Repetitive document handling
Specifications, drawings, certificates and forms get read by a person and re-typed into another system. It is slow, easy to get wrong, and nobody wants the job.
Alpha builds narrow, reviewable AI workflows for small and mid-sized companies in manufacturing, engineering, metal fabrication and professional services. We start with one clearly defined pilot, keep a person on the final result, and expand only once it has earned that.
None of this is exotic. It is the everyday work that quietly consumes a small team.
Specifications, drawings, certificates and forms get read by a person and re-typed into another system. It is slow, easy to get wrong, and nobody wants the job.
Requests arrive as email threads, PDFs and spreadsheets. Someone has to read each one, pull out quantities, materials and deadlines, and decide who picks it up.
The answer already exists — in a past quote, a project folder, or a colleague’s head. Finding it often takes longer than writing it again from scratch.
Qualifying one company means checking its products, markets and signals across a dozen tabs before anyone writes a single line.
Product data, manuals and outreach have to exist in several languages, and every update means doing the same job again in each one.
Each of these is scoped to a single workflow and delivered with a review step, not as an unattended system.
Field extraction and question answering over your own documents, with the source passage shown next to every answer so a reviewer can check it in seconds.
Read incoming requests, pull out the fields your team actually uses, and hand them to the system you already work in — with the original attached.
Assemble a structured company profile and a first-draft message for a person to review, edit and send. We do not send anything on your behalf.
Keep product data, documentation and outreach consistent across languages, with a review step before anything is published.
If your workflow does not match the list, we scope a proof of concept against a fixed sample and an agreed way of judging the result before we build.
Five steps. The last one is a real decision, not a formality.
We walk through the actual steps, inputs and exceptions — including the ones nobody has written down.
One workflow, a fixed sample, and an agreed definition of “good enough” settled before anything is built.
We build against that sample and measure the result on it, rather than describing what the system should do.
Your team checks the output on their own work. Where it is wrong, we say so, then adjust or stop.
Expanding is a decision you make on evidence at the end, not a commitment you made at the beginning.
These limits are part of the offer, not fine print underneath it.
Our systems draft, extract and rank. People approve. We design for review, not for unattended output.
Some steps stay manual because that is the honest answer for your data and your risk. We will tell you which ones, early.
Work that is still experimental is described as experimental. We do not present a research prototype as a finished product.
Scope, retention and deletion are agreed in writing before we touch a sample. Nothing is reused across clients, and nothing trains a model without your written consent.
Alpha is an AI software initiative by Xi’an Qiuwen Network Technology Co., Ltd. (西安丘闻网络科技有限公司), a China-based technology company.
Alpha is built for small and mid-sized companies internationally — especially in manufacturing, engineering, metal fabrication and professional services, where useful AI work is often unglamorous and very specific.
Our Chinese-language site is www.ciwnw.com. This site gives international visitors and prospective clients the same information in their own language.
Alpha is a product and project name. It is not a registered international trademark, and we do not claim overseas offices or an overseas team.
Tell us the workflow, roughly how often it runs, and what a good result would look like. If it is not a fit, we will say so.
Please do not send production data, personal data or confidential material before a written data agreement is in place.