Practical Automation Framework

AI should work for us. AI is disruptive. It disrupts in positive and negative ways. We are worried about AI's potential to replace people while simultaneously being excited about it's potential to free our attention to valuable tasks. Many of us are conflicted about using the technology, for fear of accelerating a replacement trend. I've spent a significant amount of time experimenting with AI automation in the past couple of months. I'm convinced that, like all of our technology, we want these tools to work for us, not to outsource our thinking and learning.

We should augment our people instead of replacing them. If this is the goal, we should NOT automate out all our repetitive tasks. By now, we all know that struggle fosters growth and builds strength. We know that we need to be more comfortable with boredom. Creating AI bots that act as junior employees will rob us of the opportunity to bring and up skill team members that will become the next generation of leaders.

Boredom and frustration are noisy signals. Boring tasks can teach you how things work or give you new information. Hard tasks train your mind and help you build new skills. Investing in hard work now will pay off in the long-term. But it does suck sometimes. We gain speed and ease through automation. It is tempting, but we avoid the discomfort at the expense of growth and precision.

We SHOULD automate some things that hoard our attention. Busywork may start off as a learning opportunity, but the returns diminish quickly. We need our attention free to tasks that require human judgement and our expertise. Some repetitive tasks benefit most from automation, others from developing systems. Still others are so infrequent and complicated that they will require creative problem-solving with novel approaches. I've created the following framework to help me think about these distinctions.

Automation Framework Chart

This 2x2 framework of repetition and learning/growth potential can help to guide in thinking about whether a task is worth automating.

Expertise

The lower-right quadrant is where the most novel of activities occur. Things like negotiations, crisis decision-making, and handling sensitive personal issues would fall here. This is where you benefit from leadership or a variety of expertise to devise novel, creative solutions. You want to document the decision-making and process for reflection and future use, but you almost certainly wouldn't devise an automated workflow. It is foolish to trust AI, at least in its current form, over human expertise and creativity. In this quadrant, skilled humans rule.

Discipline

The lower-left quadrant also indicates a low level of repetition, but it is for tasks that don't yield much in learning outcomes. These would be things like filling out one-off forms or updating online registrations. There may be some learning that takes place, likely for junior employees, but they diminish extremely rapidly. The tedium of these tasks make automation tempting, but they occur too rarely to invest the time. Documenting process, roadblocks, and tips is helpful in case these items come up again in the future, but the payoff in investment for automation is low.

Systematize

The upper-right quadrant of high repetition and high learning is a good spot for systems. Things like project postmortems, strategy sessions, or marketing reviews would fall into this category. Focus on developing playbooks, SOPs, frameworks, facilitation guides, and the like, to help focus participants in the most aligned ways. These are not set in stone, so adapt as needed to fit goals. Human expertise and judgement reign in this quadrant. Some automations may be appropriate, maybe in scheduling or the packaging and sharing of materials, but you should avoid automating research (or otherwise relying on AI research) to inform these activities until a level of expertise is achieved within the team that allows them to effectively critique and iterate outputs.

Automation

The upper-left quadrant is for repeated tasks with little to no learning potential. The first few times someone completes them, they will be building some knowledge and skill, but returns diminish almost immediately. Things like formatting invoices, completing timesheets, entering expenses, and other routine data entry, fall in here. The goal is not to replace the junior employee, but to get them the basic knowledge needed to get to higher value contributions.

-----

The ability to automate depends on your tool stack and skill level of your team. Almost all solutions require degrees of technical knowledge. Automation could be sending some instructions to a script that executes a pre-determined workflow automatically. It could also involve AI "agents" that are given some autonomy in your systems. I've shared some insights that I've gained into this (see my post on using AI agents).

As a leader, your job is not to avoid all difficulty. It is possible, however, to reduce friction while preserving learning. I think this framework is useful when making decisions around automation that augments rather than replaces.