Compressor World turns 4,090 manuals into a paid AI agent
AskCW.ai answers from manufacturer documents and names the file and page. Buyers get 5 free queries a month.
Industry
Industrial equipment e-commerce
Location
Massachusetts, USA
Client since
2025
Status
Finished
4,090
manuals, spec sheets, FAQs
3
document sources
2
build phases in 2025
5
free queries a month
On the AskCW.ai screen above, a sample question describes a workshop’s job. It lists impact wrenches, a dual-action (DA) sander, a paint spray gun, 4 to 6 hours a day, 230V single-phase power, an indoor install and a budget of about $3,000. AskCW.ai turns that into a requirements list and selection criteria, then gives options from the manufacturer documents, with the file and page under the document facts. When a spec isn’t in the documents, it says so, and price questions go to Compressor World’s sales team.
AskCW.ai at a glance
Client
Compressor World, LLC, an online store for industrial air compressors since 2000, Massachusetts
Product
AskCW.ai, an AI agent on its own website, askcw.ai
Product type
AI assistant over a technical product catalog, with a free tier, a paid plan and sponsored product cards
Documents
4,090 manuals, spec sheets and FAQs from Google Drive, Dropbox and Google Sheets
Engagement
AI product development in 2 phases: from April 14, 2025, then from July 31, 2025
Built with
Next.js, NestJS, PostgreSQL, AWS, LlamaIndex, OpenAI and the Bridge knowledge base
Compressor World and the AI agent it asked for
Compressor World has sold industrial air compressors online since 2000, from Massachusetts. Its senior applications specialists have 30 to 56 years each in compressed air, and buyers reach them by phone, email and live chat. By 2015 the company had built its own offer and counter-offer sales software, Haggle It, and was offering it to other online stores.
In March 2025 Compressor World’s founder, Matt Mazanec, asked Brocoders for a specific AI agent for its business and niche, with the front end and the back end. At the April 8 kick-off, one goal was to promote partner brands through conversational advertising, and phase 2 built it as sponsored product cards. The work was an AI product development project on Compressor World’s own documents.
“We built AskCW.ai: a domain-specific AI assistant that provides accurate, document-based answers to technical queries, 24/7.”
How AskCW.ai answers from 4,090 compressor manuals
AskCW.ai is a retrieval-augmented generation (RAG) assistant: it first finds the passages in Compressor World’s own documents that match the question, then writes the answer from them. More on the method in our list of RAG development companies.
1. Documents stay with the client’s team
Compressor World keeps and updates its manuals, spec sheets and FAQs in Google Drive, Dropbox and Google Sheets. The sheets map product categories and series to their documents.
2. Scheduled ingestion
The pipeline picks up new and changed files through the Bridge knowledge base, Brocoders’ own retrieval platform. It copies them to AWS S3 and indexes PDF, CSV (comma-separated values) and Excel (XLSX) files. By July 2025 it held 4,090 documents.
3. Retrieval
LlamaIndex finds the passages that match the question. The retrieval layer can’t read images directly, so for diagrams the model works from the figure descriptions in the document.
4. The answer
OpenAI GPT-4o writes the answer from those passages. NestJS runs the API, accounts and logs. Next.js runs the chat, and PostgreSQL holds users, logs and query limits.
Answers and names the file and page, for example “Source: up6-411-kw-oilflooded-rotary-screw-compressors.pdf, page 2”
A spec that isn’t in the documents
Says so and points to the sheet to check: “Airflow/CFM: Not found in the documentation; check model-specific spec sheets.”
A price
Sends the buyer to sales: “Pricing isn’t available here. Please contact sales.”
Anything
Shows the note “AskCW.ai may not always be 100% accurate. Please double-check important details or consult a live expert when needed.”
Source lines sit under the document facts. The closing recommendation sums them up and has no source line of its own.
AskCW.ai phase 1, from kick-off to May 2025
Phase 1 built the assistant itself: the document pipeline, the chat, accounts and a minimal admin panel.
April 8, 2025
Kick-off. The team asked for typical buyer questions with the answers Compressor World expected. It also asked for the product data in Excel or CSV, the manuals and spec sheets, the target users (80% technical, 20% non-technical) and the rules for sponsored content.
April 14
Build starts.
April 18
The user interface (UI) design is finished in Figma. OpenAI’s o3 model is connected to the product, and o4-mini is added for comparison.
April 25
Authentication, the chat interface, hosting, secure connections (SSL) and a staging environment are done. The first sample conversations are shared with the client.
May 2
The product is renamed AskCW.ai. The bulk import of the knowledge base is being prepared, and the chat widget design is being adapted for embedding.
AskCW.ai phase 2: a free tier, a paid plan and sponsored product cards
Phase 2 started on July 31, 2025 and built the business model Compressor World asked for.
Plans
A counter tracks free queries. The plan page lists a Free plan with “5 AI assistant queries per month” and a Pro plan with “Unlimited AI assistant access”. Compressor World sets the Pro price on that page.
Sponsored listings
A sponsored product card appears inside the answer, marked “SPONSORED”, below the buyer’s requirements.
Follow-ups and documents
Both plans list follow-up suggestions, for example “Need help sourcing?” and “Talk to an expert”, and access to manuals and PDF datasheets.
What AskCW.ai gives a compressor buyer
The wording below is the client’s own, from askcw.ai and its plan page.
Ask in plain words
Describe the tools, the hours, the power supply, the noise limit and the budget. AskCW.ai turns them into a requirements list before it suggests anything.
Compare options
The selection criteria come first: duty cycle, airflow in cubic feet per minute (CFM) at 90 pounds per square inch (PSI), noise and power. Then come options from the current documentation, each with its source file and page.
Get the document
Access to manuals and PDF datasheets on both plans.
Talk to an expert
Follow-up suggestions that point to Compressor World’s people, and price questions handled by its sales team.
Plans
Free gives 5 AI assistant queries a month. Pro gives unlimited access and faster response processing. The page adds: “No long-term commitments.”
How the AskCW.ai team ran the build
Project manager
1
Front-end engineer
1
Back-end engineers
2
Designer
1
Quality assurance (QA) engineer, phase 2
1
DevOps engineer
1
User stories had acceptance criteria and a minimum, average and maximum estimate, with time added for communication and risk. The plan set a written status report every Friday and sprint demos every 2 weeks, with the work tracked in Jira and Slack.
What Compressor World has in production
The product
AskCW.ai live at askcw.ai, with chat, accounts and plans (checked October 16, 2026). A Free plan with 5 queries a month and a Pro plan. Sponsored product cards inside answers. Follow-up suggestions.
The documents
4,090 manuals, spec sheets and FAQs indexed from 3 sources (July 2025). File and page under document facts, and a “not found” answer when a spec isn’t in the documents. Access to manuals and PDF datasheets.
AskCW.ai technology stack
FAQ: an AI agent over product manuals
What does AskCW.ai do when a spec isn’t in the documents?
AskCW.ai writes its answers from the passages it finds in Compressor World’s documents and names the file and page under the document facts. When a value isn’t in the documents, it says so and points to the spec sheet to check. The chat screen carries a note asking buyers to double-check important details or consult a live expert.
How does AskCW.ai handle price questions?
Price questions go to Compressor World’s team. AskCW.ai replies, “Pricing isn’t available here. Please contact sales.”
Which document formats can it read?
PDF manuals and spec sheets, CSV and XLSX files, and Google Sheets. The retrieval layer can’t read images directly, so for diagrams the model works from the figure descriptions in the document.
How does the knowledge base stay current?
Compressor World keeps its documents in Google Drive, Dropbox and Google Sheets. The ingestion pipeline re-indexes them on a schedule through the Bridge knowledge base, so a changed spec sheet reaches the assistant without a manual upload.
How does the paid plan work?
The Free plan gives 5 AI assistant queries a month. The Pro plan gives unlimited AI assistant access. Compressor World sets the Pro price on its plan page.
Can an assistant like this run inside an existing online store?
AskCW.ai runs as its own website at askcw.ai. In May 2025 the team was also adapting the chat widget design for embedding. For when this kind of assistant fits a catalog, see our guide to AI assistants for technical documentation.
When did the work on AskCW.ai happen?
Phase 1 started on April 14, 2025 and covered the document pipeline, the chat, accounts and a minimal admin panel. Phase 2 started on July 31, 2025 and built the plans, the sponsored product cards and the follow-up suggestions.
If your buyers keep asking questions your manuals already answer, we’ll scope a first phase on one product line: its documents, the questions your experts answer most, and what the assistant should hand to sales.
