Generative AI Product Strategy
In the dynamic world of product development, Generative AI is more than a trending buzzword - it represents a transformative force reshaping our strategic landscape. But, are we truly leveraging its full potential?
TLDR
Generative AI isn't just an addition to the traditional product strategy stack; it's a paradigm shift, introducing new dimensions to how we define and execute ideas and vision. The process of HOW to build process has been simplified with lots of technical resources, tutorials and support available. However, we are still in the infancy of WHAT we need to build and WHY we should build a product. In other words, there is little to no innovation on the strategy, product vision, team alignment and execution.
This article deep dives into:
Generative AI product strategy stack
Three rapid strategies to get started
Levels of Artificial General Intelligence (AGI) & metrics blueprint
Resources to build the foundations of product teams
If you like the content of Hatch Labs newsletter so far, we would be grateful if you refer a friend to subscribe and spread the word!
Generative AI product strategy stack
Here is one of the main questions that conquers board rooms and executive meetings these days as we prepare for 2024 planning:
Is it actually feasible to have positive ROI in 2024 when implementing these research generative AI products?
The traditional product strategy stack would look into:
company mission: is this the right problem to solve based on the company’s mission?
company strategy: is this the right solution? Does it provide quantitative moat compared to existing benchmarks or alternatives?
strategic differentiation: why you, your company and your product will stand out?
product roadmap & goals: what is the sequence of steps that deliver the proposed outcomes?
The four pillars of generative AI product strategy stack are:
Let’s explain each of these.
Technical feasibility → HOW?
Technical feasibility answers the question of HOW. This is a multi-dimensional question. To answer that, leverage resources such as the generative AI Flywheel to gather requirements and define the product roadmap based on data, AI models and user feedback.
Value → WHY?
The product value answers the question of WHY we need to leverage AI to build this product and WHAT the value of the product is in the AI value chain. Is it a new vertical AI application, new foundational model provider? A helpful resource to position the product is the AI value chain.
It is important to address the strategic differentiation as part of your value, which can be deep domain expertise in an industry vertical, unique customer base, defensible IP (this will help with how you tackle hallucinations), exclusive data resources (proprietary datasets) and distribution channels. Currently, AI lacks a distribution channel and it will be either incumbents or startups that leverage traditional sales channels that will emerge.
Adoption → WHAT?
Adoption answers the question of WHAT the most valuable use cases are that have the quickest time to value.
Adoption criteria can be grouped into:
internal: what are the organizational skills, leadership and/or cultural shifts that need to happen for this product to be adopted?
external: what are regulatory, ethical and AI governance aspects that may impact adoption rates?
technical: what is the innovation/novelty of this product?
Risks → WHAT?
There are multiple risks associated with the generative AI strategy. A useful resource is to look at the AI governance framework aspects that are applicable for your use case.
There are lots of risks and vulnerabilities that need to be addressed, but it is important to start small and iterate multiple times until you have defined a defensible strategy.
Three rapid strategies to get started in generative AI strategy
Below I present 3 rapid strategies that I have implemented in my product career.
1️⃣ Rapid Wins with Pilot Workstreams: Identify where Generative AI can most effectively demonstrate ROI. Kickstart pilot projects to unveil its immediate value.
Example: At a large software company, I started with a hackathon project and built a team of 32 people with an initial prototype for recognizing mechanical parts based on machine vision.
Start small, a hackathon, an innovation day, a pilot work stream and you'll be amazed by what you can achieve afterwards!
2️⃣ Embrace a Data-Driven Roadmap: Shift from intuitive decisions to a quantitatively approach to strategy. Align feature rollouts with data-driven insights and market trends.
Some metrics you can track:
engineering effort
impact
time
Although these may seem qualitative metrics and it's more natural for us to define those metrics qualitatively.
In my opinion, it is absolutely crucial to quantitatively define metrics and the team should be aligned on these metrics.
Example: Recently, I was able to help an advertising startup achieve 150% revenue increase. Establishing concrete metrics has been the major driver of my product launches - as early as 2 weeks from team formation.
3️⃣ Simplify to scale
Large organizations can struggle with imprecise definitions and objectives. Simple and structured mockups are key to steer alignment in large teams.
Example: I helped a government organization define their digital twin strategy by using a structured incubate framework. This strategy led to multi-million investments.
Levels of AGI & metrics blueprint
AGI, which stands for Artificial General Intelligence, means that a machine has capabilities to understand, learn and apply its intelligence to a wide range of problems, similar to a human being. Unlike specialized AI (such as computer vision, natural language processing), which is designed to excel in a specific task or set of tasks, AGI is characterized by its versatility and adaptability, enabling it to perform any intellectual task that a human can do.
The development of AGI would mark a major milestone in AI research, potentially leading to profound changes in technology, society, and how we interact with machines. Recently, Google DeepMind released the framework with the six levels of AGI. According to their definition, ChatGPT and Bard are emerging AGI, and no level beyond that has been achieved.
Measuring the performance of today’s models is a challenging task and assessing model performance will still be more difficult.
In a previous post, we discussed multiple categories of metrics used to quantify the performance of Large Language Models (LLMs):
🔢 Quantitative performance metrics
🩺 Operational & health metrics
👥 User engagement metrics
🏛 AI governance metrics
⚖️ A/B testing metrics
Generative AI metrics blueprint
There are three components when defining your generative AI metrics blueprint:
Product objective (AKA OKR in large organizations)
What is the main goal your product achieves?
Example: Improve customer satisfaction by providing quicker and more efficient customer support with AI chatbots
North star metric
This is the main metric that showcases product success
Example: Reduce response time of AI customer service chatbots by 30%
Guardrail metric
List the potential adverse effects that need to be monitored and controlled. A guardrail metric is a secondary metric used to ensure that the pursuit of the North Star metric does not lead to unintended negative consequences.
Example: Maintain a customer support complaint rate below 5%.
KPIs
All other metrics used to measure performance and success rate towards achieving your objective. There may be multiple KPIs and multiple product objectives, but only one North Star metric for each product objective.
Example KPIs:
average resolution time (operational & health metric)
customer satisfaction score (user engagement metric)
chatbot escalation rate (operational & health metric)
number of hallucinations per 100 chatbot responses (quantitative performance)
compliance adherence rate (AI governance metric)
Let’s connect the dots now and see some KPI examples per each AGI level.
Conclusions
There is no definitive guide or framework on what a concrete generative AI strategy entails as well as a metrics blueprint for product leaders and founders to follow. This post is just the beginning on what Hatch Labs is working on. Product strategy is about solving the right problems. If you're not focusing on the problems worth solving, you're just creating more noise.
As Warren Buffett mentioned:
In business, I look for economic castles protected by unbreachable 'moats'.
👉 We provide further guidelines and frameworks as part of the generative AI bundle offered in our referral system.
The recipe to success is to start small and build on safe grounds. Below is a list of sample resources to build the foundations of product teams.
Resources to build the foundations
👉 Understanding Product Failures: Learn from Marty Cagan's insights on the root causes of product failures. Watch here: https://lnkd.in/eqKZXzTw
👉 Modern Product Discovery: Dive into Teresa Torres's book on continuous discovery habits. Find it here: https://lnkd.in/e4qQv-PV
👉 Lean Product Playbook: Explore Dan Olsen's playbook for creating products that customers love. Check out the book: https://lnkd.in/evkiQpPe
👉 Crossing the Chasm: Understand Geoffrey Moore's perspectives on taking disruptive products mainstream. Link here: https://lnkd.in/eKTme28J
👉 Hacking Growth: Learn growth strategies from Sean Ellis, applicable for fast-growing companies. Check it out here: https://lnkd.in/eMt77gbW
How do you plan to build your AI strategy into a defensible moat in this competitive landscape?
Hatch Labs actively helps startups & enterprises to define their generative AI product strategy and product market fit. If you think you or your colleagues can benefit from this, let's have a conversation. Happy to chat!