How do you choose an AI consulting company in Shanghai?
Judge the company by the diagnostic, not the pitch. A credible AI consultancy should be able to explain what it will inspect, what it will deliver, who owns the work, how success will be measured, and what it will not promise. Ask for a scoped first stage, a written acceptance process, and evidence from comparable work.
1. Check how they define the problem
A vendor that starts with a tool or a model is selling a product. A consultancy should start with the business goal, the current workflow, the people involved, and the bottleneck. Ask them to describe a recent diagnosis in concrete terms: what they inspected, what they found, and what they recommended not doing.
2. Check the first-stage scope
A credible first stage is bounded: a diagnostic, a pilot, or a specific deliverable within a defined time. Be cautious with engagements that begin with a large transformation program and no measurable first milestone.
Ask what will be delivered at the end of stage one, who will review it, and what would cause the engagement to stop or change direction.
3. Check the evidence
Ask for anonymized cases with context: the starting point, the work performed, the metric changed, and the limits of the result. A case that only lists a logo and a percentage is not enough. A credible vendor should be able to explain what did not work as well as what did.
For enterprise work, ask whether the case was delivered by the people who will work on your project, or by a separate sales team.
4. Check the delivery roles
Clarify who diagnoses, who designs, who executes, and who reviews. A senior person in the pitch who disappears during delivery is a risk. Ask for named roles, decision rights, and the cadence of review.
If the work involves your team, ask how knowledge transfer will happen. A good consultancy should leave a workflow your team can run, not a dependency on the consultant.
5. Check the boundaries
A credible company should be explicit about what it cannot promise: traffic, leads, revenue, headcount reduction, or regulatory approval. It should also explain how it handles data, confidentiality, and errors.
The strongest signal is usually a willingness to decline a project that does not fit. A vendor that says yes to everything is not diagnosing.
Common questions
Should we choose a local Shanghai company or a remote provider?
Choose based on access to your team and the relevance of the work, not location alone. Local presence helps for on-site diagnosis and enterprise coordination; remote delivery can be efficient for advisory and content workflows.
What should a first engagement include?
A bounded diagnosis with a current-state map, a prioritized use case, a deliverable definition, a baseline metric, and a decision on whether to proceed. It should not require a large upfront transformation budget.
How do we compare proposals?
Compare the problem definition, the first-stage deliverable, the people doing the work, the review cadence, and the explicit exclusions. A lower price with a vague scope is usually more expensive.
Do you work with companies outside Shanghai?
Yes. AndyAwake is based in Shanghai and works with clients across China and internationally. Most advisory work is online; on-site enterprise work is scoped case by case.
About AndyAwake
Company background, method, founder, and the boundaries we work within.
AI Chief Andy
Former AI product lead at Tencent and Baidu, VP at a large-model company, co-founder of 流量 AI. Founder of Shanghai Andy Xingjue Technology Co., Ltd.