W1
Week One Labs
Knowledge AI for Business

An AI assistant that actually knows your business

Your docs, policies, catalogue, and past tickets already hold the answers your team repeats all day. I turn them into an assistant that gives the right answer, shows exactly where it found it, admits when it does not know, and brings in a person for the rest. Fixed $12,000. Live in 14 days.

Book a free scoping call →See the UN case study
✓ Every answer shows its source✓ Admits when it does not know✓ Tested on your real questions before launch✓ Fixed price. Your accounts. Your code.
The Knowledge AI Pilot in one paragraph

One job, one sprint. Take the questions your business answers over and over (customers at 2am, staff digging through folders, everyone reading the same 40 documents) and hand them to an assistant that answers from your own files, shows its sources, and passes the hard ones to a person. Built by the engineer behind BBC election results pages used by millions and UN internal tools that save $30,000 a month. Before launch, you see it tested on your real questions. $12,000, fixed. Live in 14 days.

You probably tried a chatbot tool already

Most businesses reach for an off-the-shelf AI chatbot first, and the reviews tell the story: answers invented from outside the documentation, fine on simple questions but lost on anything specific, per-resolution pricing that doubles in a busy month, and setup that quietly requires a developer anyway. The tools are not bad. They just do not know your business, and they answer anyway.

The assistant I build does the opposite: it only answers from your own content, gets tested on your real questions before launch, costs the same every month, and is handed to you as code you own.

What ships in the pilot

  • Your files, wikis, and databases loaded in and kept in sync
  • Answers pulled from your content, with the source shown every time
  • A chat window your team or customers actually use, in your brand
  • It admits when it does not know, instead of making something up
  • Hard or sensitive questions get passed to a person automatically
  • A test report: how it scored on real questions from your business
  • Everything runs in your accounts. Code, docs, and logins are yours

14 days, three phases

01

Pick the job, write the test

We pick one job for the assistant, I go through the documents behind it, and we collect real questions from your business to test it against later. You approve the plan before any code.

02

Build, days 3 to 10

I load your content, build the assistant, wire up the sources-shown answers and the pass-to-a-person path. You see a working version mid-sprint, so day 14 is never a surprise.

03

Test, go live, hand over

The assistant is scored on your real questions, goes live in your accounts, and your team gets a walkthrough and the docs. You own all of it.

Transparent pricing

Published upfront because you should not need a sales call to learn a price.

AI Readiness Sprint
$1,500 – $2,500

One week. I check whether your documents can support a good assistant and you get an honest yes or no, plus a fixed plan. The fee counts toward the pilot.

Knowledge AI Pilot
$12,000 fixed

The core offer. A working assistant that answers from your own documents, tested on your real questions and live in your accounts in 14 days.

Expansion Systems
$8,000 – $20,000

The next job: sorting leads, drafting quotes, searching files, processing documents. Fixed price per system.

Care Plan
from $750/month

I keep it accurate: re-tested when AI models change, new documents synced in, small improvements monthly. Cancel anytime.

Proof, not promises

United Nations
Internal tools saving $30,000/month

Custom admin tooling for UN South Asia offices, built in a single sprint. Read the case study →

ApplicoAI
LLM document pipeline to first paying client in 3 weeks

Resume parsing and scoring over messy real-world documents, demoed on Day 14. Read the case study →

Frequently asked questions

How much does custom RAG development cost?+
Market rates for custom RAG development run $10,000 to $75,000 depending on scope and who builds it, with agencies clustering at $40,000+. The Week One Labs Knowledge AI Pilot is a fixed $12,000: retrieval over your documents, a production chat interface with citations, refusal and human-handoff behavior, and an evaluation report against your own real questions, live in 14 days. Expansion systems after the pilot run $8,000 to $20,000 fixed depending on integrations.
Why not just use Chatbase, Botpress, or Intercom Fin?+
For simple FAQ deflection, those tools can be enough, and I will tell you so on the scoping call. The documented failure modes show up when questions get specific to your business: answers invented from outside your documentation, resolution rates well below marketing claims, per-resolution billing that spikes with volume, and configuration that ends up needing a developer anyway. A custom system is grounded only in your knowledge, refuses what it cannot support with a source, escalates to a human, and runs at flat, predictable cost.
How do you stop the AI from making things up?+
Three mechanisms, all verifiable. Retrieval grounding: the system answers only from passages actually retrieved from your documents, and cites them. Refusal behavior: when retrieval finds nothing relevant, the system says so and hands off to a human instead of guessing. Evaluation: before launch we build a test set from your real questions and measure grounded-answer rates on it, and you get that report. Any change after launch is measured against the same set.
What counts as "our knowledge"?+
Anything your business already has in written or structured form: policy documents, product catalogues, FAQs, support ticket history, contracts, SOPs, manuals, internal wikis, spreadsheets, and databases. Typical pilots cover 50 to 5,000 documents. If your knowledge is mostly in a few key people's heads, the scoping call includes how to capture enough of it for the system to be useful.
Where does our data live, and who can see it?+
The system deploys into your own cloud accounts, and your documents stay in your infrastructure. LLM calls go to the provider you approve (Anthropic, OpenAI, or an open model hosted in your account when requirements demand it) under their business terms, with no training on your data. A mutual NDA before scoping is standard, and you own every credential from day one.
What happens in the 14 days?+
Days 1 to 2: scope freeze, document audit, and the evaluation question set drawn from your real workload. Days 3 to 10: ingestion, retrieval, the assistant, and its refusal and handoff behavior, with a mid-sprint checkpoint. Days 11 to 14: evaluation against the test set, fixes, deployment into your infrastructure, and a working handoff session with your team.
Do we own the system?+
Fully. Code, prompts, evaluation sets, infrastructure configuration, and documentation are yours: clean repo, deployment runbook, and onboarding notes. There is no Week One Labs platform dependency and no lock-in. The optional care plan exists for teams that want monitoring and updates handled, not because the system needs me to run.
What if the accuracy is not good enough for our use case?+
That is what the readiness sprint is for. Before you commit $12,000, a short fixed-fee scoping engagement audits your documents and runs a feasibility check on a sample of your real questions. If your use case is not a fit, you find out for $1,500 to $2,500 instead of after a full build, and the fee credits toward the pilot if you proceed.

Stop answering the same questions by hand

A 30-minute scoping call tells you whether your knowledge is ready, what the pilot would cover, and exactly what it costs. No deck, no discovery phase, no surprise quote.

Book your scoping call →
Free weekly newsletter

I know which AI tools are worth your time.

I build with AI every single day. I will send you what actually works, what is overhyped, and what you should be paying attention to next. No fluff, just signal.

Delivered every weekUnsubscribe anytime

Get the AI signal. Drop your email below.

No spam. Just useful AI intel for builders.

Related: AI agent development cost in 2026 · RAG vs fine-tuning · AI app development company · UN case study

← Home