The Product Engineering Evolution, Part 2: Navigating the AI Talent Dilemma
Build, borrow, or buy?
Look ahead three years. The landscape of software development is fundamentally different. Anyone can build software independently with the assistance of ubiquitous AI tools. The software we consume is ephemeral and deeply personalized, suited to our specific individual needs and unique situations, used and discarded once it has provided value to its user.
Software development no longer depends on teams of highly specialized people separated by their expertise into distinct roles–product manager, designer, architect, developer, quality engineer, etc. Instead, humans are generalists who rely on the superior expertise of AI tools that perform complex planning, develop code, and execute sophisticated go-to-market strategies.
Development methodologies like Agile are no longer necessary because the complicated workflows designed to manage handoffs across separate product management and engineering teams, and the need to alleviate the inevitable gaps when transferring knowledge across disciplines are no longer necessary. Tools like Jira are obsolete; its epics, stories, and dashboards are replaced by working prototype code built by product managers who no longer waste long hours writing user stories no one fully grasps and filling backlogs no team can realistically achieve.
A self-funded company with a single employee is the market’s darling after reaching $1B in annual revenue and celebrating a successful IPO1.
As we wrote in the kickoff to this series,?player 3?has entered the game. The tools altering the game forever aren’t theoretical; they are here.
As a product leader peering out to the horizon of 2027, are you freaking out?
We aren’t, and here’s why.
While it’s tempting to speculate that AI tools obviate the need for humans to accomplish work, what they are actually doing is changing the nature of the work we will do. You’ll still have a product organization organized into teams, only they’ll be accomplishing vastly more than today’s teams. The auto industry automated car production, and robots took over repetitive manual tasks, but ultimately automation created as many jobs as it displaced–just with very different kinds of jobs.
The pace of discovery, innovation, and growth will astonish AI cuspers (you know, people who remember the world before AI) and one day will mildly disappoint AI natives (you know, the ones who don’t).
If you can see a possible future, you can work backward from it to do the things that make that possible future inevitable.
What’s the most obvious problem to solve on the path to our hypothetical future? Your team needs AI skills. Where and how to acquire them?
There are three options: build the skills within your existing teams, acqui-hire the talent, or a hybrid of the two.
Super-skilling
Acquiring a new skill is onerous and time-consuming for humans. Learning happens individually, and it’s tough to distribute that new skill across a team.
For AIs, not so.
领英推荐
Mo Gawdat?describes the way AIs acquire skills?as a sort of shared neuroplasticity–when one machine learns a skill, all machines learn that skill.
Machines already know how to write code, create marketing copy, and persuade people to buy your products. They are learning from one another and?getting better at it every day.
We’ve hypothesized a possible future in which product managers develop code to create prototypes instead of user stories that are handed off to designers and developers. To reach that end state, ‘super-skill’ your product managers to use AI tools in novel and innovative ways.
This ‘super-skilling’ requires teaching product managers at least two new skills.
The first skill is initiating and holding conversations that direct an LLM to comprehend your vision, understand your requirements, and help you create an implementation plan.
The second skill is pair programming with an AI.
For now, LLMs are like hyperintelligent coworkers who forget every conversation you have with them. Interacting with them is challenging. Once an AI has a context and a task, it executes with startling speed: outcomes that used to take days or weeks appear in seconds.
Budget time and money to build what we’ll call Learning Engines. Use these to find the people in your organization who are already good at using AI and tap them to spread that goodness. For a short time, teaching becomes their full-time job. Even if the critical work they leave behind slows, spreading their AI knowledge and skills is more important. Shifting your team’s priorities from product work to teaching and acquiring AI skills is likely your most important decision over the next year.
How to Build Learning Engines
First, set a goal and establish a timeframe: find your team's most gifted and innovative AI users in the next thirty days.
Answer these questions:
Who in your organization is adept at using AI? How well can the best users enable other team members to become better AI users? Which applications of AI will be most beneficial to the organization?
An AI Learning Engine combines the natural curiosity of humans with the elastic intelligence of AIs. Your first two Learning Engines are thirty-day experiments to see if you can:
Structuring Your Experiment
The experiment should help you learn quickly and iterate...