The repetitive steps are handed over to AI, and the energy can go back to the more important things.The customer throws a question in the group, and the engineer actually sits down to analyze the failure, and how much work is left in the middle?
In one of Bosch's research and development departments, the answer used to be: find the messages scattered across groups and private chats, organize them into a comprehensive description in multiple dimensions, judge by experience who to find, then meet again, and then write the conclusions into a report.
The failure analysis has not yet begun and most of the day has already passed.
Today, a large section of the road in between has been handed over by engineers to beapoints and beapoints partners.
01
This is the first time I've seen this.
How to get AI to work after communicating
The R&D department of the automotive industry chain often encounters the following situation: an engineer has to do three things: R&D details are needed, and there are many daily repetitions that need to be done.
A team of engineers in a business department at Bosch needs to connect different customers to provide a comprehensive system solution for customers. Faced with the same customer, different projects will have different groups, testing, routing, producing any link, a message emerges: which project, which software, what fault codes reported, what errors occurred.
The engineer received the problem information, containing the field to be repaired ((defective, demo indicated) image source: beans packing work
He had to integrate the fragments in several different dimensions, including the respective projects, software versions, vehicle configurations, test conditions, correction times, diagnostic screenshots, and then, based on his own experience, or transfer them to the research and development team responsible for the corresponding modules.
It was two things that made this engineer decide to change his mind.
One is knowledge dependency. Experience is embedded in the individual minds of engineers, cloud documents, knowledge bases are not there, but everyone's templates and habits are different, projects are rushed, no one can look into them.
If this is no longer a matter of consciousness, whether it's a message or a meeting, the AI can automatically remember it in the unconscious state, which can save a lot of time and energy.
The second is that customers do not ask questions according to templates. The problem is that the flow occurs, the message goes on and on, and no one can give fault codes, screenshots and fourteen-five fields in one message at a time.
These issues have been addressed by the involvement of soybean workers and soybean partners.
Human-powered collaborative workflow illustrates image source: beans packaging work
In the real-life work of an engineer, the problem usually starts with a private chat, where the engineer can configure timed tasks in a bean bag job, inspect the private chat messages he receives, identify fault types of information, and then forward them personally to the special topic group for problem reporting.
In this process, the customer's communication habits do not need to change, engineers have less manual transportation, and things outside the process are pulled back into the process.
The message enters a special group, each issue is a separate topic, the material of the different tasks is unmixed, naturally forming a "one thing, one thing" interface.
The beans pack partner crawled through the historical context of the clusters, describing the complete problem of assembling the flowing pieces, then calling on the skills to generate well-packaged reports, creating a preliminary draft according to standard templates, while registering them in a multi-dimensional spreadsheet, reminding the corresponding engineers to process.
In a real problem-processing record, the AI organizes the message into a flyer document, the problem status tag is "in progress"; the drive form, the drive pattern, the incomplete information, is kept "to be added", and the engineer needs to verify and complete the initial draft without having to start with a blank document.
Producing problem reports and registering problem status ((defective, demo shows) image source: beans bag working
Some details can only be finalized at the meeting.
Engineers can invite their buddy to participate in a problem communication meeting, which will be listened to throughout the session, and after the session, will synchronize the clarified information directly into reports, discussions, records and reports, avoiding the link of artificial translation. According to the engineer's own estimate, the AI initial draft can cover 78% of the content, leaving the remaining 2% to the judgement of the person; the conclusion of the meeting is completed, the report is basically alive, and then sent to the customer through artificial review.
After the meeting, complete information, supplemental report ((defective, demo shown) image source: beans bag work
The report is simultaneously archived into the team's knowledge base, and the status and progress are stored in a multi-dimensional form: the form key is followed up, the knowledge key is stored.
It can use two categories of knowledge: one is external industry knowledge, such as what the General Failure Codes each represent; the other is the history of the team itself, even the problems that have been dealt with by departing engineers that can be cited.
The project took just over a week from demand to landing. In fact, it really took just one day for the partner to get addicted, and after a day of working with the engineer, the project was in operation, and the next week was used to fine-tune the details, such as which reports were written too rough, which content shouldn't be put in, which descriptions weren't accurate enough, an engineer could tell it: remember this next time.
Currently, there have been 56 problems reported after the beans bag working partners, and it has been able to stabilize at the level of 780.
The remaining twenty are left for the final shutdown. How to determine the cause of the failure, how to arrange the verification plan, what conclusions can be drawn from the outside, these critical tasks are still to be manually audited, supplemented and confirmed by engineers.
02
In a remarkable job,
Hiding the true value of AI
After the introduction of bean bag work, engineers can move less messages, write less reports, encounter new problems, look for processing experience from historical work, give repeated steps to AI, and focus on problem analysis, verification, and communicating these more important things to customers.
In a team of three people, these are real capacity improvements.
For the team, a deeper change is in the flow of information. Last time communication ended, information was scattered in chat logs, personal clouds and someone's minds; now messages are automatically turned into tasks, reports and knowledge base entries, once the records are processed, and from personal files into colleagues' shared assets.
Combining historical knowledge libraries and conference inputs to form a problem reference 已脱敏,demo indicated
In the past, it has been easy to focus on how AI can do complex professional tasks for humans, how it can code, diagnose, and program. But in real knowledge work, professional judgment is only part of the time, and more time is spent after communication: sorting out fragmented information, recording meeting conclusions, passing questions to the right people, keeping track of progress, and archiving traces. These things do not have a technological ring, fragmentation, repetition, but actually occupy human energy on the ground.
Just as in the collaboration between Bosch and Bean Pack, AI creates value rather than being at the heart of the workflow by leaving communication, judging and solving professional problems to people, and handing over the organizing, recording and action after communication to AI as much as possible, so that engineers can get out there and do what they really need to do.
AI doesn't have to do the most important things to create the most important value.
* Source: The work of the bean bag
This article was originally published by Geekpark GO.
A question from the PolesAI is crumbling.
Is it the best option for industrial landings?
