PM #009: Product Management In The Age of GPT and AI Agents
Get fit with GPT-based technology or you’ll get sidelined sooner than you think.
I have installed AutoGPT a few days ago (link to GitHub repository) and — like everyone else who has tried it — I was completely blown away.
An immediate question that comes to mind is how will it change product management and the product building process in general?
We are all witnessing an exponential technology growth curve and everyone with access to the internet understands that the world is never going to be again the way it was just 6 months ago.
Continuing this pace, the world will again look very different in 6 months from now.
To analyze the influence of AI on product management we can look at what is already possible today regarding the different stages of the product building process: discovery, definition, development, and delivery.
For each of these stages tasks can generally be grouped into
- finite tasks: a one-time task that needs to be executed, ideally, as fast as possible.
- repetitive or iteratively optimizing tasks: e.g. the continuous optimization of a conversion funnel.
A lot of finite tasks may be automated or sped up by ChatGPT (Plus).
Examples already possible today:
- Carry out a consolidated market research
- Create a summarized competitor analysis
- Analyze, group and categorize customer feedback
- Ideate and create design drafts (e.g. with Uizard autodesigner, Dall-e, Midjourney) for yourself or to communicate with your design team
- Generate edge case analysis for a new feature or product
It is important to realize that until now each of these tasks took you several days or even weeks to carry out (depending on the depth) or you would have supporting roles who worked on them for a while.
In other words, work that until now took, say, 10 days, can be done in 1–2 days.
That’s a hefty 5–10X acceleration of work, output and ultimately speed-to-delivery.
For the second category of tasks which are iteratively optimizing the gain is even more mind-blowing.
If we set up AI agents such as AutoGPT or BabyAGI we can autonomously execute on creating
- weekly or monthly customer feedback analysis
- weekly crisp analysis of analytics regarding your product, website, engagement metrics etc.
Then for instance, have an agent combining the above to draw conclusions and
- suggest detailed A/B tests for your product
- then write and execute the A/B test
This way you can create a team of agents that optimizes entire parts of your conversion funnel.
Side note:
As ChatGPT, GPT-4 in general and AutoGTP are also able to speed up the coding aspect of product, the overall increase in speed and delivery is hard to grasp.
See more examples here:
- Twitter following #AutoGPT
- Twitter account of Significant Gravitas (creater of AutoGPT)
- Twitter following #BabyAGI
- How to use ChatGPT in your PM work by Lenny Rachitsky
How will AI affect the job market for product managers?
The obvious is that things can be built now much faster and with fewer resources.
On the other hand more people can build stuff with less knowledge but more help from AI systems.
Listen to famous venture capitalist Chamath Palihapitiya here (jump to minute 11 in the video):
It’s not clear to me how you start a company anymore. I don’t understand why you would have a 40 or 50 person company to try to get to an MVP. I think you can do that with 3 or 4 people (…).
Thus, you’ll need fewer people to get stuff done.
Obviously, you’ll need the people who know how to leverage AI technology.
You as a PM and your engineers can all 10x yourselves and thus 3–4 people can do what today 30–40 people are able to do.
I have engineers in my network who tell me today that over the last 6 months their productivity increased by 30%.
And that’s while the tools are still evolving and the engineers are still learning and integrating them into their workflow.
Let that sink in for a moment.
Same goes for PM tasks.
One consequence of the AI-era is that large companies will suddenly have to compete with much smaller companies that offer similar products or services but at a much lower costs.
Note that this does not even require innovation on the product itself but simply optimizing the product building process.
TL; DR:
Understanding how to leverage GPT-based technology in your workflow is probably the most important skillset to learn 2023 as a PM and engineer.
If you’d like to learn more, there are 3 ways I can help you:
- Subscribe to my newsletter or follow me on Medium if you want to get actionable tips on software product management.
- Email me at danielschmitter@substack.com to provide suggestions about blog topics.
- Follow or connect with me on LinkedIn here.
