Your AI Writes Beautifully. Your Reports Still Bounce Twice Before They Go Out.

Engineer Your Freedom

July 2026  |  Issue No. 8

The Institutional Knowledge Problem

Your AI Writes Beautifully.
Your Reports Still Bounce Twice
Before They Go Out.

The gap was never the writing. It is that nobody ever wrote down how your firm writes.

A senior engineer on my team was reviewing another firm's draft of a regulatory site application last week. He had it open for a few minutes before he called it. It read like a third-year EIT had assembled it. Half the forms were missing. It met the minimum and stopped there. He made a point of saying it was a good firm and that he was not trashing anyone. But you could see who built it before you ever saw the name on the cover.

That is just as true inside your own walls. Your best PM turns in a basis-of-design report that reads tight and lands clean. The new EIT turns in one that is technically correct and structurally a mess, and you already know it is going to bounce off your reviewing PE twice before it reaches the client. The engineering underneath both is fine. The difference is in how it was written. And that difference is costing you margin every week.

The AI Pilot Problem

When I first put AI to work on our narrative deliverables, I assumed the problem would be the writing itself. Early on there was a version of that. It was wordy. It ran sentences on. It padded a clean thought into a full paragraph. But that turned out to be the easy part. You learn to train it out in about a week. Once the prose was tight, what I was left with was the harder problem. It read like a competent stranger had written a report about water infrastructure, because that is exactly what happened. The model could write. It just had no idea how we write.

This is the part that is easy to misread. A firm runs a pilot, the prose looks sharp, leadership calls it impressive and hands out licenses. Then the deliverables still come back needing heavy editing and nobody can quite say why. The prose was never the bottleneck. The bottleneck is that your firm has a specific way of structuring a technical report, a specific tone for a response-to-comments letter, a specific set of criteria sections your reviewers expect to see in a specific order, and none of it is written down anywhere a machine can read. It lives in the heads of two or three people who learned it by doing it for fifteen years.

"AI does not have a writing problem. It has an instructions problem. The firm that writes down how it writes turns its best deliverables into something any engineer on the team can produce."

The Real Cost of Knowledge Locked in Heads

That knowledge walks out the door at five o'clock, and it walks out for good when one of those people leaves. This is not a soft cost.

1 day

Lost per week per knowledge worker hunting internal info

(McKinsey)

42%

Of what an employee knows is unique to them and irreplaceable

(Panopto)

In an engineering firm, that 42 percent is usually your best PE's hard-won judgment about exactly how your deliverables should be built. In the meantime it is the reason your senior engineers are rewriting drafts instead of reviewing them. A response letter that bounces three times between a junior and a PE before it is right is not three small edits. It is a PE spending loaded hours doing work a junior should have produced correctly the first time, on every project, forever. That is the same margin leak as a team leader fixing a drawing himself. It just hides inside a Word document instead of a CAD file.

What We Built at Modern Engineering to Fix It

Not a better prompt. Not a folder of past reports someone liked. A structured library of writing skill files, one for each deliverable type the firm produces, living inside our internal operating platform and connected directly to our AI agents.

How It Actually Works

From engineer request to correctly structured deliverable — automatically

1

Engineer Requests

"Help me draft a response-to-comments letter"

→

2

Agent Scans

Scans the skill library for the matching deliverable type

→

3

Skill Loads

Loads the full skill into context, announces which skill it loaded

→

4

Output Produced

Writes using your firm's structure, tone, and criteria order

The engineer does not select a template from a dropdown. The agent does the routing.

Each Skill File Follows an 11-Section Structure

Purpose

Scope

What Good Looks Like

The Critical Rules

Workflow Checklist

What Failure Looks Like

Trigger Field

Agent Scope

Never Do

How We Know It's Holding

Owner & Escalation

Refinement Is a Single Conversation

When a PE wants to refine how a deliverable is built, the process is a single conversation. They tell the agent what the rule is. The agent drafts the skill file. The PE approves it. It is live for every agent in the system on the next request, no publish step, no cache to clear. Every edit is tracked with version history, author attribution, and conflict detection so nobody silently overwrites someone else's work. Any authenticated staff member can contribute a refinement, not just admins, because the people closest to the work are the ones who know when something needs to change.

Three Approaches Compared

Approach	First Draft Quality	Scales with the Firm	Result
Folder of past reports	Inconsistent	No — static	Still bounces
Better prompts	Improved prose	Partial	Still bounces
Skill file library	Firm standard, every time	Yes — compounds	First draft lands clean

This is the difference between keeping a folder of past reports you happened to like and giving every person in the firm a way to produce a report or memo exactly the way the project in front of them requires. The folder is static. The skill library compounds. Every good edit your best people make becomes the floor for everyone behind them.

The Compounding Advantage

When the way your firm writes lives in a system instead of three people's heads, your newest engineer's first draft starts where your best engineer's used to. That is throughput you do not have to hire for.

The AI is not writing the report. Your engineers still own the judgment, the numbers, and the seal. What changes is that the institutional knowledge of how your firm writes stops being something that walks out the door and starts being something the firm owns, maintains, and improves. Every project makes it better. Every good PE who touches it leaves something behind that the next engineer gets to start from.

"This is what Phase 2 looks like for written work. Not a faster typewriter sitting on top of the same firm. The framework underneath it, owned by the firm, getting sharper every time someone uses it."

Diagnostic Question This Week

Which single deliverable in your firm eats the most review cycles? The one that always bounces, the one your most senior people keep quietly rewriting because it is faster than explaining it again. That is the first thing worth writing down. That is the first skill worth building.

Reply and tell me which deliverable that is for you. I would bet you already know.

If you want to see what a skill file looks like before you build one, reply to this email. I will send you an example from our library.

contact an engineer

Michael Groselle, PE

CEO / Founder, Modern Engineering Solutions

contact an engineer  |  mod-eng.com  |  (214) 833-6748

Modern Engineering Solutions  |  McKinney, TX  |  mod-eng.com

You are receiving this because you opted in to Engineer Your Freedom.  |  Unsubscribe