Think Big, Start Small
đĄIdeation for AI Products
LLMs are reshaping the way we work, but turning big ideas into actionable solutions can feel overwhelming. As builders, we can sometimes get stuck between ambition and execution. This post will guide you through identifying impactful areas to apply generative AI and provide a detailed exercise to spark actionable ideas for integration.
đ§ Why start small?
Simply put, the possibilities with Generative AI are overwhelming. If youâve got an idea or are passionate about a problem, there likely exists a genAI tool that can help you build your solution. From no code to LLM driven coding interfaces to entire flows managing and orchestrating agentic LLMs, a tool exists to facilitate building your product. If youâve never coded before, and this was your barrier to building, well, now you donât have to, as long as you understand the fundamentals. If youâre decent at software development, used right, LLMs can bring down your coding time to as low as 20%. From ideation to production, several new products entirely built by LLMs have now hit our app stores. If youâre leading an existing product or service, everyone including your customers, your employers, and your employees, expects you to use AI â have you seen the newest line of AI powered vaccum cleaners (!)?
If building excites you, all this can be overwhelming. If youâre anything like me, this feels like a time of endless possibility, but also a time of rapid paradigm shifts. Things are changing so fast itâs hard to keep up. There are so many ideas, so many experiments you want to try, that youâre starting things just to abandon them halfway before moving on to the next big idea. If thatâs been your mental state, Iâd urge you to breathe, and focus on starting (and finishing) small.
Focusing on, and iterating on, a small problem to solve, is usually a great way to develop the toolkit required to build a larger solution. Itâs tempting to aim for the moon when exploring GenAI. But revolutionary change often begins with evolutionary steps. By identifying a small, impactful area to implement generative AI, you reduce the risks of overreach, speed up iteration, and build confidence in your ability to execute.
For example, an AI-driven customer support chatbot might start with FAQ generation. Once the feature demonstrates value, it could expand to handling complex customer interactions, learning from live agent interventions, and even having access to tools over time, creating incremental value at each step.
The key is to let the technology serve the problem, not the other way around. Instead of building a flashy, overcomplicated solution, focus on solving a single, meaningful pain point. Iterate, gather feedback, and expand incrementally.
đŻ Identifying Impactful Areas for Generative AI
Before diving into brainstorming, itâs crucial to understand where generative AI can make the most impact. Start by asking:
What are the repetitive tasks in this product?
Example: In a customer support platform, generating responses to common queries can automatically reduce the workload on agents and improve response time. LLMs could generate comprehensive templates for troubleshooting, or dynamically create email drafts tailored to customer concerns based on past interactions. Beyond this, they might analyze call logs and automatically create scripts for escalation scenarios, or even invoke tools to directly guide users to finish their tasks, instead of requiring a human agent.
Where does personalization matter most?
Example: In an e-commerce app, using generative AI to create dynamic product descriptions based on individual user preferences or past purchases can elevate the customer experience. For instance, LLMs could rewrite descriptions to highlight eco-friendly attributes for sustainability-conscious buyers, or suggest complementary products for repeat customers.
What unmet needs exist for better content creation?
Example: In a marketing tool, LLMs can enable drafting social media posts or ad copy tailored to specific audience segments. The could help test variations of campaign content in real-time, learning what resonates most with each audience segment. This could bt taken further by integrating performance data from past campaigns to adjust tone and style dynamically, offering iterative improvement over time.
Where could insights lead to better decisions?
In a financial analysis platform, LLMs (augmented with existing data systems and reports) can be used to summarize trends from large datasets and generate reports with actionable recommendations. In the product discovery phases, they can synthesise user research documents to drive insights and provide meaningful recommendations for what to build next.
By focusing on these areas, youâll uncover opportunities where GenAIâs capabilities align with real-world needs.
đ§ A Comprehensive Brainstorming Exercise
Letâs take a more in-depth approach to ideation. This extended exercise will help you explore opportunities for generative AI integration through a small step, iterative approach.
Step 1: Pick a Product Youâve Thought About Deeply
Choose a product youâve worked on, use frequently, have studied, or really want to build. It could be anythingâa project management tool, an e-commerce platform, or even your favorite fitness app. If youâre unsure where to start, pick the product you interact with daily, as familiarity often leads to deeper insights.
Step 2: Map Out Core User Journeys and Supporting Workflows
Document the main workflows accomplished by the product. For example:
For a project management tool: creating tasks, assigning resources, generating progress reports, integrating with time-tracking apps.
For an e-commerce platform: browsing items, comparing products, completing a purchase, post-purchase support, and feedback collection.
For a fitness app: tracking workouts, planning meals, analyzing progress over time, connecting with fitness communities, and sharing achievements.
As much as possible, break these workflows into micro-tasks to reveal nuanced inefficiencies AI could address. Additionally, consider how workflows intersectâfor instance, how the browse and save behavior on an ecommerce catalog can affect future purchase behaviors (spoiler â more saves = more purchases).
Step 3: Identify Pain Points or Inefficiencies
For each task, ask:
Where do users spend unnecessary time?
Where do errors or bottlenecks occur?
Whatâs currently âgood enoughâ but could be exceptional with GenAI?
How does scaling impact these workflows?
For example, in a fitness app, LLMs could streamline planning by generating meal suggestions that align with dietary restrictions, nutritional goals, and user preferencesâall while learning from logged meals to improve future suggestions. Scaling this across millions of users could offer hyper-personalized plans without additional manual effort. Additionally, the system could proactively recommend substitutions based on seasonal ingredient availability.
Step 4: Map GenAI Capabilities to Each Pain Point
Using the questions from earlier, brainstorm how generative AI could enhance or replace parts of these workflows. Here are some examples:
Content Creation: Automatically generate templates for recurring tasks. Example: GenAI could create interactive project timelines that adapt dynamically based on real-time input from multiple collaborators.
Summarization: Summarize meeting notes or user reviews. Example: In a conferencing app, LLMs could identify patterns across multiple meetings and provide strategic insights, like recurring blockers or themes.
Personalization: Suggest tailored next steps based on user behavior. Example: In a fitness app, AI could evolve into a virtual coach, generating not just workout plans but also motivational prompts based on detected patterns in user engagement.
Prototyping: Quickly draft mockups or design ideas for review. Example: In a design tool, GenAI could produce end-to-end workflows based on wireframes, including user onboarding sequences.
Step 5: Prioritize and Refine
Evaluate:
Scalability of the GenAI solution.
Expected ROI based on user feedback and market trends.
Technical complexity and data requirements for implementation.
Consider creating a feature backlog where each proposed AI enhancement is tagged by difficulty, data dependency, and potential user value. Use this to map out short-term experiments and long-term strategic goals.
đ From Ideation to Execution
Once youâve brainstormed and prioritized ideas, itâs time to bring them to life. Here are some possible next steps:
Prototype and Test: Use existing tools or APIs to create a minimum viable version of your idea. Leverage APIs and low code tools to generate functional but fast prototypes that can be immediately validated with real users.
Learn and Iterate: Analyze feedback to refine your solution. Think of where your prototype users got âstuckâ, what challenges did they face in usage, and what they found easy enough. Define a âviableâ product that solves for pain points and stumbling blocks.
Scale Strategically: Once validated, scale features thoughtfully. Consider localization, compliance with regulations, and infrastructure optimizations.
Monitor and Maintain: Build ongoing monitoring pipelines for your AI features to ensure quality over time. Use dashboards to track user engagement, AI-generated errors, and performance metrics.
The transformative potential here isnât about creating something that wows people on paper. Itâs about delivering tools and experiences that feel indispensable in practice. Start small, but keep a clear vision of how your solution could scale and evolve into a broader ecosystem of AI-driven innovation.
đ Ready to Start?
Generative AI offers a playground of possibilities for builders. By thinking big but starting small, you can tackle meaningful problems, unlock new value, and set the stage for innovation that scales. Remember that GenAI is more than a toolâitâs a mindset shift. The intersection of user insights, AI capabilities, and strategic execution can lead to breakthroughs that redefine industries.
So, how would you integrate GenAI into a product you know well? The answer might be the seed of your next big breakthrough.
Read these to dive deeper into building â
