The biggest bottleneck of the enterprise information department is definitely not R&D efficiency.
The business side says the requirement is urgent, the IT team works overtime to develop and launch the system, but then no one uses it — the bottleneck is not how fast you can build it, but whether the value closed loop can be formed and operate effectively.
Most enterprise IT teams must be familiar with this scenario: the business team rushes over in a great hurry saying "this requirement is extremely urgent", the digital team works around the clock and launches the system in just a few days. Then what happens? The system that the business team claimed to be urgently needed just lies idle on the server with no users. Two weeks later, the business team comes again, putting forward an even "more urgent" requirement...
Why are the developed systems left unused? Sometimes the business team blames the development team for "insufficient R&D efficiency", so the development team spares no effort to deploy low-code platforms and AI Coding tools to further speed up delivery to meet the business' "requirements". But reality often deals a harsh blow: the system that took one month to launch before and had no users now takes only one week to launch, and still no one uses it. The more advanced the technology and the higher the development efficiency, the more useless garbage systems will be generated.
Thus it is not hard to see that the real bottleneck for the IT department is not "how fast you can deliver", but "whether what you build is correct, whether people will use it, and whether it can generate value after being used". The combination of these three points is called the value closed loop. Efficiency is the most dispensable part in this loop.
I. High Efficiency Does Not Equal High Value
If the delivery efficiency of an enterprise's digital transformation is even low, there must be something wrong. But high efficiency does not mean high value. The IT departments of many enterprises are trapped in a vicious circle: the business side says the requirement is urgent, the development team works overtime to launch it, no one uses it after launch, then new requirements are raised, and the development team continues to work overtime to launch new systems. This is never an efficiency problem, but the absence of the second half of the value closed loop.
The emergence of AI Coding has further improved the speed from requirement to code, but it cannot solve the following problems: are the business side willing to use the system? Can it generate business results after launch? Can the data quality support normal operation? Will the organizational process be adjusted accordingly?
Therefore, the more powerful the tools are, the easier it is for the IT department to "produce more unused systems faster". R&D efficiency is an underlying capability, not the core bottleneck. The core bottleneck of enterprise digital transformation lies in the value closed loop, i.e. whether the whole chain from requirement discovery, verification, launch, usage to business results can operate smoothly.
II. Apart from Requirement Quality and System Adoption, There Are Eight Other Types of Pitfalls
If you think the top priority for enterprises in the digital transformation process is to solve the problems of requirement quality and system adoption, your understanding is quite good, but apart from these two points, the IT department also needs to face the following eight major pitfalls
First, Lack of Value Definition
No one clarifies before launch for many requirements: whose problem to solve, what the success criteria are, which indicators to track, and where the data comes from. The launch is regarded as the end of the project, and when reporting at the end of the year, the team can only count "how many systems have been launched" but cannot tell "how much business growth has been brought".
Second, Mixed Authentic and Fake Requirements, Disordered Priority
The sources of fake requirements are very fixed: requirements raised just because of a leader's casual remark, requirements put forward by the business side to cope with inspections, requirements to follow competitors' actions, requirements for the partial interests of a single department, and requirements that the technical team wants to practice their skills on. All these are packaged as "extremely urgent" but have not been verified by real users.
Third, Absence of System Operation
A large number of systems have no promotion plans, no training, no seed users, no feedback channels, and no iteration mechanisms after launch. Once the development team leaves after the project is completed, the system will naturally be abandoned completely.
Fourth, Poor Data Infrastructure
Digital transformation and AI implementation both rely on data, but many enterprises face severe data silos, conflicting indicator calibers, messy master data, dirty historical data, and poor interconnection between systems. The data is inaccurate and the user experience is poor after the system is launched, leading to the final abandonment, and the root cause usually lies in this aspect.
Fifth, Disconnected Organization and Collaboration
The business and IT teams talk at cross-purposes, the IT team passively receives orders without decision-making power, there is a lack of "interpreters" who understand both business and technology, the business side takes no responsibility for the requirements they put forward, and all the blames are shifted to the IT team.
Sixth, Technical Debt and Architecture Sprawl
To "respond quickly", teams keep adding features, more and more isolated chimney-style systems are built, redundant construction is rampant, integration complexity rises sharply, and the core system becomes increasingly fragile. After three to five years, even modifying a single requirement becomes extremely slow, and the efficiency is dragged down by the problems accumulated before.
Seventh, Risks of Security Compliance and AI Governance
The widespread application of AI brings a new set of issues: data privacy leakage, wrong decisions caused by model hallucination, out-of-control permissions, algorithm bias, and unexplainable models. If no assessment is made in the early stage, major problems may arise in the later stage.
Eighth, Inaccurate AI Scenario Selection
Too many enterprises follow the wrong path of "looking for scenarios with existing technology" instead of "looking for appropriate technology for existing scenarios". They choose scenarios with cool technology but low business value, force the deployment of AI without sufficient data, get no cooperation from the business side so that the AI outputs are not used, and hold overly high expectations that AI can replace humans to make complex judgments.
III. How to Improve the Situation?
This is a long-discussed issue that Lao Yang has mentioned many times in his previous articles, and the key points are briefly summarized as follows:
1. Reposition the IT department, transforming it from a Delivery Center to a Value Center
The responsibility of the IT department is not "to build systems", but "to enable the business side to generate measurable value through digital transformation".
2. Establish requirement governance and value review mechanisms
No subjective ideas or requirements from any business department can be directly developed, and the requester must clarify clearly: what the business problem is, who is affected, how much loss is caused by the current situation, the expected result, success indicators, data sources, and how many resources the business side is willing to invest to cooperate. A digital product committee composed of representatives from the business, IT, data, and finance teams must be established to conduct unified review and priority sorting. It is also necessary to implement the "requester responsibility system": the business side that puts forward the requirement must be responsible for the usage rate and final results. If no one uses the system after launch, the business side shall take the corresponding responsibility instead of shifting all the blames to the IT department.
3. Verify requirements with lean and product thinking
Do not pursue large and comprehensive functions, but verify the authenticity of the requirement with small costs first: user interviews, on-site observation, DEMO, gray release, and A/B testing. The only principle is to verify that "there are users who are willing to use the system and the system can solve the problem" before investing in large-scale development. AI Coding can play a key role here: quickly generate prototypes, internal tools, and data dashboards for the business side to try out before deciding whether to approve the project.
4. Build the operation closed loop
Launch is not the end, but the starting point of operation. Promotion plans, seed user cultivation, training, feedback channels, key behavior monitoring, and regular reviews are all indispensable parts.
5. Transform from project-based management to product-based management
Do not leave after the key system development is completed. It is necessary to assign a dedicated product manager, formulate a long-term roadmap, and carry out continuous iteration. In this way, the system can be optimized with usage instead of reaching its peak right after launch.
Final conclusion:
The biggest bottleneck of the digital department is not R&D efficiency, but the value closed loop. AI Coding has made efficiency no longer the main contradiction, but makes the problem of "whether the direction is correct and whether the value has been verified" more prominent. What enterprises need to do is not to deploy more tools, but to adjust the positioning, assessment, governance and operation culture, to promote the digital transformation work from "delivering projects" to "delivering real business results".
This article is from the WeChat official account "Xiangjiang Digital Review" (ID: benpaoshuzi), written by Lao Yang, and authorized for release by 36Kr.