“People don’t know what they want until you show it to them.”
“You’ve got to start with the customer experience and work backwards to the technology.”
– Steve Jobs
Steve Jobs famously made these statements to describe his product development strategy at Apple. One gets the sense this is where AI is now, as it transitions from hyperscale compute buildout to actual enterprise deployment across firms, large and small. AI has so much potential, but right now, it is still a fairly blank slate. Do we use it for research, content creation, building agents to perform tasks, creating software code or entertaining us? The answer is likely all of the above. AI can be anything we want it to be. That is both the opportunity and the problem. It is built to keep you engaged and constantly feeling on the edge of greatness, in the paraphrased words of Morgan Housel. We are reaching a critical time where we individually or as part of an organization should decide if we are going to take the reins from AI and tell it, in the words of Steve Jobs, what we A) want it to be by building agents that take tasks off our plate, B) wait for somebody else to show us what we want or C) stay stuck in purgatory of constantly feeling on the edge of greatness, but not really progressing in any meaningful or enduring way. Let’s break down what each of those choices could look like.
A. Building Agents
In its simplest form, an AI agent is a custom piece of software that uses AI to perform a task with some level of autonomy. For example, you build an agent that, when prompted, either pre-scheduled or on-demand, will go and interpret a designated data set for a certain query that you seek. A radiology-focused agent might be trained to, every night, go to a designated file location and read every imaging file posted within the last 24 hours and output a description. Now, that agent must have been trained on perhaps tens of thousands of images and the correct diagnosis associated with each one to be effective, much the same way a human would be, but once it knows what it is looking for, it theoretically can conduct the task in a highly repetitive and autonomous fashion. So, a hospital system could build that agent itself, or wait for the marketplace to build it. The build vs. buy decision. Either way, the radiologist can come in the next day to review the findings and increase productivity. Everybody wins.
B. Wait for Someone Else to Show Us
If we don’t roll up our sleeves and crack the agent-building code ourselves, we will need to wait for the marketplace to put solutions in front of us. This is the historical software model that Steve Jobs and others developed so well. With good taste and deep knowledge of an industry workflow, software companies were able to build one-size-fits-all solutions that companies just had to plug into. They spent years building a competitive moat around their code and then just rolled it out to as many subscribers as possible at zero marginal cost. That model is likely to remain in the form of pre-packaged agentic solutions, but the competitive moat for these packaged solutions will probably not be as large as the legacy software companies once enjoyed. For that matter, the legacy software companies still are likely to enjoy a sort of data record-of-truth status for some time. But it remains incumbent upon them to use their legacy status to introduce the agent solution option into the mix, or else the market will do it for them.
C. Stuck at the Edge of Greatness
Everybody wants to feel like they are great at what they do. AI knows this and is programmed to reinforce engagement by making you feel like your ideas are promising. Social media operates similarly. There is a risk in this echo chamber, and it is important to remain in the real world regarding what is measurably viable from a productivity and margin standpoint and what is not. There are stories of companies blowing through their annual AI spend budgets in mere months because everything looks promising in AI’s eyes. This is the hook. AI encourages engagement all the while the meter is running.
In summary, we live in unbelievable times; AI is changing so much and so quickly. It is very important, however, to know where the conflicts of interest exist so that you can navigate the best possible solutions and use cases for AI in your businesses and daily lives.