In first-half new signings, AI application development accounted for 51%. This review discusses the three changes behind the numbers and what we will bet on in the second half.
It is mid-year inventory month again. Let me be upfront: in the first half of 2026, AI application development accounted for 51% of Dushi Technology's newly signed projects, surpassing traditional custom software for the first time. We did not expect this number to come so quickly either.
Change 1: Clients no longer ask "what can AI do" but "do this for me"
Two years ago, when discussing AI, clients asked "can you do it?" This year they ask "help me automate contract review," "cut my customer service outsourcing in half," and "make inventory forecasting accurate." Requirements have moved from concepts to specific tasks, which shows the market is truly maturing.
Accordingly, our pre-sales approach has changed: we no longer present features with demos, but bring clients' real data into a demo environment, run a small scenario they care about on the spot, and discuss contracts only after they see the results.
Change 2: The delivery team structure has changed
In the first half of the year, we hired mostly two types of people: business-savvy data annotation and knowledge governance engineers, and integration developers who can connect large models into existing systems. Traditional "backend programmers" were not expanded much.
The reason is straightforward: in AI projects, the models themselves are ready-made; what really takes time is data organization, business rule mapping, and system integration. Skilled workers in these three areas are still scarce in the market.
Change 3: Renewals from existing clients underpin the base
In first-half revenue, additional purchases and renewals from existing clients accounted for more than 60%. Many old projects were inventory management and membership systems built years ago; this year, the first thing clients ask when they return is "can you add an AI customer service to this system?"
This path gave us a judgment: for a large number of SMEs, AI does not mean replacing old systems but adding an "AI plug-in" to them. In the second half, we are specifically building a lightweight integration solution that lets existing systems connect to large model capabilities with minimal changes; it is already being piloted in three industries.
Three bets for the second half
First, the knowledge middle platform: if enterprise data does not enter the knowledge base, AI just spins idle. Second, deep operations of AI customer service: going live is only the beginning; continuous tuning is the moat. Third, AI transformation of old systems: this is the biggest opportunity in the existing market. The industry understanding Shenzhen Dushi Technology has accumulated over 16 years will all be bet on these three lines. Existing clients and new friends are welcome to talk with us about which part of your business is most worth connecting to AI first.