A saying goes that startups only do two things: create a "product" and sell it.

In the "selling products" phase, AI+sales and AI+SDR have become popular choices—take 11x as an example. Its "digital employee" Alice can automatically collect customer lists and then mine potential customers through emails and LinkedIn messages. It is reported that 11x's ARR has reached $10 million and has just completed a $50 million Series B round led by a16z. Another lead generation tool, Clay, integrates over 75 data-rich providers and uses agents to crawl, analyze, and summarize web information. Currently, Clay is valued at $500 million and has received two rounds of funding from Sequoia Capital.

In the "creating products" phase, AI is also expected to reduce costs and increase efficiency.

On one hand, R&D expenditures are huge. The latest UK statistics show that R&D expenditure reached £71 billion in 2022, with £50 billion coming from the business sector. In the US, this figure is approximately $886 billion, with $690 billion coming from corporate investment.

On the other hand, substantial R&D expenditures are essential. A 2023 McKinsey report predicts that within the next five years, one-third of sales (totaling approximately $30 trillion) in major industries such as automobiles, telecommunications, and consumer goods will be attributed to new products.

The Harvard Business Review has pointed out that if a company allocates 70% of its resources to core project innovation, 20% to adjacent innovation, and 10% to disruptive innovation, it will outperform its peers. Additionally, studies have shown that technology companies should increase investments in the latter two types of innovation.

So, what role will AI play in product development? Recently, FT published a research paper titled "AI and the R&D revolution" exploring how AI can excel in various stages of product development. Without altering the original meaning, we have compiled and moderately rewritten the article.

All enterprises are essentially technology companies

The core of innovation lies in mindset. Sean Ammirati, a professor of entrepreneurship at Carnegie Mellon University, states that fostering an entrepreneurial culture within R&D departments can drive innovation, even in large enterprises. Teams with this mindset are more likely to propose disruptive rather than incremental product development solutions.

Having founded multiple machine learning startups, Ammirati believes that many companies fail to invest sufficient resources in adjacent and disruptive innovations. In the current era of technology proliferation, every company should consider itself a technology company when planning its R&D budget.

Clarify your goals and target customers

If you are a door and window manufacturer, who are your target customers?

In Iowa, a door and window manufacturer named Pella does not consider homeowners as its primary users but instead targets window and door installers. They have developed a window product that can be installed indoors, reducing the risk for construction workers working on high-rise buildings.

Setting clear goals is equally important for the R&D process, including determining whether your objectives are to reduce material costs, engineering costs, speed up time-to-market, or a combination of these. Keeping your goals in mind will help you set KPIs to evaluate the process.

The more data you have access to, the easier it is to understand customer needs

Customer needs are always the inspiration for product development.

For example, Taobao operators can clearly understand what products 25-30-year-old women like to search for at 9:30 p.m. every night. However, Taobao sellers cannot access this data.

In contrast, the main value of branded independent websites (where brands host their own websites) lies in retaining and interpreting customer data. This real-time insight can help companies optimize and expand their product lines.

For instance, the lingerie brand Lively noticed that many women were searching for strapless bras and subsequently added them to their product line. Tommy John observed that women liked to buy men's underwear and subsequently launched women's "boxer briefs," T-shirts, and pants.

During data collection and organization, ensuring data accuracy is crucial. Especially when deploying AI systems, which are essentially statistical models based on data, any laxity in data input can lead to continuous subsequent issues.

With technological advancements, more and more products can be tailored to specific user groups. For example, Fenty Beauty, founded by Rihanna, uses image recognition and machine learning to provide cosmetics tailored to customers' skin tones. Nike's "Nike By You" service allows customers to customize their sneakers from various styles, colors, and designs.

"Co-creation" goes beyond personalized services. In this process, customers can not only suggest modifications based on preset options but also directly communicate their true needs to the company, similar to what IKEA does.

At an AI summit hosted by Bloomreach, Azita Martin, Vice President and General Manager of AI for Retail and Consumer Goods at NVIDIA, noted that the future challenge for many companies is how to quickly fulfill highly personalized orders—essentially serving a "single-user market." From accurate demand forecasting to improving distribution center efficiency and last-mile delivery, AI can provide retailers with great flexibility.

Even without access to data, you can still conceive product ideas independently.

However, with AI, you can brainstorm from an idea or even a vague concept, refine product concepts, conduct market research to see if similar products exist, and analyze competitors to obtain suggestions for differentiation.

AI can also provide multiple versions and improvements of a product to meet specific market demands—some of which may be niche areas that human designers have never considered.

As the product concept matures, AI can assist in formulating market testing strategies, accelerating product design and testing processes, and providing suggestions on materials, procurement, and manufacturing processes. This is particularly helpful for startups and small businesses. For example, the founder of Skittenz, a manufacturer of decorative ski glove covers, used AI to explore suitable materials and manufacturing processes and brought their innovative products to market.

Here are some commonly used tools in R&D:

1. Market Research: Digitally distributed market research tools (such as SurveyMonkey, Google Forms, and Typeform) have extensive reach and enhanced analytical capabilities.

2. Product Design: CAD-CAM software used in engineering, architecture, and manufacturing (such as Autodesk, Siemens, and Trimble) is introducing AI capabilities. Generative design technology utilizes machine learning algorithms to propose optimized design solutions based on various parameters (such as material economy or structural strength). Engineers only need to set basic requirements and constraints (such as manufacturing processes, load-bearing capacity, and flexibility), and the system can provide multiple design variants, some of which may be innovative. The combination of AI and these tools enables R&D teams to drive innovation with unprecedented speed and efficiency.

3. Digital Prototyping: Purely digital prototyping significantly reduces R&D costs. Companies can conduct more digital experiments with cloud-supported machine learning models. Additionally, cloud computing means companies no longer need to invest heavily in internal servers. This technology brings about truly significant innovation—providing democratized technology accessible to small entrepreneurial teams.

4. Simulation Testing: From Marketing Strategies to Product Performance

Simulation technology is widely used for testing marketing strategies (such as A/B testing) to help companies determine the best way to communicate with customers. Simultaneously, it can be used in product development to test different variants of materials and designs. For example, engineering software such as Ansys and Matlab can help designers create virtual objects and test their performance in various environments, including fluid and thermodynamic behaviors. The value of such software lies in the ability to conduct tests without physical prototypes.

Market acceptance analysis is also an important application of simulation testing. Unlike marketing strategy testing, it focuses on product feasibility. For instance, it compares product pricing and performance to assess the impact of different economic environments and competitors. Although simulations demand high processing power, cloud computing can provide support.

5. 3D Printing: Lowering Production Thresholds

After successful simulation and testing, companies can utilize 3D printing to advance product development without immediately investing in actual production. In the past, prototyping required multiple steps from drawings to clay modeling; today, digitization has significantly lowered production thresholds.

The combination of generative AI and 3D printing will further enhance design flexibility. For example, designers can ask AI to generate multiple improved versions. The AI-supported 3D printing technology developed by Bosch can even adjust material input, temperature, and pressure in real-time, making prototype quality comparable to final products. This technology supports small-batch production, reducing the risk of large-scale facility investment. Additionally, Bosch's ceramic 3D printing technology accurately simulates the different shrinkage phenomena that occur during ceramic firing, ensuring high-precision manufacturing.

6. Digital Twins: From Theory to Lifecycle Management

Digital twins can dynamically update throughout a product's lifecycle, providing suggestions for improvement. This technology is widely used in managing supply chains, power plants, and other facilities, helping predict a product's performance over its lifetime.

According to Fortune Business Insights, digital twins are expected to drive the market size of computer-aided design and product lifecycle management to $26.4 billion by 2030, with the North American market accounting for approximately one-third. Although still in its early stages, EY emphasizes that the "Industrial Metaverse" is the next phase for digital twin technology. The "Industrial Metaverse" involves mapping and simulating highly complex systems such as machines, factories, and cities, providing optimal solutions to real-world problems and propelling the digitalization and intelligent development of industries into a new phase.

7. Collaboration Software: Driving Team Innovation

Basic tools like Slack support team discussions, while more specialized innovation management software (such as Miro, Braineet, and Ideanote) helps companies record ideas and data. While these tools cannot directly generate concepts, they can help teams learn from failures and improve products faster through knowledge recording and sharing.

The key to innovation is not success but the ability to fail quickly and learn from experiments. By retaining records of failed experiments, companies can avoid unnecessary reversal decisions and ensure data accuracy. Additionally, AI can help companies check for infringements and violations.

8. Future Outlook: AI + Quantum Computing

In the future, quantum computing is expected to excel in material and drug development, simulating chemical reaction processes and optimizing existing materials. Quantum sensors are also beginning to see commercial applications, such as tracking brainwaves or analyzing geological structures through gravity sensors. Battery researchers have already used quantum sensors to analyze microcurrents and improve production efficiency.

Although the widespread application of quantum technology faces challenges such as high environmental sensitivity, high energy consumption, and technical complexity, within the next 2-5 years, as related technologies advance, its commercial potential will gradually be unleashed. The fusion of AI and quantum technology will usher in new breakthroughs in R&D, pushing the speed and depth of innovation to new heights.

In July this year, McKinsey pointed out that analytical and generative AI are expected to significantly enhance innovation outcomes, including increasing market fit by up to 50%, improving product performance by 15%-60%, increasing work efficiency by up to 50%, and shortening time-to-market by up to 40%.

It is evident that the future of product development will involve "human-AI collaboration." To avoid the interference of AI's "hallucinations," a correct approach is to view AI as a co-founder or as Human plus AI. AI is more like your brainstorming partner rather than a "clone" that does all the work for you.

To win in the future, enterprises must embrace the AI revolution. After all, we are already at a "technological inflection point," and most companies that rejected the internet have since disappeared.