The Four Layers Under MES, Top Down

Why does most AI at engineering firms sound like a chatbot?

Because the firm skipped the foundation. Most firms I talk to start with AI agents and bolt them onto the software they already have. The agent can search a few documents and summarize an email thread. It cannot tell you whether a project is on track, because nobody ever told it what "on track" means for this firm.

That is why their AI sounds like a chatbot pretending to be the firm, not the firm answering a question. In an earlier piece I wrote that AI is architecture, not a feature. Several readers wrote back with the next question: what is actually inside, and how do the layers connect? Here they are, top down.

What is the operating system layer?

It is the firm's brain in written, structured form. It holds what we believe, how we work and what we measure ourselves against. Ours is organized in five pillars:

1. First principles: vision, mission and values, the foundation everything else is read through.
2. Strategy: our ideal client profiles, revenue strategy and quarterly plans.
3. Standards: the load-bearing rules for sales, delivery, brand, HR, finance and IT.
4. Execution: playbooks, decision rights and cadence, meaning how work actually happens.
5. Targets: KPIs, revenue goals, staffing ratios and bonus tiers.

Every other layer is interpreted through this one. A project's gross margin is just a number until it is read against the target margin the firm wrote down. Then it means something. The operating system tells the AI what should be, not just what is.

What does the operational spine hold?

It holds the firm's own records: projects, people, clients, money and activity. Everything else we track hangs off those five.

Time entries tie to a project and the person who logged them. Payroll comes from those hours and the HR rules for each person. Bonuses come from the same records, read against the bonus tiers in the operating system. Milestones and tasks hang off projects. Meeting notes, messages and call summaries land in one activity history.

The result: one question can pull together a project's hours, its profitability, the engineer who did the work and every email that engineer sent the client. No exports, no stitching spreadsheets together.

Because the spine is ours, the software stack collapses. We built our own time tracking and HR tools on top of it. Our CRM, project management and nurture email run there too. We shut HubSpot down in August 2026. The firm holds its own data instead of paying half a dozen vendors to hold pieces of it.

What sits at the external edge?

The systems we should not or cannot replace. Some data has to live where it lives. Our edge has four parts:

• Accounting, for invoices, estimates and payments. We read it; we do not copy it.
• Our Microsoft 365 account, for sign-in, files, email, Teams and calendars, read and write.
• Web analytics, search data and YouTube, for marketing and lead signals.
• Prospect research tools that tell us about companies before we call them.

Anything regulated stays in its own system. Anything else we bring home. The rule is simple: own what we can own, and sync what we must sync.

What do the agents actually do?

They read the three layers below them, and they carry no knowledge of their own. We run one main assistant and a set of department agents. None of them has the firm hardcoded into it. Every fact in every answer comes from looking something up.

Each agent has tools for the operating system, tools for the spine, tools that reach the edge, and tools for its own department. Every agent must check the operating system before it answers an operational or strategic question. Every claim it makes cites the standard it came from, with a link straight back to the source.

When the marketing agent drafts a LinkedIn post, it pulls our brand voice, the current client profile, recent project outcomes and the last week of relevant activity in one pass.

When the finance agent runs a quarterly profitability review, it reads hours, project budgets, payroll and bonus rules together. Then it tells me which projects went sideways and exactly why.

That is what I mean by less admin, more engineering. The agents do the gathering. Our engineers and I spend the time on judgment.

Why does the shape matter more than the tools?

Because the tools change every year and the shape does not. Operating system on top defines truth. The spine in the middle holds reality. The edge syncs the outside world. Agents read all three.

Most firms invert it. They buy the agents first and hope the foundation shows up later. It does not.

Almost every firm I have looked at has a spine of some kind, no operating system at all, and a few agents bolted onto rented software. If that sounds like you, the missing layer is not the AI.

Limits: this is how we run our own firm, written as of 2026. It is a pattern, not a product, and the tools under each layer will keep changing.

Frequently asked questions

What is an operating system layer for an engineering firm?

It is the firm's vision, values, client profiles, strategy, department standards, playbooks, decision rights and targets, written down in a structured form software can read. Every agent reads it before answering an operational question. It separates an AI that knows the firm from one that is pretending.

Why did MES shut down HubSpot?

Because our operational records now live in our own system. CRM, project management, time tracking, HR and nurture email all sit in the same place the agents read from. Once the spine is built, paying several vendors to hold disconnected pieces of it stops making sense.

What is the difference between the spine and the edge?

The spine is records the firm owns outright: projects, people, clients, money and activity. The edge is regulated or outside data that stays in its own system, like accounting, email and analytics. Own what you can own, sync what you must sync.

I own a 15-person civil firm in Texas and we just bought an AI tool that nobody trusts. Where should I start fixing it?

Start with the operating system, not the tool. Write down your target margins, your standards and who decides what. Until those exist, the AI has nothing to judge your data against, so its answers stay generic.

I'm a principal at an engineering firm with good project records but no written standards. Can we add agents now and write the standards later?

You can, but you will get the chatbot problem. The spine tells an agent what happened. Only the operating system tells it whether that was good. Writing it down costs discipline, not software, so do it first.

Related resources

• Your AI Already Knows More About Your Business Than You Think It Should
• How I Built a Civil Engineering Firm That Doesn't Follow Industry Norms
• You need more engineers and you need AI
• Why Speed Is a Design Constraint: How Modern Engineering Firms Deliver Infrastructure Faster

Want an engineering partner that spends its hours on engineering?

We built these layers to cut our own admin, so more of every project's hours go to the design work you hired us for.

We specialize in:

• Water engineering
• Wastewater engineering
• Site civil engineering for land development
• Construction administration
• Specialty scope and overflow production for engineering firms

Modern Engineering Solutions, McKinney, Texas. Contact: (214) 833-6748 or mod-eng.com