The Agents of Change: Where Real AI Efficiency Lives

The response to my article on whether AGI will come about has been interesting. The most common response has been to send me articles on AI and software development. But there's a conflict of interest here: most are penned by CEOs whose company valuations depend on trumpeting AI’s success. I don’t disagree with the efficiencies, but let’s be honest about the incentives. I don’t necessarily disagree with them, all I am saying is that there is a conflict of interest there. That said, I don’t dispute the efficiencies that AI has brought to software development in the article. The article was about AGI and not AI. 

But let’s pause and ask ourselves, what do we mean by a transformational technology? For me it means that the technology must change the way we do things at a fundamental level. It must change human behavior. To take an analogy, a faster car is not a transformational technology - it just helps us do something faster without changing behavior. 

On the other hand, let’s look at the railways. There was a time in history when we intuitively knew that the railway networks were a transformational technology. We just didn’t know how. So everyone started to lay railway tracks to here, there, and nowhere. They were betting on someone somewhere figuring out what to do with these tracks. Many fell by the wayside, few survived.

The railways did bring about a fundamental change, but in the most unexpected aspect of our lives - time zones.

Until the railways came along there was no need to have a standard clock because we didn’t need timetables. 

The reason people adopted cars wasn’t because they could travel faster, it was because there was less horseshit on the roads. Without Hotmail and Lotus 1-2-3, the desktop computer would have never become ubiquitous. 

We know intuitively that AI is a transformational technology, but it is hubris to think we know where the transformation is going to be. My bet is that it’s going to be in the most unlikely of places.

As a technical writing team, we asked ourselves where AI can really transform our processes or bring in the most efficiency. The obvious answer would be to use it to generate content.

But is that the real bottleneck?

We looked at where we spend most of our time and discovered that it is spent waiting for information or making corrections after we get information that should have been provided earlier. We asked ourselves whether we can solve that piece of the puzzle. 

What if we were to provide our agent with all the information that the product team has about the feature? Link it to JIRA, Confluence, Slack, and Figma. Would the AI help reduce wait times or discovery of new information too late? Not surprisingly, AI is great at this. An AI agent that takes a JIRA ticket id as input and runs at regular intervals is a game changer when it comes to improving efficiency. Of course, this requires the product team - PMs and devs - to record all decisions on either JIRA or Slack. This ensures that even people who weren’t part of the conversation get the update. The agent’s output is a summary that contains an introduction to the feature, important dates, important contacts, changes suggested and accepted since the last time the agent was run, and links to related tickets, conversations and designs.

Such an agent is not useful only to the documentation team. Even sales, marketing, program management, release management, and design teams can use it to stay abreast of what is happening in the project. 

In conclusion, before you jump into using AI to help with your work, ask yourself what the real bottlenecks are and how AI can drive efficiencies there. More often than not, your bottlenecks are things you are not thinking of because they are not visible. While a Ferrari may not be transformational over a Suzuki Alto, it helps if we can get somewhere faster.