Finding a Past Project Means Finding the Person Who Ran It

An engineering consultancy reached out to us recently with a problem we hear constantly. Eleven people, twenty years of projects, about four terabytes of files. When a partner needs to reference a past job — same region, same materials, same problem — the process is: figure out which project manager ran it, email them, wait while they dig through their folders, and hope they attach the right thing.

Everything exists. Reports, analysis data, drawings, proposals — two decades of hard-won engineering knowledge. But it's spread across shared folders and personal machines, named eight different ways, in file formats spanning twenty years of software. In practice, the company's knowledge lives in the memories of whoever's been there longest.

That's not just an inconvenience. When partners think about growth, succession, or a future sale, an archive nobody can search is worth a lot less than the same archive made answerable.

Ask Your Files a Question, Get an Answer With Sources

The setup that fixes this is now well established: an AI index of everything the firm has ever produced, with a chat interface on top. A partner types "find the projects in Denver that used this concrete mix" and gets back the four matching jobs — with links to the actual files — in seconds.

A few things about how this works that surprise people:

The Hard Part Is Your Files, Not the AI

Installing the AI is the easy step. The real work — and the real cost — is in the files themselves. The same project might appear as "Denver Hospital," "Denver Hosp," and "DH-2015." Some data is in modern formats, some in software nobody has run since 2009. And most firms genuinely don't know which of their data matters until someone goes and looks.

The mistake to avoid: trying to clean up and reorganize everything before doing anything. Twenty years of files never gets fully reorganized — that project dies of exhaustion. The right move is to structure only the data people actually query, and leave the rest searchable as-is.

This is also why honest quotes for this kind of project come as a range until someone has inventoried the files. The AI is the same either way; the condition of the archive decides the effort.

Start With the Questions, Not the Technology

The firms that get this right start by writing down the five questions they most wish they could ask their own history. "Find similar past jobs" is almost always first — and it's also one of the easiest to build. That's a good first target: high value, quick to prove, and it builds the team's trust in the system before anything more ambitious.

Sitting on an archive you can't search?

Book a free assessment. We'll look at what you have and tell you honestly what it would take to make it answerable.