If it were 2023, you'd be wise to consider your AI readiness before diving in. Today? Not so much.

Two problems make “AI readiness” irrelevant by now

For one thing, there's the sheer irrelevance of your Readiness. AI is happening. It is forced upon you from the outside, whether you like it or not, so the universe doesn't care if you are ready. You have to act anyway. In most cases, asking about AI readiness is just dithering.

By the way, this isn’t unique to AI. A basic business building law might be “The Law of Knife’s Edge”:

The Law of Knife’s Edge:

When to launch new efforts is always the balance of two unbendable forces: Readiness and Necessity.

A new, high-risk effort will launch at the earlier of perceived inside-out readiness and observed outside-in requirement, weighted based on the risks of being too late or too early.

If nothing forces a launch and the risk of being too early is high to executives, the initiative will never launch.

(Yeah, I need to write that up separately.)

Anyway, the risk or cost of being too late the party of the AI era is high, and the outside forces that eventually demand that you get going already are also high. In the end, you must launch, and your readiness is utterly irrelevant to the rest of the universe.

For another thing, Readiness is not a finite, defined thing, no matter that consultants' assessments want to make you think so, and you certainly would like to believe it.

As new capabilities appear, you will, by necessity, be unprepared for them because they never existed before. Right now, since those changes happen from a number of providers, both Frontier Labs and specialized companies, you can't possibly keep up with all of them, even though each of them might only release something new every few months. It gets worse because changes pile on top of new changes. As soon as one generation of AI capability has become routine, a totally new generation becomes possible, and the cycle repeats. Even if you can keep up with the linear change of incremental capabilities, you will quickly fail with the exponential change of new generations of AI-related capabilities.

Beyond that pace of change, there's also the problem of fractal-like nuance. Readiness is specific to your specific situation, the individual teams and jobs, and the specific way you have set up your apps, infrastructure, org design, and capabilities. It is not a binary on/off thing. We all will be getting massively better over time.

In the face of that, "readiness" is an arbitrary point in a three-dimensional space. You can't possibly know what is "ready enough” with any kind of objectivity. That means your company's executives will debate it endlessly and simply stall the process.

At that point, psychology makes it worse. It's well established (look it up, very interesting) that adults have a terrible time with the two capabilities you really need to succeed here:

  1. Being wrong
  2. Admitting that they don't know

This gets worse the older we get, the higher our titles, and the higher our education, all of which are usually markers for knowing the right answer and not making mistakes. Plus, we have accumulated a series of results in our lives that came from not being wrong. This kind of gathering and over-engineering is perfectly natural, but the universe ultimately doesn't care.

Yes, you do have alternatives of course

A lot of ink has been spilled about how to start on your AI efforts or take them to the next level, away from the comfortable rut that you've possibly fallen into. But squint and you see that the bottom line always is:

Learn by making, get started, find something to do, and do it in the scope of what matters to your world.

"Great," you'll say, "but it's not as easy.” The question is where to start, and that is true. That's where a lot of the advice gets contradictory, and the relevance to you is a lot harder to figure out. So far, I've only found two rules that hold universally:

Certain wins are always nice

This is the very practical contribution and toolkit of Tobias Zwingmann. (Check out his work.)

In short, anything you try to do with AI takes longer than you think, just as has always been true for any kind of new initiative, even before AI. It's a safe bet that anything beyond the trivial will take you 10 hours. A good way of thinking about whether it's worth that is to be clear on and write down whether the time investment of that is worth the while. The specifics will vary based on the job and the pay of the people involved, but a rule of thumb that Zwingmann suggests is that anything worth a meaningful AI effort should be worth $10,000 per person per year in benefits.

So ask: will this effort clearly save us $10,000 or earn the equivalent in revenue? If so, go for it.

These are your “no-regrets” moves, if you use that language.

Each of these efforts will be small individually, but if done right and coordinated somewhat, they will pile up and feed on each other, getting you greater and greater benefits over time. Some people like to use the "1% better every day" rule. That's a little bit of a rule of thumb at best, but it makes the point.

Must-dos on your economic leverage points get you real progress

There's another kind of effort that makes sense, but let's first talk about what doesn't work.

The more simplistic prioritization efforts may use pretty language but boil down to an "effort versus impact" prioritization. This often appears in the form of a 2x2. The real problem with that is that it only works if you have a massive number of options and ultimately don't care which ones you pick. The point of the effort is simply to break through to figuring out a starting point.

In any other situation, that kind of prioritization always fails because it turns out that impactful things are also hard, and easy things also don't make that much of a difference. In other words, the 2x2 collapses to a line. That doesn't tell you anything.

But that problem also tells us how to break through it. All you need to do is not worry about the effort. Easier said than done, of course, but the way you achieve that is this:

Bring AI to something that is both high impact and a true must-do for your business.

Ideally, pick the one with the highest ability to correct for errors and problems with the AI early on. That's less about “low risk” and more about “high recoverability”.

To find what is a must-do, identify the 3-10 end-to-end workflows that generate the most value inside or outside your business (whatever your personal scope happens to be). Those will be the ones that include the economic leverage points, to borrow McKinsey’s term. You might also call this your core “value generators” or something else. Independent of term, those are the topics that define the essence of your value contribution to your company.

Some of those will be so carefully or precariously built that any disruption will instantly destroy value. Those are probably not the ones to start with. When you're learning, things will go wrong. Instead, you want to pick the end-to-end processes that have some give in them so that you can compensate for the time, effort, and issues it takes to learn. If you're really lucky, you might even be able to parallel-path.

And, to state the obvious, of course there is nothing that perfectly achieves this. Otherwise, you would probably have been seen as being very wasteful in the past. That's why, further above, I pointed out that our readiness is irrelevant. There are certainly such efforts that are less fragile than others, and you simply need to pick your least fragile one, or the one with your best team, or the one you otherwise think you can compensate for when things inevitably go wrong (when you inevitably learn about things that you haven't known about before).

For example, in my startup work, an economic leverage point is the ability to get the attention of overly-busy industrial plant managers, despite all else they have going on. And an end-to-end process that helps us gain that leverage is “trend signal ingestion to thought leadership output”. That nerd speak simply means that we learn what problems plant managers currently worry about most, filter down to the ones where we can help, and turn our ability to help into sales assets and marketing materials.” This is also a must-do because plant manager attention is an always-changing must-have to unlock sales conversations and potential clients.

So I automated that work. It’s hard. And AI is simply not as good as promised for many of its steps. But I haven’t stopped. Gradually, I’m getting there, building us a plant manager attention engine that is as automated as currently feasible and helpful to our team. It’s starting to look pretty good by now.

That example points out to you that when you have a must-do effort, you won't automate it in one fell swoop, and you may not want to automate it ever. You might want to lean the other direction of "augmenting" humans with AI rather than replacing them, and that's certainly true here. But if something honestly is an effort that you must succeed at, you will keep chipping away, even when you keep failing, even when you keep not knowing the answer, and even though that is uncomfortable. Your standard for success needs to be ongoing progress and eventual high value, not instant results. Even better if you can interweave the new way with your current process so it can assist people even as they do their core daily job, rather than distracting them from it.

You can create your own “AI readiness”

What matters most about these two ways to build AI readiness through doing rather than preparing (even when you need to take the plunge and get going before you're ready) is that they build tremendous power and control over your fortune in you and your team.

Even worse than the worry of embarrassment through failing or not knowing is the feeling of not being in control of your own life. That's where you get individual burnout and collective loss of morale, but when you have a way to internalize the decision of where you start, you are no longer a victim to the vagaries of AI frontier labs or others. You can be in charge of your own fate.

As a leader, all you need to do is be sure to define success correctly:

  • Making sure that your initial efforts have sufficient clarity, low stakes, and near an ROI
  • That you and the team will keep going and celebrate progress, even as you struggle, and even that is under your control in the way you lead

So don’t worry. Create your own AI readiness.