Beyond 'Violent Task Churn': Thoughts towards a Human-first AI Transformation

JPMorgan's recent prediction of "violent task churn" from AI implementation was definitely interesting and worth taking note of. Jacob Manoukian, JPMorgan's U.S. head of investment strategy, projects that AI will follow the same trajectory as previous technological revolutions: "Violent task churn, then broad productivity growth." While I agree with JPMorgan's optimistic long-term outlook, I believe we can do better than accepting the "violent" part as inevitable.

Having spent the last 20+ years scaling organizations through digital transformations, I've learned that how you manage technological transitions matters just as much as the technology itself. The difference between "violent churn" and strategic transformation often comes down to leadership approach and organizational readiness.

The Stakes Are Higher Than Historical Precedent

While JPMorgan draws parallels to steam engines and electricity, there's a critical difference with AI: speed and scope. JPMorgan estimates AI will drive productivity gains in less than seven years, compared to 61 years for steam engines, 32 for electricity, and 15 for computers. This acceleration means we have less time to adapt, but also less time for society to naturally absorb displaced workers.

In my experience with building companies, I've seen firsthand how rapid technological implementation can either energize teams or create organizational trauma that takes years to recover from. The companies that thrive don't just implement technology—they orchestrate human-centered transformation. Those who simply spout "we're now an AI company, and version X of ourselves" and then proceed to cut staff in half just don't get it. Or another favorite is then just checking who is using their chatGPT account more than others. Just like hours, more doesn't necessarily mean better productivity.

Lessons from the Field: Overhauling everything

At the point where the GPT-hype cycle began, I started regular evaluations of those particular models in relation to the day-to-day work that was going on. These were more hyper-focused efforts, identifying a particular use-case and then pushing the boundaries. Eventually, the models reached a point where they could do more and more within appropriate boundaries, but the real key was reworking the entire flows of work to take advantage of the technology.

The results were telling: some activities that took weeks could be done within an hour. Baseline implementations in the product no longer waited on the coding, but waited purely on the admin of setting up accounts and paperwork. And the overall quality of work was improving piece by piece.

A Framework for AI Transformation

This is the basics of how I went about it. Now there are a couple of critical things to remember when reading this.

• First, I prefer smaller team sizes for productivity, and a higher quality of person in the role. Prior to AI, this was beneficial in a lot of ways, but the benefits are exponential when then applying AI. If you go cheap or hire only yes-men, then be prepared for the amount of crap produced to go up.
• Strong guidelines were in place for anything around the product and technology. I specifically use the word guidelines. By having placed strong people in roles with guidelines, they were able to build a solid base of valuable information that can be used to amplify the results of AI. The AI is great at replicating things, so make sure what it's replicating is already solid.

And with that, let's get into it:

1. Task Audit Before Technology Deployment

Before implementing anything, conduct a comprehensive audit of current roles and then what those people are doing on a day-to-day. You may have done something like flow diagrams of processes already for getting from idea to production. Categorize tasks into three buckets:

• AI-Replaceable: Routine, rule-based activities with clear inputs/outputs. These REALLY need to have a solid base to work off of.
• AI-Augmentable: Complex tasks where AI can enhance human capabilities. I put my critical paths for Product as well as then Engineering tasks in here. I'll provide a more in-depth example in another article.
• Human-Essential: Activities requiring creativity, emotional intelligence, or complex judgment.

This audit should involve the people actually performing the work. The paper is always a bit off from the reality. Sometimes you have a unique person in a role that is operating as a hub for the other team members, or just filling a gap that wasn't written down.

2. Design Transitions, Not Replacement Plans

For each role affected by AI implementation, create specific pathways that leverage existing knowledge while building new capabilities. Remember that change evokes a loss response in most cases and the most successful transitions happen when people see opportunity, not just change.

An example of this would be generally how we treated the Software Engineering role. It became a "craftsman" role, where the AI became more like a tool for making both the bigger cuts and refinements under the hands of the engineer. The focus of the work transformed into deeper thinking, creating architects out of everyone. While we continued to work towards certain deliverable goals, we piece for piece started to push the raw implementation work down into the AI and people moved into these new roles.

3. Implement Co-Creation, Not Consultation

Rather than developing AI solutions in isolation and then "consulting" with affected teams, involve them in the design process. Nothing magical, but generally better results than a top-down "just do what I say" approach.

Whenever we went about overhauling a flow or role, it was done in tandem with someone performing the work. When I was changing something about my work, I also involved someone else who was doing something similar to get another perspective on it. That reduction from weeks to an hour I mentioned earlier was a result of this work in tandem.

4. Measure Impact Alongside Technical Metrics

Track not just productivity gains and cost savings, but also employee engagement and skill development. Create feedback loops that allow you to adjust implementation based on the impact your seeking. Again, labelling yourself simply as AI won't cut it, and if your doing something silly like tracking activity instead of outcomes, you're going to make bad decisions.

The Economic Reality Check

JPMorgan's optimism about job creation isn't unfounded. As Manoukian notes, "businesses will likely reinvest some of the savings derived from AI into new growth areas" and "expect more hiring from companies that build software applications and data infrastructure as well as those that integrate AI tools into workflows and systems."

However, this transition isn't automatic. It requires intentional effort from leadership to ensure that AI-driven productivity gains are realized and then reinvested into the business.

I've been fortunate over the years to be part of great digital transformations as well as the "jerk the wheel", overreaction to simply a trend resulting in gutting a team's morale and inevitably killing the company's future.

Leadership Imperatives for the AI Transition

The companies that will thrive in this transition are those whose leaders recognize that AI transformation is fundamentally an organizational development challenge with a technology component, not the reverse.

Start with Strategy, Not Software: Before selecting AI tools, understand how work currently flows through your organization and where AI can create the most value with the least human disruption.

Invest in Your People Early: Incorporating them into the entire process both ramped their expertise quicker and made the change more sticky. Do treat it as an investment though. If you're spraying & praying your way by just buying a bunch of tools, then hope that lotto ticket works out for you.

Create Psychological Safety: At this stage, it should be a no-brainer. Fear-based environments lead to poor adoption. Generally, you have created a lower quality team, so don't expect much there.

Plan for the Long Game: Your size ultimately dictates how quickly you can do this. Even in a small team though, you may find that you need 6+ months to get to that initial place where you can ramp quickly. Once you've made the breakthrough for your org, then it can really move. Until then you may need a few targeted starts to get it right.

Beyond the Churn: Building Antifragile Organizations

The goal isn't to avoid all disruption—it's to build organizations that become stronger through change. In my experience, nagivating change properly means that you can really get an org into a healthier, more resilient spot.

The alternative to "violent task churn" isn't resisting AI adoption—it's approaching it with the drive and approach to any major business transformation. This means considering human capital development as seriously as technology selection, measuring success by organizational health as much as efficiency gains, and recognizing that sustainable productivity improvements come from building up the stack with AI: Low-level, repetitive, soul-sucking tasks get pushed into the AI first, and you build upon that to tackle more and more complexity.

Questions for yourself

As you consider your AI transformation:

How are you currently measuring and tracking the human impact of technology implementations alongside business metrics?

What mechanisms do you have in place to identify and develop new roles that leverage both AI capabilities and existing employee expertise?

How might your organization's approach to AI implementation serve as a competitive advantage in attracting and retaining the right talent?

The transformation ahead is inevitable, but the human cost isn't. With thoughtful leadership and strategic implementation, we can capture AI's productivity promise while building stronger, more adaptable organizations. The choice between "violent churn" and strategic evolution is ours to make.