AI Implementation Without a Goal: Why Instructions Don't Work
A case from Habr: a company requested AI instructions and prompts but forgot to define the goal of implementation. The tools were handed out, but employees don't use them.
On October 30, Habr published an analysis of a case: a company with three client-server products and three product owners tasked an employee with the "practical implementation of AI into the processes of creating and maintaining products." The request included a specific list: a set of instructions on using neural networks, ready-made prompts, starter skills for managers — prototyping and working with hypotheses, for development — code generation, testing, bugs, refactoring. The author notes: the request looks responsible, there is analysis and breakdown by roles, but there is no answer to the first question — what is the goal of AI implementation.
The difference from the typical approach is that here implementation began with tools, not with a task. The author compares this to putting a product into operation before testing and before writing a technical specification. According to him, management sees that everyone is implementing AI, fears falling behind, and tasks a familiar employee with authority to "implement AI." He conducts the analysis, deploys the tools, issues instructions — but colleagues don't use them and look unfriendly. This is not an isolated case: the request that the author hears increasingly often repeats the same mistake — starting from the end.
In practice, this means that without a sold idea of change and without an answer to the question "why," any instructions and prompts remain dead weight. Employees do not see the connection between the new tools and their daily work. The author does not provide figures on time or money losses, but emphasizes: first you need to define the goal and metrics, then choose tools. Otherwise, implementation turns into a simulation of activity: there is reporting, but no changes.
It remains unknown what exact goal of AI implementation was eventually formulated in the company and whether the situation changed after publication. There is also no data on how many employees actually began using the proposed tools.