Ex-Google Program Manager: Use AI to Build Long-Term Growth, Not Chase Short-Term Savings
Invest in people, not cuts: Use AI to augment top talent and expand capacity, which lifts revenue faster than chasing capped cost savings.
AI rewrites what’s possible: Large migrations and complex targeting can be automated, but you still need strong leaders and engineers because AI amplifies good or bad design.
Focus on adoption, not headlines: Build AI literacy, embed tools into daily workflows, and share real success stories to drive meaningful, lasting transformation.
In this conversation, Jonathan shares how AI is reshaping leadership and org design — and why you should be using AI to uplevel your employees instead of using it to cut costs.
From big tech to startup: How Jonathan Conradt is rethinking leadership with AI
I've been blessed to work for Amazon, eBay, Google, and Microsoft on products people use every day.
Keep Reading—and Keep Leading Smarter
Create a free account to finish this piece and join a community of forward-thinking leaders unlocking tools, playbooks, and insights for thriving in the age of AI.
Already a member? Log In
I was the engineering program manager for Chromebooks — which my grandkids now use at school — as well as the Mac and Linux versions of the Chrome browser. At eBay, I was Director over product for merchandising and marketing.
At Amazon, I moved from leading marketing for all Amazon devices to building AI solutions, and then to training Amazon VPs and Directors on AI. Seeing a greenfield opportunity with AI and leadership, I left Amazon in 2024 to build Shiftwell AI, a startup around career and leadership development with a heavy AI component.
Building Shiftwell AI to help leaders grow their people
The insight we had at Shiftwell AI is that managers are overwhelmed. They rarely have the time, and often lack the expertise, to help their direct reports develop their leadership skills and grow their careers.
We expect that AI will lead companies to flatten their organizational structures as tools come online to summarize the flood of information leaders receive each day. Office productivity improvements will allow a manager to have more direct reports, but those direct reports still won't have the support they need to grow.
Shiftwell AI is a replacement for the mass marketing employee engagement surveys that have failed to deliver on productivity and retention. Applying AI to an engagement survey takes a bad approach and makes it worse. We are taking a new approach enabled by AI and decades of research that delivers for employees and managers the insights and tools they need to thrive.
We expect every employee to have AI teammates and partners who assist their work and automate tasks. We see the best companies using AI to also build their people. Because the companies that are able to attract, develop, and maintain top talent will thrive.
How AI is transforming leadership and organizational design
A key lesson I taught Amazon leaders was that their years of experience and intuition about what is impossible or too expensive is now invalid.
For example, Amazon went through a painful transition from Perl and C to a largely Java codebase. This transition was very expensive and took a long time. But with AI, their move from an old version of Java to the latest version of Java was almost completely automated. This quantitative leap in productivity has real impacts on how we think about workforce composition and scale.
In my view, there are three ways to respond to the AI economy:
- Ignore AI and hope it goes away, a strategy that Sears appeared to take with the Internet.
- Rush headlong into it and quickly reduce your workforce with the expectation that AI's increases in productivity will allow you to maintain the same level of delivery at a higher profit margin. This has a short-term benefit to the stock price, and ultimately leaves the company with the same capacity as before but at lower cost.
- Retain top talent and enhance them with AI; so they can do more, faster.
I’ve always argued for that third path.
In every engineering project I have ever worked on, we needed to prioritize the customer demands and then draw a line at our capacity to deliver. We always left some customer demands unmet because we had finite resources.
Companies that retain their talented and experienced staff and increase their productivity through AI will deliver more customer benefits than those that go for the short-term stock bump.
This approach does not immediately reduce costs, but it does quickly increase revenue as a company accelerates past its competitors. In my experience, there is enormous headroom for increasing revenue — but cost savings are always capped.
How AI transformed Amazon’s email marketing — and how they accelerated adoption
Here's another example. Amazon relied on hundreds of marketing managers to manually configure targeting parameters for marketing campaigns. This labor-intensive process involved brainstorming which criteria and behaviors mattered, and then hundreds of hours of report generation and analysis. Teams were reluctant to try new things or run multiple experiments because the investment was too high.
I was fortunate to meet an AI scientist who had the idea of using AI to combine hundreds of measures into a prediction of who was going to purchase in a category in the next two weeks, then who was most likely to respond positively to an email.
I leaned into this and created experiments with Amazon's device campaigns to validate and refine the model.
At first, marketing managers were wary of a solution that claimed to know their customers better than they did. AI wasn't a common tool and was generally misunderstood at the time. So I created a class for marketing managers to explain the basics of the model, and we spoke at internal conferences.
Ultimately, by making it available in their existing tools and widely publishing success stories, we were able to gain momentum. As teams shared their successes and employees moved between roles in the company, the adoption rate grew. Within 18 months, every team was using it, and all new country launches included the requirement to field the model as soon as customer data was available.
And in the end, all Amazon email campaigns worldwide were required to use this model in their targeting. It dramatically reduced the effort to field new campaigns, which encouraged teams to experiment with multiple campaigns, resulting in new insights into customer needs and improved sales.
The big insight for us was that humans would never have considered nuanced approaches involving hundreds (later thousands) of variables. But by applying AI to the problem, we could do more for customers with the same or less effort — as long as our teams adopted it.
Why guardrails are needed when software teams use AI
As the CTO of a startup, it is on my shoulders to write the code.
AI code tools like GitHub Copilot in VS Code, plus Ollama running locally with Open Web UI, frequently shock me with their ability to seemingly read my mind and suggest large code blocks that are exactly what I was about to write.
At the same time, the AI is happy to continue down a terrible path and amplify mistakes in architecture or design.
That's why having well-trained and experienced software engineers is vital to producing the right results. AI can accelerate and build on great design or bad design. It will amplify the level of talent in your organization like a carnival funhouse mirror.
If you invest in top talent, your AI tools will amplify their talents and produce great work faster. If you have failed to develop your people or have hired poorly, AI will assist your workforce in producing poor solutions quickly.
AI can accelerate and build on great design or bad design. It will amplify the level of talent in your organization like a carnival funhouse mirror.
Why readability matters more than ever in the age of AI-generated code
At Shiftwell AI, we have chosen Go as our programming language of choice. It is well supported by all AI tools. It is also very readable.
I was blessed to observe the Google Chrome team engineers as an engineering PM. The code review process, still visible to everyone via the Chromium project, was an excellent environment for learning, training, and improving code. A key step in the review was increasing the readability of the code. Could an engineer read and understand the solution within a reasonable amount of time?
The Go language team at Google took this to heart and built some readability guidelines into the language. While languages like Rust or C++ can produce faster or smaller solutions, the Go solution is generally the most readable.
We expect to have fewer engineers than we would have had ten years ago, with as much or more code than a similarly sized startup would have had. Both the AI tools and this highly readable language help with this. Because, when an AI proposes code, the engineer has to read and understand the code.
I suspect that companies will soon move toward languages that are easier to read and understand to avoid AI tool mistakes.
How Shiftwell AI uses AI tools across development, marketing, and support
As I mentioned, I use GitHub CoPilot in VS Code extensively. It has made it much faster to bootstrap a good set of tests for my code. It's impressive how frequently it generates smart tests that are easy for me to expand with scenarios I want covered.
As an organization, we also use it to brainstorm marketing messages and things like elevator pitches. And when we get stuck on how we want to say something, we turn to AI to generate additional ideas.
We fully expect AI to be integral to our customer support flow. We plan to support 35+ languages with an English-first support team. Providing native-level customer service will only be possible with AI.
The hidden risks of AI consolidation and how smaller companies can thrive
AI appears to be driving consolidation, as larger companies that have high-quality training data have an advantage in training and applying AI to their businesses. What will happen to the mom-and-pop companies?
Historically, photography had a similar impact on the art world. A small number of camera, film, and developing companies could outproduce all of the artists in the world with a high-quality product with incredible fidelity. This made every camera owner an artist, albeit normally a poor artist.
At the time, this led many to expect that the value of paintings would collapse and artists would cease to exist. They didn't.
As AI leads to consolidation and diversity disappears, products will normalize and become more generic. This is the market opportunity for small and medium-sized businesses that can provide unique and bespoke solutions. Small companies that focus on their customers and take the time to understand their needs deeply can still thrive in an AI economy.
The most important advice for leaders navigating the AI economy
My advice for this moment in time is to focus on using productivity enhancements from AI to deliver more for customers, instead of cost savings to drive profits.
This is an opportunity to accelerate past your competitors who make the mistake of stripping talent and experience out of their company in the pursuit of short-term gains. If an employee who once delivered 7X their salary in revenue can deliver 20X their salary, what fool would sacrifice that revenue to save their salary?
With that said, I think AI will have a surprising impact on Director- and VP-level jobs as it reduces the cognitive load of leadership, allowing fewer people to do the work.
The best companies will drive a decrease in administrative expenses and reallocate these savings to investments in employees, directly driving customer value.
AI will have a surprising impact on Director and VP level jobs as AI reduces the cognitive load of leadership.
Follow along
You can follow Jonathan’s work on LinkedIn and explore his company, Shiftwell AI, where they’re building the future of leadership development with AI at its core.
More expert interviews to come on People Managing People!