Andrew Ng's Advice on How Manufacturers Can Get Started With Computer Vision

Computer vision research and systems have a surprisingly long history, dating back to the late 1960s(!). It could be said that the practical application of computer vision for real-time face recognition brought the technology into the public consciousness in the 2000s. In the 2010s, vast improvements in computer vision accuracy further increased its potential to address real world challenges. Today, many are aware of the technology, perhaps having seen its application in self-driving car systems like those of Tesla and surveillance schemes in authoritarian states.

Nonetheless, you could be excused for viewing computer vision as a “futuristic” technology: More should be done to educate the AI-curious about the potential and limitations of computer vision.

Ex-Google Brain Lead, Andrew Ng has been a leading force in educating the mainstream about computer vision, publishing the AI Transformation Playbook to help business leaders demystify what it takes to stand up AI projects in their organizations. This post covers the key steps from that playbook.

The AI Transformation Playbook

Ng lays out the following steps to get going with AI:

Steps from AI Transformation Playbook

  1. Execute pilot projects to gain momentum
  2. Build an in-house AI team
  3. Provide broad AI training
  4. Develop an AI strategy
  5. Develop internal and external communications

He mentions that these steps don’t have to be taken in the order listed above — perhaps some could be done in a different order or in parallel — but that this is one sequence which can work for most companies.

Step 1: Execute Pilot Projects To Gain Momentum

Interestingly, Ng’s guidance on selecting a pilot project is to find a project that is highly likely to succeed rather than one that is highly valuable. This makes sense considering the aim here is to build momentum rather than prove value.

At Google, Ng started with a speech recognition project which was viewed as a less-valuable area of Google’s bottom line than, say, online advertising. However the success he could demonstrate in this area drew the attention of other, more valuable product groups.

Show traction within 6-12 months (can execute in-house or outsource).

Step 2: Build An In-House AI Team

Follow a matrix set up: Build a centralized AI team which can collaborate with leaders across various BUs. The rationale behind this is that BUs are not necessarily AI experts, and that AI talent can make an impact across BUs.

An AI/computer vision team can be set up to report to the CEO or the CIO, CTO, CDO (even a new CAIO) and collaborate with various BUs, responding to their biggest needs.


Equip the AI Team with sufficient resources and mandate them to have the biggest impact possible.

Step 3: Provide Broad AI Training

Not only engineers need to understand AI – executives, senior business leaders and leaders of divisions working on AI projects should all understand how AI can impact their business objectives.

Executives and senior business leaders should understand how to set AI strategies. All leaders who have AI projects within their domains should understand the resource allocation needs of AI.

The right training should be curated (not created) for business leaders.

Step 4: Develop an AI Strategy

Ng mentions that although many executives want to set a strategy as their first step, having this as step 4 is intentional: It’s important for leaders to get their feet wet before they make decisions about AI objectives and resource allocation.

An AI strategy should leverage AI to create an advantage specific to your industry sector. For manufacturers considering using computer vision, this could include using computer vision to increase product quality or labor productivity to industry-leading levels.

Also consider creating a data strategy. This could include strategic data acquisition (e.g. data collected by manufacturers of complementary products) and creating a unified data warehouse.

An AI strategy could seek to use computer vision to decrease product defect rates by 50%, thereby creating savings which can be passed on to customers, providing a competitive price advantage.

AI and computer vision should be seen as tools to help achieve competitive advantage.

Step 5: Develop Internal and External Communications

Communicating your company’s AI direction and strategy with investors, relevant government agencies (if applicable to your industry), customers, talent and internal teams can be crucial to achieving success.

Embarking on an AI strategy to achieve industry-leading defect rates can be important information for investors to correctly value your company.

Understand the impact of your AI strategy on relevant stakeholders and communicate that.

Learn More and Get Started With Computer Vision

If you’d like to learn more, we can recommend Ng’s course, AI For Everyone which is available on Coursera and is a great introduction to AI and its potential.

If you’d like to get started with a computer vision pilot project, schedule a demo with JourneyApps to learn how our platform supports rapid development and deployment of computer vision models to edge devices.

← Back to all posts


The development platform

for industrial apps

Try For Free