I'm Ma Shu. For the past few years I've done one thing: get inside a company's operations, find the single link that should move first, then build it with the team until they run it without me.
I'm not selling software and I'm not here to talk trends. It's judgment — what's worth doing, what comes first, and what counts as done.
The self-assessment tells you three things: where you stand, the path for your industry, what to move first.
Why now
Productivity growth today comes almost entirely from AI. With demand tight and inflation global, the one lever left for cost and efficiency is AI — that's the only road still open.
More concretely: AI is moving into every link of how a business runs. Companies with no AI interface are simply waiting to be displaced. This isn't a technology choice — it's survival.
But do it right
Which segment AI takes, and which segment people keep, is what decides the outcome. Here's how one content-to-lead chain actually gets split:
Break down competitors, build templates, generate differentiated angles at volume
AI-ledCopy, visuals, finished video — produced to spec at scale
AI-ledAccount matrix, one-click distribution, scheduled review
AI-ledAuto-reply, intent detection, capture contact details
AI-ledAuto-reply, scripts adjusted to each customer
AI-assistedAI supplies the customer profile; closing stays with people
Human-ledThe dividing line is judgment: whoever has to decide, carry the risk and read the other side's situation stays human. At the execution layer, AI is a different order of magnitude. At the judgment layer, people are still irreplaceable.
Where to start
First decide which one you're after. The investment, the acceptance criteria and the payback period are completely different.
You already have mature, standard, repeating processes: order entry, contract review, government filings, line-side inspection.
You have a team, but output is capped by headcount and speed: content, demand generation, marketing, customer service.
Core capability is locked in veterans' heads and data sleeps on hard drives: manufacturing, engineering, professional services.
How we work
Think it through, build it, then run it with you until it stands on its own. AI is a management matter, not a one-off purchase. Three stages, contracted separately, each with an exit gate — a first engagement only needs the advisory stage.
Delivered: diagnostic report · scenario priority list · data and knowledge base plan · AI feature menu
Delivered: working system · knowledge base and admin console · API documentation · runbook
Delivered: weekly review cadence · data reviews · role training · iteration releases
Why advisory comes first: buying software alone spins in place; advisory alone doesn't stick. Both sides need time to align on how to work, pin down the real need, and prepare the data and material for AI to land on. None of the three stages can be skipped, and the order can't be flipped.
Track record
Seven projects, all live. The first three you can look at directly — layout, fields and flow are as delivered, with data desensitised. The other four are described only.
Intake, work order, material prep, factory hand-off, QC — five posts threading eleven stages. Which one was waiting on which, and which material had not arrived, used to live in one coordinator’s head. Forms now travel with the stage, overruns escalate on their own, and a skipped step lights up.
Your judgement, your cases and the way you speak become three libraries. A team of digital staff then relays the work to a finished cut. A human appears twice: to approve the topic, and to sign off the final.
A volunteer speaks for 30 seconds on site; a student reflects for 60 seconds afterwards. Both are desensitised before transcription. Voice produces observations only — scoring stays with a human, dimension by dimension, with reasons written down.
Where: custom development and day-to-day maintenance of the core system
Problem: the core system is huge and every new customer needs changes. Understanding and changing it sat with a handful of senior engineers — slow, expensive, and it breaks the day they leave.
Where: collecting, pre-extracting and filing regulatory submissions
Problem: the same information gets typed into many forms, formats keep changing, and one formatting slip can cost a filing window. Most of that information already exists in the system.
Where: first response on direct messages and online enquiries
Problem: enquiry volume is high and whoever is free answers. Real prospects churn while waiting, and staff can't cover the full volume anyway.
Where: assembling bids, qualifications and past-performance records
Problem: qualifications, past performance, parameters and cases had to be reorganised every single time, and rework was high. The know-how stayed with individuals and reset when they left.
Tools built
What a company often lacks isn't a big system but a small tool that sits on the process and opens straight into use. Both below are single-file, zero-dependency and run offline — send it to anyone and they can just open it. (Interface is in Chinese.)
Turns the 8 stages of a prototyping workflow into a feature-level design tool: for each node you define which fields to capture, who fills them, who approves, and what condition lets it through. Export JSON and a process document, and use it directly as a build spec.
Open the demoSplits design, prototyping and delivery into 3 stages and 8 steps. Each step defines only three things: what gets handed over, what has to be checked before it passes, and who reviews what. Configure per task type; overdue items flag amber, then red.
Open the demo10 minutes · free · no contact details needed
15 questions, a check-up across six dimensions. The report comes out immediately: which stage you're at, how the path runs in your industry, roughly where peers sit, and what to move first.
About
Inside a company my role is closer to a consultant who walks the floor — not standing outside talking trends, but sitting down to see where this order actually jams, what that person repeats every day, and whether it's worth touching at all.
And after the judgment, I don't leave. Behind me is an AI transformation SWAT team of consultants and engineers — reading the client's situation, proposing a fitted path, delivering the platform fast, and running it alongside you, the whole way through.
Former overseas executive at Huawei · senior management consultant
I know how large organisations actually run: who sets direction, who carries the risk, where resources flow, and where a process breaks. Which is why I can turn a tangle of business logic into a single line, fast.
You can't fix a process until you can read the organisation.
Founder of a high-tech company · 10 years building businesses
I know the whole operating chain: product design, R&D, after-sales; business model, order growth, management; hiring, systems, incentives. I look for root causes instead of treating symptoms.
Judged from an owner's seat: is this actually worth it?
Long-term advisor to SMEs · has also coached many student teams
I've coached not only companies but many student startup teams. I know exactly where a method fails to land inside a company: which step, which missing person, which sentence left unsaid. And I know how to teach the method so the company grows the capability itself.
Not just answers — a team that can solve it without me.
Let's talk
You don't need a requirements document, and you don't need to know what to buy yet. Just tell me which step eats the most people, runs the slowest, breaks the most often — I can usually tell whether it's worth touching, and where to start.
Add me on WeChat — straight to the point
Usually the same day. If it's a concrete problem, bring numbers.