Use caseConstruction project software

An embedded document AI assistant

Contractors want to ask Brickline's project files a question and get the page back. Two prototypes later, the parts list keeps growing.

Answers cite the document and the pageThe assistant reads only what the user may readThe model changes with a setting, not a release

Actions in the project page
The assistantStairwell doors, level 7

Stairwell doors need a 90-minute fire rating with self-closing hardware. Door 7-S2 also needs panic hardware.

source · spec 08 71 00 p. 12, door schedule rev. D p. 3
Where the answer comes from

Found in the documents · each statement links to its page

01

The starting point

The core idea

4,812 documents sit in one Harrow Build project alone: specifications, drawings, change orders, requests for information. Brickline sells the project management software that holds them, and its customers want an assistant that answers from them. On a roadmap that looks small. Behind it sit search by meaning, a model to choose and host, one contractor's files kept from another, and a record of every answer. That is an infrastructure project, and it is not Brickline's product.

Brickline calls those parts instead of building them. Project documents go into the repository by API with the rights Brickline already manages. The assistant answers through an API, reads only what the asking user may read, and cites its pages. Brickline builds the panel its users see; the pipeline behind it is already there.

Brickline · Riverside TowerWorkspace
A panel in Brickline's design. The assistant lives in the project page, next to the documents it answers from.
Brickline · Assistant prototype v2Workspace
The second attempt, mid-sprint. One line works. Each open line is a system to build, host and run before a customer asks anything.

Three moves carry this page: the documents arrive with their rights, the task is written once in the console, and every answer in Brickline's panel comes back with its pages. Priya makes the first two in a week; Luis meets the third on level 7.

02

The connection

Priya Raman, staff engineer, has a week before the internal demo and three things to connect.

Step 1Intake

Project documents go in with their rights

Brickline's back end sends each project document to the repository through its REST API: the file, its project, its discipline and the groups allowed to read it. Content and metadata are indexed on write.

SpecificationsDrawingsChange ordersRequests for information
Harrow Build · Riverside TowerRepository
Each document carries the groups that may read it, from Brickline's own user model.
Rights are set once, where they already liveBrickline does not build a second permission model for the AI. The groups it manages travel with each document.
Step 2Console

Choose the model by where it runs

In the console, Priya picks the model for each task. Providers are grouped by hosting and flagged by capability. Harrow Build's contracts ask for US-only hosting, so its tasks run on a sovereign model hosted in the US.

Model providersAI console
Each provider shows its hosting, its latency and a compliance chip.
The model is a settingWhen a better model comes out, Priya changes it in the console. Brickline's product is not redeployed.
Step 3Console

Write each business task once

Priya and Daniel Osei, VP of product, write three tasks in the prompt library: answer from the project documents, draft a reply to a request for information, and summarize the change orders. The instructions ask for a citation on every statement, and for a plain sentence when the documents do not hold the answer.

Prompt editor · Answer from project documentsAI console
The instructions are edited in the console, with no release of Brickline's product.
Picture next spring: the assistant gets an answer wrong. Who on your team can change its instructions that afternoon, and does it take a release?
Product people own the wordingUsers click a named task. Nobody on the site composes a prompt, and the product team tunes it without a sprint.
03

The first answer

The panel is in Brickline's project page. Luis Ortega, project manager at Harrow Build, takes it onto the site.

Step 4Assistant

The question goes out with the user who asks it

Brickline's back end calls the assistant API with the question, the project and the user. Permissions apply before the call: the assistant reads only the documents that user is entitled to see, and searches them by meaning.

Assistant requestDeveloper portal
The user is on every call, so the rights are checked every time.
One contractor never reads another's filesThe search runs inside the asking user's rights. A document outside them is never read, so it can never be quoted.
Step 5Viewer

On level 7, the citation opens the page

Luis is on level 7 with the door installer. He asks a follow-up in the panel on his tablet and taps the first citation. The specification opens in the viewer on page 12, inside Brickline.

Spec 08 71 00, door hardwareViewer
Every answer points to the document and page it comes from, with its version.
The check takes one tapLuis reads the clause before he tells the installer anything. The answer is a pointer to the source.
Step 6Assistant

A reply to the subcontractor, drafted and corrected

A request for information arrives from the drywall subcontractor about the wall rating around stairwell B. Luis clicks Draft an RFI reply. The draft highlights the values it took from the documents. He changes the partition type, adds the site meeting date and sends it under his name.

RFI-0412Workspace
The values injected from the documents are highlighted, so the person checks them first.
A named person answers the subcontractorThe assistant writes a draft. The reply leaves under Luis's name, and he stays responsible for it.
04

The rollout

The demo goes well. Brickline opens the assistant to customers, then keeps an eye on what it costs and what it said.

Step 7Permissions

The customer admin closes the commercial folders

At Harrow Build, Keisha Brooks manages Brickline for the whole company. She keeps specifications, drawings and requests for information open to project managers, and closes bids, subcontract prices and claims to every group but management.

Specifications: openDrawings: openBids and prices: management onlyClaims: management only
Harrow Build · Folder rightsWorkspace
The same rights decide what a user opens and what the assistant reads for them.
One rule covers the file and the answerKeisha sets no separate AI scope. A folder a project manager cannot open never shows up in their answers.
Step 8Reporting

The product team watches consumption by task

Every Monday Daniel reads the console dashboard: requests, conversations and tokens across every surface, the model mix, and consumption by task. Brickline uses it to size its plans and to see which task its users actually rely on.

DashboardAI console
Consumption is tracked by use, so the cost of each task is visible.
Cost is read per taskDaniel sees what each feature consumes before he prices it. Figures on the dashboard are sample values.
Step 9History

A disputed answer is checked months later

In March, Harrow Build asks why an answer from November quoted door schedule revision D, when revision E now exists. Priya opens the exchange: the question, the answer, the provider, the model, the date and the cited source are all kept. In November, revision D was the current one. The check takes minutes.

Exchange history · Luis OrtegaAI console
Every model call leaves a dated, attributed entry.
Customer questions are answered from the recordWho asked what, which model answered, from which page: the exchange record holds all three.
05

The verdicts

Every question to the assistant ends in one of three places. Each one tells the user what to do next.

Cited answer

Found in the documents

Each statement links to the document, the page and the revision it comes from.

What follows: the user opens the page and checks it before acting.
Not in the documents

Nothing relevant found

The task's instructions ask for a plain sentence, with no citation and no guess.

What follows: the user asks the architect or files a request for information.
Outside the rights

A folder the user cannot open

Permissions apply before the call, so the document is never read for this user.

What follows: the answer is built from what the user may see, or not at all.
Where the AI stops

It sends nothing to a subcontractor. It grants nobody access. It changes no document. Across the nine steps on this page, every message that leaves Brickline is sent by a named person, who read the draft first.

The assistant searches, answers with sources and drafts. The rights are Brickline's and its customers', and the AI works inside them.

06

What it is worth

Three counts from Brickline's first customer, Harrow Build. A product with other documents and other customers will tally its own.

0 permission modelsbuilt for the AI. The groups Brickline already manages travel with each document.
1 settingto change the model behind a task, with no redeploy of Brickline's product.
3 taskswritten once in the prompt library and used wherever the panel appears.

Where these numbers come from. They are counted on the scenario described on this page, with example data. They are not a customer result.

Why these numbers hold

Answers you can checkTrustEvery statement points to its document and page. The user reads the source in one tap.
Rights before the callSecurityThe assistant reads only what the asking user is entitled to see.
Governed in one placeControlModels, tasks and every exchange sit in the console, with no release to change them.
07

Who it is for

A document assistant ships inside a product only when the product lead, the engineers and each customer admin are on board. Their reasons differ:

Product leadership

Heads of product

The assistant on your roadmap keeps turning into an infrastructure project.

  • The panel is yours, in your design.
  • Tasks are named business actions, with no prompt to compose.
  • Instructions change in the console, without a release.
  • Consumption is visible per task, before you price it.
CIO and architecture

CTOs and staff engineers

Search, models, rights and logs: four systems to build and run, none of them your product.

  • Documented REST APIs and SDKs.
  • Documents indexed on write, searched by meaning.
  • Models chosen by hosting, several side by side.
  • Your content is never used to train a model.
Your customers

Customer admins

You are asked to trust an AI with every folder your company owns.

  • The assistant reads only what each user may read.
  • Closing a folder closes it to the answers too.
  • Every exchange is kept with its source.
  • US-only hosting when your contracts ask for it.

A good first step

With one task and one customer. Take the question your users ask support most often, and answer it from one project's documents in a test environment.

Then put the panel in front of an internal team before any customer sees it, and read the exchanges together.

Next step

See it inside your own product

Bring a sample of one customer's documents and the questions their users ask. We will get your team to a first cited answer in your own screen.

The screens on this page are drawn to show one step at a time, with example data. Brickline, Harrow Build and the people named are fictional.

Uxopian Software gives software companies the content and AI building blocks behind their products, through APIs, without giving up control of their data or their infrastructure.