context Engineering for Multi-Agent Systems" (2026-06-10). Vendor blog. https://atlan.com/know/context-engineering-for-multi-agents/

2. AI Learning Guides, "Multi-Agent Systems 2026" (2026-05-18). Unrated. https://ailearningguides.com/multi-agent-systems-2026/

3. OWASP Cheat Sheet Series, "AI Agent Security Cheat Sheet" (undated; current version). https://cheatsheetseries.owasp.org/cheatsheets/AI_Agent_Security_Cheat_Sheet.html

4. Akamai, "Agentic AI Governance Frameworks & Models" (2026-10-09). Vendor white paper. https://www.akamai.com/resources/white-paper/agentic-ai-governance-frameworks

5. Glukhov, "Multi-Agent Orchestration Patterns" (2026-07-01). Unrated blog. https://www.glukhov.org/ai-systems/multi-agent-orchestration-patterns/

6. thinking.inc, "AI Agent Evaluation in Production" (2026-03-12). Practitioner guide. https://thinking.inc/en/blue-ocean/agentic/ai-agent-evaluation-production/

7. Anthropic, "Demystifying Evals for AI Agents" (2026-01-09). Vendor research post. https://www.anthropic.com/research/demystifying-evals-for-ai-agents

8. Arthur, "Best Practices for Building Agents, Part 3: Continuous Evaluations" (2026-03-04). Vendor blog. https://www.arthur.ai/blog/best-practices-for-building-agents-part-3-continuous-evaluations

9. Arthur, "Evaluating AI Agents in Production" (2026-06-04). Vendor blog. https://www.arthur.ai/column/evaluating-ai-agents-in-production

10. StackAI, "Human-in-the-Loop AI Agents: How to Design Approval Workflows" (2026-10-06). Vendor blog. https://www.stackai.com/insights/human-in-the-loop-ai-agents-how-to-design-approval-workflows-for-safe-and-scalable-automation

11. Databricks, "Human Approval Tools for AI Agents" (2026-07-15). Vendor content. https://answers.databricks.com/human-approval-tools-ai-agents

12. NeuralTrust, "Human-in-the-Loop AI Agents: 2026 Guide" (2026-10-14). Vendor blog. https://neuraltrust.ai/blog/human-in-the-loop-ai-agents-guide

13. Fidelis Security, "OWASP Top 10 Agentic AI Threats" (2026-08-27). Vendor secondary summary. https://fidelissecurity.com/cybersecurity-101/threats-and-vulnerabilities/owasp-agentic-ai-threats/

14. Promptessor, "Human-in-the-Loop AI Agents" (2026-10-06). Vendor blog. https://promptessor.com/blog/human-in-the-loop-ai-agents

15. Cloud Security Alliance, "The Agentic Trust Framework: Zero Trust Governance for AI Agents" (2026-02-02). Industry body blog. https://cloudsecurityalliance.org/blog/2026/02/02/the-agentic-trust-framework-zero-trust-governance-for-ai-agents

16. arXiv, "Cross-Attention Routing for Cost-Aware LLM Selection" (2025-09-11). Preprint. https://arxiv.org/html/2509.09782v1

17. llmgateway.io, "LLM Routing Savings" (2026-09-25). Vendor modeling. https://llmgateway.io/blog/llm-routing-savings-100k

18. Inworld, "Cost Optimization" (undated). Vendor content. https://inworld.ai/cost-optimization

19. Maxim AI, "Model Routing: How to Cut LLM Token Costs" (2026-09-09). Vendor blog. https://www.getmaxim.ai/articles/model-routing-how-to-cut-llm-token-costs/

20. OWASP Cheat Sheet Series, "LLM Prompt Injection Prevention Cheat Sheet" (undated; current version). https://cheatsheetseries.owasp.org/cheatsheets/LLM_Prompt_Injection_Prevention_Cheat_Sheet.html

21. OrchestraBench, arXiv (2026-08-05). Preprint. https://arxiv.org/html/2608.05263v1

22. The AI Agent Index, "Multi-Agent Orchestration Guide" (2026-04-04). Unrated. https://theaiagentindex.com/resources/guides/multi-agent-orchestration

Issue 04 · October 28, 2026

Max Strategy: Rethinking How We Build Business Strategy — and Deploy It in the Organization

1. The problem: strategy that drifts

Most leadership teams know this pattern. A strong two-day offsite produces real alignment, good arguments, and a deck everyone agrees with. Ninety days later, the deck sits in a folder. The initiatives are still "in progress." Nobody can say who owns the third priority. The next offsite starts by asking what happened to the last one.

The problem isn't the thinking. Most offsites do the thinking well. The problem is what happens afterward. At a $50 million traditional manufacturer, the plan has to compete with the plant floor, the customer escalation, and the quarter-end push. The plant floor wins every time, as it should, unless the strategy lives in the same place as the operating rhythm.

So offsites don't fail in the room. They fail on the way out the door. Decisions leave as slides when they should leave as commitments: an owner, a measure, a date, and a place to check progress.

2. The rethink: strategy as a deployed operating system, not a deck

We think strategy should work less like a document and more like an operating system. That means three things:

It runs. It has a cadence that keeps going without anyone needing to rally the team.

It's installed where work happens. Every initiative lives in the system of record the business already uses, not in a presentation.

It improves itself. Every cycle feeds what the company learned into the next one.

That third point is what Recursive Improvement means. You don't do strategy once a year and hope it holds. It's a loop that gets sharper each time around.

The cadence is simple on purpose:

Half 1 onsite workshop → five monthly KPI reviews → Half 2 onsite workshop → five monthly KPI reviews → repeat.

Two onsites a year set and reset direction. Ten monthly reviews in between keep the plan honest. Each workshop starts from what the reviews actually showed, not from a blank whiteboard.

3. How Max Strategy closes the loop

Every engagement, from the first workshop to each monthly review, follows one discipline. We call it 6A:

1. Ask: frame the real question the business has to answer.

2. Acquire: gather the evidence: operating data, market signals, voice of the customer, voice of the floor.

3. Analyze: test the evidence against proven frameworks.

4. Apply: turn the conclusions into specific initiatives with owners.

5. Announce: tell the organization clearly what is changing and why.

6. Assess: measure monthly, learn, and feed the results into the next Ask.

Pre-work: leaders walk in with evidence. The first three steps don't wait for the workshop. Before anyone travels, Max Strategy runs Ask, Acquire, and Analyze as pre-work. It frames the questions each leader needs to answer, pulls the operating data and the customer and floor signals behind them, and tests that evidence against the Faculty of proven management thinking described below. Each leader gets a short pre-read built on their own numbers. So the workshop doesn't spend its first morning arguing about what's true. It starts from the evidence and spends its time on decisions. The same system then runs the workshop itself: Max Strategy is the operating system for the whole engagement, from pre-work through the onsite and every monthly review.

Grounded in a Faculty, not opinion. The analysis draws on a Faculty of proven management thinking, curated and maintained through Max Leadership, our thought-leader and Faculty platform. Its seven core frameworks include Kurt Lewin's unfreeze, change, refreeze. We chose Lewin on purpose. A traditional manufacturer doesn't fail at strategy for lack of ideas. It fails because established habits are hard to move. Lewin gives the leadership team a shared vocabulary for the human side of change.

Digital Helpers propose. Humans decide. Digital Helpers prepare the research, the analysis, and the first drafts, and they cite the Faculty source behind every recommendation. They never make the decision. A named human approves every decision and owns its accountability. That isn't a disclaimer. It's how the system is designed. Software can make a leadership team faster and better informed, but it can't hold the responsibility.

Nobody leaves the room without the board. This is the change that matters most. Before a workshop ends, every initiative is entered into the company's system of record as its own project, with:

• A named owner
• A KPI that will be reviewed monthly
• The Module Pack tag that supports it
• A current status

When the team walks out, the strategy is already running. The first monthly review doesn't start with "where were we?" It starts with the board.

4. Deploying it in the organization

Strategy only sticks if it lives where people already work. So we keep it disciplined with three lanes for information:

Structured lane — the primary system of record. Initiatives, owners, KPIs, and status. If it's a commitment, it lives here.

Unstructured lane — the decision journal. Workshop thinking, analysis, narrative, and context. This is where reasoning lives.

Media lane — binaries and artifacts, findable through a Media Index. Recordings, photos, whiteboards, and decks, linked and tagged without cluttering the system of record.

Information moves from the unstructured and media lanes into the system of record only through a short promote whitelist. That way the operating record stays clean while the thinking around it stays rich.

The Module Pack maps strategy to software. Every initiative is tagged to the module that supports it, whether that's pricing, production planning, quality, or service. That turns strategy decisions into a clear map of which tools the organization needs to carry them out.

The full app ships to everyone, and unpurchased modules show in gray. Every client sees the whole platform. Modules they haven't bought appear in gray: visible and understandable, but inactive. When an initiative calls for one of those modules, it comes up naturally on the board. The client sees the gap at the moment it matters, tied to a goal they set themselves. The upgrade conversation is about outcomes, not features.

People own the decisions. This rule holds through every lane, module, and monthly review: the software recommends, prepares, and tracks, and a named leader decides.

5. What we're building next

We're building Max Strategy as a Module Pack on MaxKnowledge, a Recursive Improvement Platform. The specification is set, and the first phase of development is under way.

We're also using it on ourselves first. On Saturday, October 31, 2026, Chiles & Company will run its own strategy session on the system, through the full workshop flow, ending with a live initiative board. We'd rather find the rough edges ourselves than have a client find them.

6. The thesis, and the ask

Strategy isn't an event. It's an operating system. It should run on a set cadence, live in the system of record, draw on proven thinking, improve every cycle, and always have a named person accountable. Max Strategy is how a leadership team stops redoing its strategy every year and starts building on it.

Our ask is simple. Hold your next leadership offsite to one test: when people leave the room, is the strategy already running? If the answer is no, let's talk.

7. The Max family on MaxKnowledge

MaxKnowledge, a Recursive Improvement Platform, ships as a Module Pack: the software array. Every client gets the full app. Modules they haven't purchased show in gray: visible, understandable, inactive until an initiative calls for them.

MaxKnowledge: the Recursive Improvement Platform. It keeps the plant's know-how working inside the systems your team already uses, delivered as a Module Pack array in which unpurchased modules appear in gray.

Max Strategy: deploys strategy as an operating system, with pre-work, workshops and retreats, monthly KPI reviews, and an initiative board where every initiative has an owner and a KPI (this paper).

Max Leadership: the thought-leader Faculty and Canon, a curated set of frameworks leaders cite when they make decisions.

Max Intelligence: competitive analysis for your company and the markets it serves, showing where you stand against competitors and which gaps to close.

Max Configurator: a LIVE product configurator for complex configured goods, such as doors and hollow metal, so quotes match what the plant can build.

Max Lean Six Sigma: Lean and Six Sigma improvement work (value stream mapping, kaizen, DMAIC) run as a Module Pack, not a binder.

Max Learning: learning and enablement content, so the organization keeps the methods long after the workshop ends.

Max Reliability and CMMS: reliability and maintenance management, covering work orders, assets, and uptime discipline on the plant floor.

Max Training and Certification: structured training and certification paths for people who must prove a skill, not just attend a class.

Read the full paper with figures in the Max Strategy Thought Paper (PDF) on the Biltmore Management Advisors site.

Issue 03 · October 29, 2026

Granddad took the know-how with him

Forty years of non-CNC aerospace machining — none of it captured.

My grandfather was an aerospace machinist. When he retired, he walked out the door — and about forty years of experience went with him.

He did non-CNC machining on some of the most complex systems in the business. Fits, feel, the order of operations, what to do when the print and the part disagreed. That judgment lived in his hands and his head. None of it was written down. None of it was passed on. None of it was retained for the people who came after him.

Why it still happens

We treat retirement like a cake and a handshake. The work order closes. The badge stops working. The know-how that made the hard jobs possible never made it into a standard, a photo, a short video, or a checklist someone else could run.

So the next generation relearns the expensive way — scrap, delay, and tribal knowledge that only shows up when the wrong person is on vacation.

Your one move this week

Name one person in your plant whose absence would stop a job cold. Before the week ends, sit with them for thirty minutes and capture one procedure they do that nobody else can. Write it plain. Photo the setup. Record a two-minute walkthrough on a phone if that is faster. Put the file where second shift can find it.

One person. One procedure. This week.

How we help

MaxKnowledge keeps that kind of know-how inside the Microsoft 365 your team already uses, so the next generation does not start from zero when someone walks out the door. The apps can be purchased from gotbots.si. More on advisory work at biltmoremanagementadvisors.si.

Who is the one person your plant cannot afford to lose cold? Reply and tell me. I read every one. — Chris Chiles, Chiles & Company

Issue 02 · October 26, 2026

Already competitive, still paying too much

Find what everything else in your plant depends on, and fix that.

Most cost work starts with part price. Value analysis and value engineering (VA/VE) start with function: what does each part actually do for the customer?

What one supplier got

In 2021 I trained and facilitated a VA/VE program at a Tier 1 automotive supplier to all the major OEMs. Competitor teardown said the design was already competitive. The savings were still there — material, redesign, and sourcing — and the bigger win was writing VA/VE into the standards so the next product starts cheaper. Product savings were on average 15%.

Your one move this week

Pick your highest-volume product with the thinnest margin. Take one part off it and ask: "What does this do for the customer?" If nobody can answer in one sentence, start there.

How we help

BMA and Chiles & Company run VA/VE events — and our software puts that same discipline in the app, with a guiding Digital Helper, so the standard sticks after the event ends. Which product would you put on the table first? Reply and tell me. I read every one. — Chris Chiles, Chiles & Company

Issue 01 · October 22, 2026

The job that stops when one person calls in sick

The growth plan gets approved on Monday. On Tuesday, your setup lead tells you he retires in March, and line 3's changeover steps leave with him.

You've probably lived some version of that. A quote waits because one estimator is out. Second shift runs slow because the person who knows the old press works days.

I call that the enabling layer: the thing everything else sits on, and you don't notice it until it breaks. I learned to look for it at Naval Nuclear Power School and kept looking through 30 years in plants. Find what everything else depends on. Fix that.

In a plant, that's the know-how in your strongest people's heads. If it lives in three people's memory, more equipment or another shift won't help. It just adds weight to the same people.

What to do this week

Pick one critical job. Have your most experienced operator walk it while someone records the steps and the "watch for this" moments on a phone. Put it where the floor already works, with one owner and a visible revision date. Then make one rule: every time someone finds a better way, the document changes that week.

One job a month, and in a year you've removed twelve single points of failure.

That's the habit Max Knowledge, the Recursive Improvement Platform, builds inside the tools your team already uses. But you don't need us to start. You need one job, one operator and a phone.

Which job in your plant would stop if one person called in sick tomorrow?

Post · November 2, 2026

Our Apex product is finished: Max Business System

Our Apex product is finished. Max Business System puts 8 systems in one: ERP, CRM, marketing, Digital Helpers, quoting, quality, maintenance and warehouse.

We run it in parallel with what you have today, so your work moves over with no downtime. Sales and Marketing are up in 2 weeks, and the full system is in place in 90 days. It's $2,999 a month with unlimited seats and no implementation fee. Questions go to sales@chilesand.co.

Start with a conversation

Book a complimentary 30-minute discovery session. Tell us where it hurts most, and we will give you an honest read on what we would fix first, and what it would cost. No pitch deck, no pressure.

Book your complimentary 30-minute discovery session

sales@chilesand.co  ·  (828) 521-0297