Autonomous workplace assistants (AWAs) are intelligent agents that use embedded monitoring, conversation, detection, and decisioning to complete a workplace activity, often without central orchestration. The interesting part is that AWAs become real coworkers. They act more independently than robotic process automation (RPA), your basic automations that carve out short, repetitive human tasks. They support an employee or workplace activity in a closed or semi-closed loop, level 3 or level 4 process patterns.

Natural language processing (NLP), RPA, digital process automation (DPA), low-code tools, and machine learning were the primary AWA building blocks, a delight of process automation acronyms. These were developing nicely on their own, and then generative AI crashed the party. The ability to write coherently, revise text, compile software, give sound direction, and generate creative ideas gave AWAs a boost. They now apply for jobs in legal and marketing research, customer service, and hundreds of others.

Take Auto Repair — Please

Auto repair is something most of us can painfully relate to, and it is a good example of how large language models (LLMs) could affect the workplace. A real talent of the generative AI LLMs is that output can be adjusted for a target educational or skill level. This could be a huge win for repair technicians. An auto mechanic could provide symptoms or submit images from an oscilloscope and just ask, “What do I do now?” The AWA would adjust the output for a vocational teacher or beginning technician.

We spoke with Jim Fish, the vice president of Opus IVS. His company provides intelligent vehicle support (IVS) solutions for thousands of repair shops and dealerships, and he thinks today’s mechanics are already beginning to struggle. He told us: “More chips, more sensors, more complexity will drive the use of LLM-based support. The average technician can’t keep up. Your 2000 Ford F-150 had a handful of computers. Modern cars have up to a hundred that now control entertainment, fuel management, stability, and control.”

As Joe Biden might say, “Here’s the deal.” Down the road (pun intended), mechanics will need AWAs infused with LLMs. Work life will change for the 733,000 automotive service technicians and mechanics that the US Bureau of Labor Statistics tracks. On the plus side, any mechanic will be able to reach higher, to put a Superman cape on, but also, AWAs will allow employers to reach lower. Today’s median salary of $46,880 will be challenged. A less-skilled employee can turn from a C-player into a B-player. This means that the Jeff Daniels character from the movie “Dumb and Dumber” could be turned into at least a competent repair technician.

The Problem, As Always, Will Be Trust

The question for these AWAs, as with many areas of AI, will be trust. Consider this scenario: Your car is in a collision. You take it to the local body shop. These mechanics are not as skilled as auto mechanics who might diagnose an electrical problem at the dealership. Former territories of the British Empire accurately called them “panel beaters,” a term that nails it (sorry). They are good with a hammer and a torch.

After an accident, your late-model car will need adjustments to the advanced driver-assistance system (ADAS). In layman’s terms, this is your car’s self-driving support. Eleven percent of all collisions already needed an ADAS adjustment. They need to be perfect and are beyond the skills of most body shop workers. Today, without an AI bot to help, they must call in experts. A mobile technician will visit the shop or they send the vehicle to the dealership. The AWA will soon allow them to make the adjustment, but do you trust the combination of a lesser-trained human and an LLM to calibrate your ADAS, or do you want an expert technician in the loop?

Join The Conversation

If you have perspectives to share, please schedule a briefing and tell us all about them. If you’re a Forrester client and want to discuss (or challenge) our thinking on this topic, please schedule an inquiry.

Related Forrester Content