Key question New
AI-processed · Author-approved
How can traditional companies successfully integrate generative AI into repeatable, everyday business workflows?
Loads Spotify's player; Spotify may set cookies. This device remembers your choice.
Key findings
- Revealed that AI stabilization unfolds across four distinct stages: probing, patching, reweaving, and externalizing.
- Found that initial probing generates organizational awareness and exploratory use but yields minimal operational productivity gains.
- Observed that patching resolves isolated task-level bottlenecks without altering core operations, providing limited overall efficiency improvements.
- Demonstrated that workflow reweaving restructures both human roles and software configurations into tightly controlled pipelines, reducing design turnaround times from days to minutes.
- Showed that once internal workflows reach stability, they can be decoupled from original practitioners and packaged into standalone platforms for external clients.
What it means for you
CIO / IT Executive
Audit current operations to map existing human-software workflows and allocate dedicated resources to restructure tasks around both technological capabilities and human roles, rather than simply deploying individual AI tools.
IT Manager
Document the ad-hoc prompt patterns and workaround solutions currently used by employees during exploratory AI testing to identify isolated task-level bottlenecks ready for patching.
Business Strategist
Analyze core operational pipelines to identify where manual design turnaround times are slowest, and design pilot structures that tightly integrate human review steps with software configurations.
Researcher
Conduct a longitudinal qualitative study tracking how informal generative AI usage transitions through the specific four stages of probing, patching, reweaving, and externalizing within a traditional firm.
Policymaker
Establish digital transformation grant criteria that explicitly fund organizational workflow restructuring and human-AI capability integration programs rather than just software procurement in traditional industries.
Summary
This research investigates how generative AI transitions from informal trial-and-error into structured, reusable organizational workflows within traditional industries. Using an interpretive qualitative case study design, the authors analyzed four years of AI adoption at a Chinese real-estate company through 37 semi-structured interviews, system demonstrations, and internal documentation.
Problem
While individual employees can rapidly adopt generative AI for discrete tasks, organizations struggle to stabilize open-ended AI tools into cohesive business routines. This integration challenge is especially acute in traditional firms lacking mature digital infrastructure, where established workflows have historically absorbed tools rather than restructuring around them.
Transcript
Expert: Thank you, Anna. It is great to be here.
Host: Alex, let us start with the core problem addressed in this study. Many companies see their employees experimenting with generative AI, but they struggle to move beyond ad-hoc usage. Why does this happen?
Expert: While individual employees can rapidly adopt generative AI for discrete tasks, organizations frequently find it difficult to stabilize open-ended AI tools into cohesive, repeatable business routines. This integration challenge is especially acute in traditional firms that lack mature digital infrastructure. In those environments, established workflows historically absorb new software tools rather than restructuring around them.
Host: How did the researchers investigate this organizational challenge?
Expert: The study, authored by Weilin Pu, Felix Tan, Eric Lim, and Haifeng Xu for the International Conference on Information Systems in 2026, utilized an interpretive qualitative case study design. The authors analyzed four years of generative AI adoption at a Chinese real-estate company. Their evidence base included system demonstrations, internal documentation, and thirty-seven semi-structured interviews—including senior executives, designers, IT developers and project managers—.
Host: What did their analysis reveal about how AI workflows become stable?
Expert: The study identified that stabilization unfolds across four distinct stages: probing, patching, reweaving, and externalizing. In the initial probing stage, employees explore tools independently. This creates organizational awareness, but the study found it yields minimal operational productivity gains.
Host: What occurs in the patching stage?
Expert: In the patching stage, teams use AI to resolve isolated task bottlenecks, like producing early design concepts. However, because core operations remain largely unchanged, overall efficiency improvements remain limited.
Host: When does an organization begin to see measurable operational gains?
Expert: That happens during the reweaving stage. Here, the organization restructures both human roles and software configurations into tightly controlled pipelines. By standardizing prompts, integrating custom technical components, and repositioning personnel into specialized review roles, the firm reduced design turnaround times from days to minutes.
Host: That is a significant operational reduction. What happens in the final stage, externalizing?
Expert: In externalizing, once internal workflows reach stability, they are decoupled from the original practitioners and packaged into standalone platforms that can be offered to external clients.
Host: What are the key takeaways for business leaders seeking to operationalize generative AI?
Expert: First, managers should recognize that informal employee experimentation alone will not generate systemic productivity. Second, achieving meaningful operational efficiency requires workflow reweaving, which means actively redesigning human job roles and technical pipelines together. Finally, standardizing and modularizing these stabilized workflows allows companies to potentially turn internal operational solutions into external, commercial software offerings.
Host: A practical framework for moving from casual AI adoption to scalable routines. Thank you for sharing these insights, Alex.
Expert: You are welcome, Anna.
Host: And thank you to our listeners for joining us on A.I.S. Insights — powered by Living Knowledge. Tune in next time as we continue examining practical research in business and technology.
Generative AI, Workflow stabilization, Imbrication, Digital transformation, Design workflows, Qualitative case study