Step 1: Set clear goals.
Step 2: Determine pilot scope.
Step 3: Gather stakeholder input.
This week, I’d like to write about Step 4: Document current processes
In all my years, it’s rare that I’ve ever come across someone who LOVES to document processes. Most people would rather COMPLETE processes, IMPROVE processes, FIND BETTER processes, etc. But document processes? No thanks. I get it.
However, when experimenting with AI, details are critical. If you want to implement a system to do some of the work people can’t do or don’t want to do, the details are as important as anything else.
If you are training an AI system to remove mundane, tedious tasks – like we do at Humaxa – you will probably need to talk with multiple people and ask them about the details of their workflow. What are they doing that might be considered a “workaround” that could be off-loaded to an AI? They may not even consider those processes “workarounds” because they’ve been doing work that way for so long.
Let’s look at an example: Compliance Reporting. In automotive or other manufacturing industries, detailed reports are often prepared to comply with industry regulations and standards. Compiling and verifying large amounts of compliance reporting data can be time-consuming and susceptible to human error. People in your organization may already be using compliance management software that automates data collection and reporting, but it may be legacy software that was not built with AI at its core. If you are thinking of using AI to redesign the compliance reporting process, you will need to understand and document the details of how people prepare compliance reports or use software to do so. Only then can you identify areas where AI could streamline the process and enhance accuracy.
Supplier Coordination could be another place where AI might help reduce manual, tedious work. Often, it’s necessary to coordinate requests or proposals from multiple suppliers to ensure timely delivery of parts and materials. However, managing supplier relationships and tracking proposals can be complex and error-prone, leading to delays. Using AI to automate the generations of RFQs or automating the analysis of quotes and proposals can improve coordination and reduce errors, while freeing up workers’ time. In order to implement such a system, you’ll need to know the detailed steps for creating an RFQ. What are all the requirements that go in the request? Who determines what those requirements are? How are the requirements determined? What research needs to be done? By documenting all the details that go into the process, it’s possible to manufacture an AI that will generate those RFQs for you. (That’s part of what we do at Humaxa.) But without the details, steps can get left out.
Manual Data Entry is another area where AI can help reduce tedious work. Recording and maintaining detailed logs of production data, quality control metrics, and compliance records can feel mundane and like it returns little value. It can also be slow and prone to errors, leading to inaccuracies and inconsistencies. Implementing an AI that can automatically capture data that’s entered in a rote fashion can reduce busywork and streamline data entry processes. To create an AI that can do this, however, it’s first necessary to have a detailed process of how the data is entered today. Even though the data entry itself may feel mundane, sometimes there are subtle nuances built into the processes. When is it entered, how is it verified, and are there ever changes in how/when/why it’s entered? Document all of these decisions in detail for the best results.
The tools you use to document these detailed processes will vary, depending on the processes themselves. You can use flowchart software, PowerPoint, even pen and paper will work – but don’t forget to get the details around what/when/how/why – they are critical to your success.
I hope this helps and let me know how it goes!
Carolyn Peer
CEO/Co-founder
carolyn.peer@humaxa.com
