Building AI Infrastructure for an Engineering Firm: The Complete Data Architecture

Why isn't AI changing anything at most engineering firms?

Because the tools sit on disconnected data. Everyone in our industry is asking a version of the same question: we bought the AI tools, so why isn't anything changing?

At MES we hear this from recruits, from other firms and from clients. How do you actually use AI to cut admin work for your engineers?

My honest answer starts with something most people skip: data infrastructure. AI does not create intelligence. It amplifies whatever you hand it.

Hand it five department systems that never talk to each other, and you get faster silos. Hand it one connected data layer, and you get a firm where each piece of work makes the next one easier.

We drew a picture to explain this inside MES, our AI operations map. This post walks through how it works and the uses that changed how we operate.

What are the five departments every engineering firm runs?

Operations, accounting, HR, sales and marketing, at any size.

• Operations is the heart of the business: engineers, EITs and drafters delivering projects. Most billable work lives here.
• Accounting handles invoices, payments and financial reporting.
• HR runs timesheets, payroll, hiring and onboarding. It also holds interview transcripts and candidate data that most firms never think to use.
• Sales builds proposals, signs deals and tracks client relationships. In a small firm that is often the principal.
• Marketing supports proposals, creates content and builds the visibility that brings in clients and recruits.

In most firms each group runs its own system. Sales has the CRM, operations has project files, HR has timesheets, accounting has invoices, and marketing often has no system at all. None of them talk without a person copying information across.

That manual copying is a cost nobody has named yet. You pay it in late invoices, thin proposals and generic content.

How does AI move through connected data?

Think of your data as a road network. AI travels the roads to find what it needs. If the roads exist and are mapped, it can get anywhere. If they don't, the AI is driving blind, however capable the model is.

Our map shows the five departments connected through one central hub: the client. Each department has its own branches, such as projects, invoices, timesheets, deals, content and hiring, and every link between them is a road the AI can travel.

Built right, information created in one department becomes useful in every other one, without anyone sending an email. I walk through it in our video on the AI operations map.

What does connected data let AI do?

Invoice descriptions written from what actually happened

This is the one that surprises people most. At month end, someone sits down to write invoice descriptions that justify what was billed on, say, Project 35. Usually they work from memory.

With connected data, the AI doesn't need memory. Project 35's number is tagged to every email, Teams message, meeting transcript and timesheet entry. The AI reads all of it, summarizes the month and drafts descriptions in the billing style you already use.

You will be surprised what it finds from the first days of the month that nobody wrote down. It was in the emails, the transcripts and the timesheets. You can also run it on a monthly trigger for every active project.

Content built from real project data

Most engineering firm content sounds generic because it is. Someone who was not on the project wrote it from a two-paragraph summary.

Say we want a post or a video about lift station design. We search our projects tagged lift station, and the AI pulls the calls, transcripts, field notes and reports behind them. Now the draft can describe the real challenges, the approach and the outcome.

The AI is not inventing case studies. It is summarizing what our engineers said in meetings and what our files show. People review what we publish, and a client is named only with the client's consent.

Proposals written in the client's own words

This is the use with the most direct effect on winning work. A prospect creates data as they move through sales: scoping calls, email threads, notes about what they care about and what worries them. It usually ends up scattered.

Connect the CRM to proposal drafting, and the AI can write in the language the client used when they described what they wanted. A draft that covers the scope items they named on the call, uses their terms and answers the concerns they raised.

You give it past proposals as a format reference. It reads the notes, the transcripts and the sample, and produces a first draft. Your proposal writer refines it instead of starting from a blank page. The result reads like it was written for them, because it was.

Can hiring interviews double as market research?

Yes, and it is not in most firms' playbooks yet. When you record hiring interviews and keep the transcripts, you build a record of what candidates actually care about.

In the interviews we run this year, AI comes up almost every time. What tools do you use? How is the firm changing? What does that mean for my day? With transcripts connected to the content system, the AI can list the questions asked most often over the last six months. Then you write recruiting content that answers them, without a survey.

How did we build it at MES?

We mapped every connection by hand before writing a line of code. Which department needs what from which other one. Where data is created and where it has to go. Where information was dying in a manual handoff. That map became the blueprint.

Then we built one application with Claude Code, covering HR, time tracking, KPIs, content, report writing and the CRM, with AI agent workflows running along the connections. We had a working first version fast, and we change it when we need to instead of waiting on a vendor. No platform sprawl: one operating system, one data layer.

Map the connections, build the roads, then put the AI on them. Most firms do it in reverse and wonder why nothing sticks.

For us the point is less admin, more engineering. Every hour the system saves on invoices and paperwork is an hour an engineer can spend on design and on getting clients results.

This is how we did it at MES, not a product recommendation. Your tools, client contracts and confidentiality duties will shape what you can connect. Get consent before you record calls or interviews, and keep a person reviewing anything that goes to a client.

Frequently asked questions

What is AI infrastructure for an engineering firm?

The connected data layer that links operations, accounting, HR, sales and marketing, so AI can find and act on information across all of them. It is what makes AI useful beyond one-off prompts.

Do I need a custom-built system to do this?

Not necessarily. The key is structured, connected data. Some firms get there by integrating the CRM, project and accounting tools they already have. Others build a custom application, as we did with Claude Code. The right path depends on your firm's size, budget and how much control you want.

Can a small engineering firm do this?

Yes, and small firms often gain more, because admin overhead per engineer is higher on a lean team. The five-department model applies at any size.

I run a small civil firm and our project notes live in email, Teams and timesheets. Where should I start if I want AI to draft our monthly invoice descriptions?

Start by putting the project number on every email thread, meeting and timesheet entry. Once the number ties them together, AI can read a month of activity per project and draft descriptions for your billing person to check.

I'm a principal at an engineering firm and we already pay for a CRM and an accounting system. How do I tell whether connecting them is worth it before building anything new?

Map where people copy information between those systems by hand, and how often. Each manual handoff is a road you could build. If the map shows many, connect those first.

What tools did MES use?

Claude Code, to build a custom application covering the CRM, HR, timesheets, KPI tracking, content and report writing, with open-source packages from GitHub for many of the building blocks.

Related resources

• The Four Layers Under MES, Top Down
• How I Built a Civil Engineering Firm That Doesn’t Follow Industry Norms
• You need more engineers and you need AI
• Revolutionizing Infrastructure Management with AI | EP 10

Want an engineering team with less admin in the way?

We built this so our engineers spend their time on your project, not on paperwork. Talk to us about your next water, wastewater or site civil project.

We specialize in:

• Water engineering
• Wastewater engineering
• Site civil engineering for land development
• Support for engineering firms
• Construction administration

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