How We Build Your AI Company Brain: Our 6-Week Process
You’ve decided your business needs a custom AI knowledge base. Maybe you’ve read about what an AI knowledge base does and recognized the Steve Problem in your own company. Maybe you’ve already felt the cost of losing institutional knowledge when a key employee left and you’re determined to prevent that from happening again.
Either way, you want to know exactly what happens next.
This page walks through our complete build process, phase by phase, with realistic timelines, your team’s time commitment, and what you’ll have at the end. No vague promises. Just a transparent look at how we turn your company’s scattered knowledge into a system every employee can access.
Before We Start: The Discovery Call
The process actually begins before the six weeks start. We schedule a 30-minute discovery call to determine whether a Company Brain is the right fit for your business.
During this call, we ask about your team size, the types of knowledge your employees need most often, where your current documentation lives, and the specific pain points driving your interest. Not every business needs an AI knowledge base, and we’ll tell you if yours doesn’t.
If it’s a good fit, we send a proposal within 48 hours. The proposal includes a clear scope, a fixed price, and a timeline. No hourly billing surprises.
Phase 1: Knowledge Audit (Week 1)
The first phase is about understanding what your business knows and where that knowledge currently lives.
What Happens
We conduct a structured audit of your existing knowledge assets. This means reviewing your document ecosystem: SOPs, training manuals, employee handbooks, vendor agreements, product specs, process checklists, email templates, and anything else your team references.
But written documents are only part of the picture. The most valuable knowledge in most businesses is unwritten. It lives in the heads of your experienced employees. The workarounds, the client preferences, the shortcuts that actually work, the “everyone just knows” information that nobody has bothered to write down.
We identify these knowledge gaps through interviews with your key team members. These are short conversations, typically 30 to 45 minutes each, focused on questions like: “What do new hires ask you most often?” and “What information would be hardest to replace if you left?”
What You’ll See
At the end of Week 1, you receive a Knowledge Map. This document outlines every knowledge area we’ve identified, ranks them by business impact, and flags the gaps between what’s documented and what’s tribal. You’ll review this and tell us if we missed anything.
Your Time Commitment
About 3 to 4 hours total. This includes the initial kickoff meeting and scheduling time for 2 to 3 team members to do their knowledge interviews.
Phase 2: Data Preparation (Week 2)
This is the least glamorous phase and the most important one. The quality of your AI knowledge base is determined entirely by the quality of the information fed into it.
What Happens
We take every document identified in Phase 1 and prepare it for AI processing. This involves cleaning up formatting issues, removing duplicate or outdated content, breaking large documents into logical sections, and structuring information so the AI can retrieve it accurately.
A 47-page employee handbook doesn’t go in as a single block. It gets broken into individual policies and procedures, each tagged with context so the AI understands what it’s looking at. A folder of vendor spec sheets gets processed individually, with each vendor’s information clearly delineated.
For the undocumented tribal knowledge captured during the audit interviews, we create structured documents from the interview notes. Your team reviews these for accuracy before they’re added to the system.
What You’ll See
A content inventory showing exactly what’s going into the system. You review it, approve it, and flag anything that needs correction. This is your last chance to catch errors before the information goes live.
Your Time Commitment
About 2 to 3 hours for document review and approvals. We do the heavy lifting on preparation.
Phase 3: System Build (Weeks 3-4)
This is where the AI knowledge base comes to life.
What Happens
The technical build has several components.
First, your prepared documents get processed into a vector database. In plain terms, the AI reads all your content and creates a mathematical map of the information, understanding not just the words but the relationships between concepts. If the technical side interests you, we explain the underlying RAG technology in our guide to what an internal AI knowledge base is. This architecture is what allows the system to find relevant answers even when someone asks a question using different words than the document uses.
Second, we configure the AI’s retrieval and response system. This involves setting the system’s personality and tone to match your company culture, defining how it handles questions it can’t answer, configuring source citations so employees can verify answers against original documents, and building the user interface your team will actually use.
Third, we integrate the system with the tools your team already uses. For most businesses, this means a web-based interface accessible from any device. Many clients also want Slack or Microsoft Teams integration so employees can ask questions without leaving their workflow.
If you’re wondering how this compares to just using a shared drive or wiki, the difference is significant. Read AI knowledge base vs. Google Drive or wiki for a detailed comparison.
What You’ll See
A working prototype of your Company Brain. You and a small group of designated testers will get access to ask questions and evaluate the responses. This isn’t the final version, but it’s functional enough to see exactly how the system works.
Your Time Commitment
Minimal during the build itself. About 1 hour for a mid-build check-in where we show you the prototype and gather initial feedback.
Phase 4: Accuracy Testing (Week 5)
This phase separates a useful system from a frustrating one. Testing is where the real work of quality happens.
What Happens
We compile a set of 50 to 100 real questions your team actually asks in the course of daily work. These come from the knowledge audit, from new hire questions we collected, and from your managers’ input on what people ask most.
We run every question through the system and evaluate the answers against your documentation and your team’s knowledge. Each answer gets rated on accuracy, completeness, and clarity.
The target is 85% or higher accuracy on the first round. That means 85 out of 100 answers are correct, complete, and clearly stated. The other 15% reveal gaps, ambiguities in the source material, or areas where the system needs tuning.
We address every failed question. Some require adding more source material. Others need adjustments to how the system interprets certain types of questions. A few might reveal that the original documentation was incomplete or contradictory, and those get flagged for your team to resolve.
After adjustments, we run a second testing round. Most systems hit 90% or higher on the second pass.
What You’ll See
A detailed accuracy report showing every test question, the system’s answer, the correct answer, and the accuracy score. You’ll also see exactly what we changed to improve performance.
Your Time Commitment
About 2 to 3 hours. This includes reviewing the accuracy report and working with us to resolve any documentation conflicts or gaps the testing revealed.
Phase 5: Launch and Onboarding (Week 6)
The system is built and tested. Now your team needs to start using it.
What Happens
We launch the system to your full team with a structured onboarding process. This typically includes a live training session (30 to 45 minutes) where your team sees the system in action and practices asking questions. We cover how to interpret answers, how to check source citations, and what to do if they think an answer is wrong.
We also train your designated knowledge manager, the person responsible for submitting new documents and flagging content that needs updating. This role doesn’t require technical skills. It’s more like being a librarian than a programmer.
For the first two weeks after launch, we monitor system usage closely. We track which questions get asked most, where users seem to get stuck, and which answers need improvement. This early data is gold for fine-tuning.
What You’ll See
Your team actively using the system. We provide a post-launch report after two weeks showing usage metrics, common questions, and recommendations for improving the knowledge base.
Your Time Commitment
About 2 hours for the team training session and knowledge manager onboarding.
After Launch: What Monthly Maintenance Covers
An AI knowledge base isn’t a set-it-and-forget-it tool. Your business evolves, your processes change, and new information needs to be incorporated. Monthly maintenance ensures the system stays accurate and useful.
Your monthly fee covers AI hosting and infrastructure, processing new and updated documents as your business changes, quarterly accuracy audits (we re-test with fresh questions), system monitoring and performance optimization, and technical support when needed.
Most clients submit 2 to 5 document updates per month. Some months there’s nothing new. Other months, like when you roll out a new service or change a major process, there’s a batch of updates. The monthly fee covers all of it.
If you’re curious about the cost of losing institutional knowledge and how a Company Brain pays for itself, we’ve done the math in detail on that page.
What It Costs
We believe in transparent pricing. Here’s the realistic range.
The build fee typically runs between $3,000 and $8,000. The factors that influence price are the volume of source documents, the number of knowledge areas covered, and the complexity of integration (basic web interface vs. Slack plus Teams plus mobile).
Monthly maintenance runs between $300 and $800, depending on the size of the knowledge base and how frequently your content changes.
For most Treasure Valley businesses with 10 to 30 employees, the total first-year investment falls between $6,600 and $17,600. That includes the build plus 12 months of maintenance.
Compare that to the cost of even one poorly handled employee transition, one that could run $30,000 or more when you factor in recruiting, training, and lost productivity, and the math becomes straightforward.
Real-World Timeline Example
Here’s what the six weeks looked like for a recent project with a 22-person contractor in the Boise area.
Week 1: Kickoff meeting on Monday. Knowledge audit interviews with the owner, lead project manager, and office manager by Thursday. Knowledge Map delivered Friday.
Week 2: Document preparation. We processed 14 SOPs, an employee handbook, 8 vendor spec sheets, and 3 sets of interview notes capturing tribal knowledge. Content inventory sent for review Wednesday, approved by Friday.
Week 3-4: System build. Mid-build prototype demo on the Wednesday of Week 3. Client feedback incorporated during Week 4.
Week 5: Accuracy testing with 72 real questions. First round: 87% accuracy. Adjustments made. Second round: 93% accuracy.
Week 6: Team training on Tuesday. Full launch Wednesday. Two-week monitoring began immediately.
The owner’s total time investment across all six weeks was about 8 hours. The office manager contributed about 5 hours. Other team members contributed 1 to 2 hours each for interviews and the training session.
What You’ll Need to Provide
The build process is designed to minimize your team’s effort, but there are specific things we need from you to build a system that actually works.
Access to existing documentation. SOPs, manuals, handbooks, spec sheets, templates, and anything else your team references. We’ll tell you exactly which documents we need during the kickoff meeting. Most businesses can gather everything within a few days.
Time from 2 to 4 key employees. Your most experienced people hold the tribal knowledge that makes the system valuable. We need 30 to 45 minutes with each of them for structured interviews. We’ll schedule around their availability.
A designated point of contact. One person who can answer questions, approve content, and coordinate access to documents throughout the six weeks. This is typically the owner, office manager, or operations manager.
Feedback during review checkpoints. We send materials for your review at three points: the Knowledge Map (Week 1), the Content Inventory (Week 2), and the Accuracy Report (Week 5). Timely feedback at each checkpoint keeps the project on schedule.
We don’t need access to your IT systems, proprietary software, or customer data. The knowledge base is built from documents you provide to us directly.
What Happens If It Doesn’t Work
We stand behind the build. If the system doesn’t hit 85% accuracy during Phase 4 testing, we continue working, at no additional cost, until it does. That commitment has never required more than one additional week of refinement.
If after launch you feel the system isn’t delivering value, we’ll work with you to identify why and make adjustments. Our goal is a system your team actually uses every day, not a tool that looks impressive in a demo and then gathers dust.
Ready to see what a Company Brain would look like for your business? Book a discovery call and we’ll walk through your specific situation. The call takes 30 minutes, costs nothing, and you’ll leave with a clear picture of whether this investment makes sense for your company.
FAQ
How much of my team’s time does the build process require?
The total time commitment for your team across all six weeks is approximately 10 to 15 hours, spread across multiple people. The owner or decision maker typically contributes 6 to 8 hours. Key team members contribute 1 to 3 hours each for interviews and review sessions. We designed the process to minimize disruption to your daily operations.
Can we add more knowledge to the system after launch?
Yes, and you should. Your monthly maintenance includes processing new and updated documents. Most clients add content monthly as they create new SOPs, update existing procedures, or document previously unwritten processes. The system grows more valuable over time as it accumulates more of your company’s knowledge.
What if our documentation is a mess right now?
That’s normal. Most businesses we work with don’t have pristine documentation. Part of our Phase 2 work involves cleaning up and structuring existing documents. If critical knowledge is completely undocumented, we capture it during the Phase 1 audit interviews and create structured documents from those conversations. You don’t need perfect documentation to start.
Do you work with businesses outside the Treasure Valley?
Our primary focus is Idaho businesses, and most of our clients are in the Boise, Nampa, and Meridian area. However, the build process works equally well for remote clients. Discovery calls and training sessions happen over video. The only difference is we can’t do in-person knowledge audit interviews, but video calls work just as well for that purpose.
What technology does the system run on?
The system uses a RAG (Retrieval Augmented Generation) architecture with a vector database for document storage and retrieval, paired with a large language model for generating answers. We host on secure cloud infrastructure and can deploy on HIPAA-compliant servers for medical clients. You don’t need to manage any technology. It’s fully hosted and maintained.
How is this different from just using ChatGPT?
ChatGPT has no knowledge of your business. It can’t tell your new hire which form to use for Ada County permit applications or what your markup policy is on custom orders. An AI knowledge base is built exclusively on your company’s documents and trained to answer questions about your specific operations. It cites sources so employees can verify answers. And it never makes up information, if the answer isn’t in your documents, it says so.