Efficiency with AI
Merriam Webster defines “efficiency” to be “effective operation as measured by a comparison of production with cost (as in energy, time, and money)”.
Read that definition again. Let it sit with you for a minute. What stands out to you?
For me, the key word in this definition of efficiency is “effective”. Efficiency is not simply a comparison of production with cost, or ROI. In order to be efficient, your operation must be effective. If you have a low-cost, high-volume widget factory but the widgets produced do not work, your widget factory is not efficient because it is not effective.
AI has a great capacity for efficiency, but how do we ensure effectiveness? Expert human judgement.
When evaluating how I apply AI, I consider 2 questions:
Do I get value from doing this task myself?
Can AI make this task more efficient?
If the answer to question 1 is “Yes”, then stop and keep doing that task yourself. Do no automate away high-value tasks.
If the answer to question 1 is “No”, then move on to question 2. If the answer is “Yes”, go for it! If the answer is “No”, stop. Sometimes the answer is “I don’t know”, for which I suggest deciding how much time you want to invest in testing AI efficiency.
Let me demonstrate with a simple personal example. I am a time blocker. Every morning, I would go through my tasks and add time blocks to my calendar as a plan to get the work done. Do I get value from doing this task myself? Not really, I just need to have a plan in my calendar. Can AI make this task more efficient? I wasn’t sure, so I blocked 2 hours to build a time blocker agent with Claude. The agent pulls my calendar events and tasks, then generates task time blocks for the day between existing events. I, the human expert, decide if that plan is acceptable. Are there days when I decide to change the plan because of information Claude doesn’t have, like my morning mood? Absolutely, but most days AI’s plan is good enough. My time blocker AI agent meets the efficiency bar because it successfully creates a plan for me every day so I spend less time planning my day, and more time doing work.
You might now be thinking “well, that’s nice, Susan, but I’m making decisions for a company at-scale”. I’ve been there too, and the framework holds if you and your colleagues honestly answer the two questions. In my experience, companies right now are skipping the Value question and going right to the AI efficiency question. In my opinion, it is a mistake to automate away your high value “secret sauce”. Do not turn your human experts into AI reviewers. Figure out how to use AI to make all your human experts more efficient. Let them review the low-value tasks, freeing up more time for high-value tasks.
At a large company scale, there will also be a balance between individual efficiency and company efficiency. It is important to ensure the people implementing AI technology and operations understand the system effects. For example, generating more copy in multiple languages with AI may make your growth marketer more efficient, but may create a bottleneck in the Brand and/or Legal teams with approval reviews. Releasing copy that doesn’t meet the Brand and Legal requirements will not be effective. You now have broken widgets.
I am energized by finding ways to use AI for efficiency, but don’t forget to make sure AI production is effective. And please don’t automate away the things that bring you value. AI should be enriching your life, not taking away from it.