SERVICES · PROJECT KNOWLEDGE
Find the decision, not the folder
“Where did we agree on that?” It is the most expensive question in a construction company, because answering it costs somebody an hour and failing to answer it costs a rework, a credit, or a dispute you lose.
01 — THE PROBLEM
It exists in six places
The information almost always exists. It is in an email thread from March, a revision of a drawing, a meeting note, a message in a project chat. It exists in six places, none of which is searchable together, and the person who knows which one is on site.
02 — THE BUILD
What gets built
A retrieval system over what your company already produced. No migration, no re-filing, no new place to put documents.
What gets indexed. Project email and attachments, drawings and specifications, contracts and change orders, RFIs and submittals, meeting notes, and the operational records your systems already hold.
How you ask. In plain language: what did we agree about the bathroom tile on the Miller job; which revision has the current electrical layout; what did the client approve in writing about the timeline change.
Every answer cites its source. Document name, page, revision and approval status. If it cannot cite one, it says it does not know rather than producing something plausible. For contracts and change orders that is the entire point — an answer you cannot trace is worse than no answer.
Access follows your structure. Role-based. The field lead sees what the field lead should see, not everything the system happens to have read.
Audit trail. Every query and every answer is logged, so you can reconstruct what the system told whom and when.
03 — DATA
Where the data lives
In your infrastructure, under your retention policy, isolated per project. Nothing is used to train anyone’s model. That is not a footnote — it is why the architecture is built this way.
04 — SCOPE
Timeline and price
A first project-knowledge system is typically three to six weeks and starts at $5,000, depending on how many sources connect and the state of the existing documents.
05 — QUESTIONS
Before you start, you will ask
Do we have to reorganise our files first? No. Reorganising files before indexing them is the reason these projects never start. It indexes the mess.
What if an answer is wrong? You see the citation and check it in one click — which is why citation is mandatory rather than decorative. The system is built to say “not found” instead of guessing.
Who can see what? Whatever you configure. Access control is set before anything is indexed, not after.
Is our data used to train an AI? No. Deployment is client-controlled, and there is no cross-client training.
Tell me where your project information lives now.
Tell me where your project information lives now, and I will tell you what is realistically searchable.