Author: ICONIQ

Compiled by: Tim, PANews

The development of artificial intelligence has entered a new chapter: from hot topics to practical implementation. Building large-scale AI products is becoming a key battlefield for competition. The 2025 AI Status Report "Builder's Handbook" shifts the perspective from technology adoption to practical implementation, and deeply analyzes the full set of solutions for AI products from conception, implementation to large-scale operation.

Based on the results of an exclusive survey of 300 software company executives in April 2025, combined with in-depth interviews with AI leaders within the ICONIQ community, this report provides a tactical roadmap to transform the intellectual advantages of generative AI into sustainable business competitiveness.

The report distills five key chapters and how they will help teams actively build AI applications.

1. Artificial intelligence product strategy has entered a new stage of maturity

Compared to companies that have only integrated AI into existing products, AI-led companies are bringing products to market faster. Data shows that nearly half (47%) of AI-native companies have achieved critical mass and proven market fit, while only 13% of companies with integrated AI products have reached this stage.

What they are doing: Intelligent workflows and vertical applications have become mainstream. Nearly 80% of AI native developers are working on intelligent workflows (i.e. AI systems that can autonomously perform multi-step operations on behalf of users).

How they’re doing it: Companies are converging on multi-model architectures to optimize performance, control costs, and match specific use cases, with each respondent using an average of 2.8 models in customer-facing products.

2. Evolving AI pricing models reflect unique economic characteristics

Artificial intelligence is changing the way companies price their products and services. According to our survey, many companies are adopting a hybrid pricing model, adding a usage-based billing model on top of a basic subscription fee. Other companies are exploring pricing models that are entirely based on actual usage or customer usage results.

Many companies still offer AI capabilities for free, but more than a third (37%) plan to adjust their pricing strategy in the coming year to make prices more consistent with the value customers receive and their use of AI capabilities.

3. Talent strategy as a differentiating advantage

Artificial intelligence is not only a technical problem, but also an organizational problem. Most of the top teams are currently forming cross-functional teams consisting of artificial intelligence engineers, machine learning engineers, data scientists, and AI product managers.

Looking ahead, most companies expect 20-30% of their engineering teams to focus on AI, while high-growth companies expect this proportion to be as high as 37%. However, the survey results show that finding the right talent remains a bottleneck. Among all AI-specific positions, AI and machine learning engineers take the longest time to recruit, with an average time to fill of more than 70 days.

Opinions were split on the pace of recruiting, with some recruiters saying it was progressing well, while 54% said it was lagging behind, with the most common reason being a lack of qualified talent.

4. AI budgets are surging and showing up in company P&L statements

Companies that use artificial intelligence technology are investing 10%-20% of their R&D budgets in the AI field, and companies in all revenue ranges will continue to grow in 2025. This strategic shift increasingly highlights that AI technology has become the core driver of product strategy planning.

As AI products scale, the cost structure often changes significantly. In the early stages of product development, human resource costs are usually the largest expense item, including personnel recruitment, training, and skill improvement costs. However, as the product matures, cloud service costs, model inference costs, and compliance and regulatory costs will account for the majority of expenses.

5. The scale of AI applications within enterprises is expanding, but the distribution is uneven

Although most surveyed companies provide about 70% of their employees with access to internal AI tools, only about half actually use these tools on a regular basis. In larger and more mature companies, it is particularly difficult to encourage employees to use AI.

Companies with high adoption rates (i.e. more than half of their employees use AI tools) deploy AI in seven or more internal application scenarios, including programming assistants (77% usage), content generation (65%), and document search (57%). These areas have improved work efficiency by 15% to 30%.

The AI tool ecosystem is still fragmented, but is gradually maturing

We surveyed hundreds of companies to understand the technical frameworks, libraries, and platforms that are actually running in production environments today. This report is not a simple ranking, but a true reflection of the tools adopted by developers across different fields.

Here is a quick overview of the most commonly used tools in alphabetical order: