Our AI/ML waters are absolutely insane right now. It’s another case for staying adaptable instead of obsessing over processes. As the Head of Product at Jua.AI, I have a very specific take on it.
We need to stay flexible for the next 3 months. 3 months is a long time in a fast-changing market:
How do we structure a flexible process that is simple and lightweight for Jua.ai (Machine Learning / AI) to understand and use in a fast-changing market?
There are three areas to get right:
Product Strategy
Outcome-driven goals
Adaptability over Execution Speed
Product Strategy
Be prepared to have your product strategy to be challenged but the entire staff needs to be on board with it.
Assume misalignment and double-check whether everyone understands what you mean by your goals. Share your hypothesis for every goal that management came up with.
We have a specific notion document that details a first draft of the hypothesis behind every assumption about our diverse market.
Outcome-driven goals
Outcome-driven goals are the only saving grace in fast-changing markets. Spending enough time to figure out what really matters is key. This requires research and an intentional way of figuring out that our direction is the correct one before we start running off.
For early startups it makes sense to include two dimensions:
No traction: What do we do if we have no traction in the market
Traction: What do we do if we have more traction than we anticipated
The reason to differentiate between these two dimensions is to make sure that the “no traction” scenario means “all hands on the sales deck”. We only build what helps sales or acquisition. Tech debt for internal tooling and processes is getting deprioritized
The traction scenario is the opposite. If the acquisition is running we need to deal with scaling problems before they hit us. Focus on tech debt and analyze other nasty problems if everything starts to break. It will.
Adaptability over Execution Speed
Include in your goals further validation of your product strategy. Be prepared to overthrow it all if the market sends us contrary signals to what we assumed.
A good measure that you are doing this is to not have a single delivery committed to your goals.
Bigger goals should be layered, are there intermediate steps that we would consider early signals for success/failures - MVPs don’t get shipped without analytics. Our problem is learning, not shipping.
What are cross-team dependencies, and how might every single goal fail?
We check our OKR hypothesis once per month since we believe that alignment on the correct goal is more important than praying to the agile process gods.
Process
Below you can see the specific process we use at Jua for each team after we agree on a rough strategy. The goal is to give teams autonomy as long as the outcomes are clear. How they are executing on these goals is up to them if they find a great way.
We don't obsess over processes we obsess over their outcomes. If that means less process that's fine.
Well, pivot - FAST!
…assume that reality will hit you like everyone else. Don’t get attached to your goals and be ready to pivot
We recheck our OKR commitments once per month with management check-ins. Assume that you run in the wrong direction.
One very specific thing I’ve started doing is to use ChatGPT to do the very first rough dirty work when exploring new verticals which in the end leads to potentially new goals. While the output is spotty some of the hints are actually quite useful as they change how I approach domains that I’m not necessarily familiar with.
Those who figure it out faster will prevail.
Love this. Keeping some constraints on the strategy while giving the team freedom to execute on their version of "the best way" is a winning play for such a fast moving area.
Love it. Better if I could grab that template directly to try it out.