Copilot Is Working. But Is It Just an Expensive Google Search?

Your team got Copilot and it works. Someone asks a question. Copilot finds an answer. Fast. Saves time.

But Copilot might not be working as well as it should.

Here’s what happens: You’ve got standards written. Good ones. But they don’t get updated or improved. Once people knew the system, they stopped feeling like they needed to maintain them. If it isn’t broken, why keep working on it? New employees know the standards exist, but they don’t use them either. They just follow what the seniors do. The boss shortcuts the checklist too. And nobody questions it because there aren’t any metrics showing the cost.

Checklists are there, but people skip steps because of deadlines. The guy who’s been here ten years knows the process. The new hire mimics him. The boss does it faster by cutting corners. Nobody writes any of it down.

Decisions get talked about in meetings but never get documented. Lessons learned live in someone’s head. Continuity is whatever the boss or project owner remembers. File names are inconsistent. Metadata either doesn’t exist or was abandoned years ago.

And here’s the real question: Is “not broken” actually effective? Or are you just coasting?

Is one person consistently faster at a task than another? Does anyone actually know why?

If you can’t explain why John pulls together a foundation analysis in three days while someone else takes five, then that difference is knowledge. Process. Judgment. And it’s locked in John’s head.

That’s where Copilot breaks down.

Why This Matters

Copilot needs to see connections between information. It needs to know that a decision made on one project relates to a soil condition from another. That a problem you solved three years ago might apply to what someone’s working on today.

But Copilot can only do that if the information is organized in a way it can read. Right now, it might not be.

Your foundation spec is in one place. The soil reports are somewhere else. Why you chose that system is in an email. What actually happened on the project is in someone’s head. Copilot can find the spec. That’s step one.

But it can’t connect it to the decisions. The lessons learned. The reasoning. Because those things are scattered.

The Cost of Knowledge Staying in Heads

Look at the numbers. About 42% of institutional knowledge in AEC firms is unique to individual people. It’s never written down. Never filed. Never searchable.

When that person leaves, it’s gone.

This isn’t abstract. When experienced staff walk out the door, firms lose 5-15% of their project costs to rework. That’s $15 billion annually across the U.S. construction industry. And 45% of firms experience significant delays.

Most firms have one or two senior principals who hold the entire history. Not in documents. In their heads. Which subconsultants show up on time. What the real scope was on past jobs. Why decisions got made. What went wrong and how they fixed it.

When that person retires, you can’t get it back. No software recovers it. No search finds it.

Right now, your standards are written but not followed. Your checklists exist but get shortcut. Your good people work fast because they know things nobody else does. And none of it is documented.

That means Copilot sees scattered files. A spec here. A report there. An email chain somewhere else. No connections. No reasoning. No context.

Copilot can’t see what your team knows.

What Firms Are Actually Doing About This

Kimley-Horn is a major engineering firm. They handle complex infrastructure projects. Data lives everywhere: CAD files, scheduling tools, financial systems, email, Teams.

A senior engineer there said this: “The biggest challenge in complex projects isn’t finding information. It’s connecting the right information at the right time. We need to see connections between design decisions, budget constraints, and scheduling requirements that might otherwise remain hidden.”

That’s the gap they’re solving. When an engineer asks “What were the key decisions from last week’s structural review?” Copilot doesn’t just find something. It pulls meeting transcripts, emails, project documents, and connects them. The engineer gets the full picture. Not a spec. Context.

That only works because they’ve organized their knowledge in a way that Copilot can read it.

Your firm might be using Copilot to find the spec. Get the information. Done. Step one.

But here’s what’s possible: An architect asks for foundation details from a similar project. Copilot finds the system that was used. Why it was chosen. What code applied then versus now. What problems came up. What worked. What the lessons were.

That answer saves a full day. Maybe more.

Right now it’s not happening because the information exists, but Copilot can’t connect the pieces. Your standards aren’t being followed consistently. Your decisions aren’t documented. Your lessons learned live in someone’s head.

Copilot is doing exactly what it can with what’s available.

The Knowledge Crisis

7% of AEC firms miss deadlines. Knowledge loss is a big reason why.

When a senior engineer retires, you don’t just lose a person. You lose decision-making patterns. You lose why decisions got made the way they did. You lose the ability to avoid mistakes the firm already learned not to make.

41% of your workforce will retire by 2031. 90% of what they know was never documented.

Your knowledge lives in three places:

• Drawings and specs (sometimes labeled, sometimes not)
• Emails, RFI logs, change orders (scattered)
• People’s heads (why decisions were made, what worked, what didn’t)

Copilot can read specs. It can read email. It can’t read people’s minds.

The problem is that your knowledge is being held hostage by whoever happens to remember it. New engineers have to learn it all again. Copilot can’t access it.

When you make the knowledge searchable and documented, Copilot becomes useful. And new engineers can actually learn from what your firm already knows instead of starting over.

What Needs to Change

This isn’t complicated. But it requires discipline. And it has to start at the top.

Your standards need to be enforced. Not just written. Actually used. When shortcuts happen without consequences, they become accepted. That means measuring the cost. Rework. Schedule slips. Quality issues. When you can see “we skipped this step and lost $50K,” behavior changes.

Your decisions need to be documented. Not in email chains. In a document of record. Why did you choose this system? What were the alternatives? What constraints were you operating under?

Your lessons learned need to be captured when the project ends. Not stored in someone’s head. Written down. Searchable.

Your fast people. The ones who work twice as quickly as everyone else. Their knowledge needs to get transferred. Ask them why they do things the way they do. Document the process. Let new people learn it. Let Copilot read it.

Your file structure needs consistency. Project names. Locations. Building types. Document types. Make it standard. Make it searchable.

This isn’t about AI. It’s about whether your firm can actually learn from itself. And whether the people running the firm actually want it to.

The Actual ROI

Right now, your firm loses 5-15% of every project to rework. That’s because knowledge isn’t shared. People remake mistakes that were solved three years ago.

41% of your workforce will retire by 2031. 90% of what they know will leave with them if it’s not documented.

WSP Global deployed Copilot to 10,000 people. 84% report saving time. Not because Copilot is magic. Because they organized their knowledge first.

Your firm could be there. But only if the knowledge system actually works.

The investment is small compared to the cost. A few decisions about how to label files. Discipline to stick with standards. Time to document decisions and lessons learned.

The payoff is serious. You stop repeating mistakes. New people learn faster. Copilot becomes useful because it can actually see what your firm knows.

And when that senior principal retires, you still have their knowledge.

Next Step

Ask yourself these questions:

Are your standards actually being followed, or just being done the way the seniors do it?

Does anyone know why one person is faster at a task than another?

When a senior person retires, what knowledge walks out the door?

Are your lessons learned documented or just talked about?

Are you measuring what these shortcuts actually cost you?

If you can’t answer those questions clearly, your knowledge system isn’t working. And Copilot won’t work either because it’s reading the same scattered files your team is.

If this sounds like your firm, let’s talk about what your actual knowledge system looks like. Not hypothetically. What are people really doing?

The cost of inaction is real. The firms that organize their knowledge first win. Take 10 minutes with the AI Readiness Assessment at https://evergx.com/ai-ready-scorecard/ and see exactly where your system stands.