Hrishikesh Paranjape
One consequence of a rapidly evolving tech landscape is increased reliance on artificial intelligence (AI) and machine learning (ML) technologies in product development. These technologies can unlock myriad benefits, such as deeper user research and quick prototyping for idea validation, a greater understanding of the target audience, and creative solutions that were not possible just a few years ago. But as the technologies have grown in popularity, it has become clear that realizing their true potential in product development requires an effective product manager who can be the voice of the customer when ideating with interdisciplinary functional teams and who has the knowledge and experience to leverage AI and ML techniques to solve challenging problems.
Unfortunately, the learning curve associated with newer technologies can be difficult for product teams and those responsible for growing the product rapidly, including business operations and go-to-market/sales ops teams. This learning process and other challenges, such as dealing with privacy and security concerns for enterprise customers, and ethics and accountability when using outputs from large language models (LLM), often slow down AI adoption in the product development process. Hrishikesh Paranjape, a senior product manager with broad-based experience across the e-commerce, real estate, customer service, and financial industries, understands what it takes for successful product managers to effectively use AI/ML technologies to build and deliver delightful customer experiences.
Q: In general, how are product managers dealing with the rapid evolution of AI and ML?
Paranjape: In the short term, the adoption of new technology can make a product manager’s job more difficult instead of easier because of the learning curve and uncertainty about direct customer benefits. That’s why, to ensure companies get maximum advantage from the expensive training and inference required to “productionize” AI-driven solutions, it’s imperative for product managers to constantly champion the voice of the user as their primary role. It’s critical for product managers to provide clarity about a specific user pain point that needs to be solved with priority, collaborate with cross-functional teams, and effectively navigate a complex technical landscape. Using generative AI (GenAI) may be the right approach or traditional deep learning and reinforcement learning techniques could be more appropriate given the cost-benefit tradeoffs. When product managers do this, they can ensure that AI- and ML-driven products deliver delightful experiences that meet or exceed user expectations and are economically viable for the business over the long term.
Q: What steps can product managers take to maximize their knowledge of new technologies and identify user stories that could be solved with AI/ML?
Paranjape: First, it’s crucial for companies to commit to investing in training for their product managers. Yes, AI can speed up product development while making it more efficient, but AI adoption frequently comes with challenges. These challenges can be addressed by educating product managers about how to use AI responsibly and within the requisite governance framework and guardrails. When product managers have opportunities to collaborate with AI experts and access continuous learning programs, they are better equipped to assess trade-offs when using these technologies daily. Product managers should have agile implementation techniques at their disposal when prioritizing user stories in the backlog, always taking a “customer-centric” approach for AI-first solutions and doubling down on collaboration with technology and applied science partners to assess implementation readiness in their respective organizations. One additional recommendation: give product managers the leeway to test and learn prototypes early as part of their daily job duties. For example, enable them to build proof of concepts (PoCs) to solve a tangible user pain point with limited engagement from applied science teams, even if it means taking some bandwidth away from their regular work. These are a few ways product managers can improve their use of AI and ML.
Q: What about the data privacy and security concerns that surround AI usage today? How should product managers address those?
Paranjape: The first step to addressing privacy and security concerns is ensuring that policies are crystal clear. When creating policies, it’s important to have a definite separation between enterprise data sources used for retrieval augmented generation (RAG) and training data for foundation models. Another best practice product managers can adopt is to focus on mitigating the problem of hallucinations in LLMs. This is where AI models can make up details or respond with unsubstantiated answers to prompts. Possible solutions include refining prompts, improving data quality through access to enterprise data sources and company-specific sources, and engaging human reviewers at each step of product development. Finally, it is also vital to develop a proactive governance framework that details the protocol for handling potential issues. This helps avoid falling into a reactive position, where the product manager constantly puts out fires to the detriment of their other duties.
Q: As a product manager in the last-mile delivery space, how is AI improving the development process?
Paranjape: Routing technologies today have access to a broad range of datasets that weren’t available before. There is now extensive information about end-user customers. For example, we learn about their specific delivery preferences, based on information about their orders. We can also ascertain on-road inputs from delivery personnel who have been to the property before. This data includes crowdsourced information about building access and shared locations such as concierge rooms, in addition to GPS information gathered from drivers that serve those destinations. This data is then filtered through AI and ML models to generate optimized delivery routes that save companies time, ensure the highest safety standards for transporters delivering packages, and improve customer satisfaction rates. When there’s insufficient information about a property, delivery apps can assist transporters with helpful generated notes based on successful deliveries in the past.
Q: In the last-mile delivery context, could AI and ML ultimately make late deliveries a thing of the past?
Paranjape: Late and low-quality deliveries are the easiest way for companies to lose business. Fortunately, this lateness can be eradicated with effective inputs and advanced optimization within specific business constraints. ML does provide companies with a new advantage by using data to foresee traffic jams or other obstacles in a specific area. Additionally, learning models can predict likely outcomes using previous delivery details and recommend more optimal routes when necessary. This same technology allows for more accurate delivery time predictions, which is crucial when customers eagerly await a package to arrive. AI and ML technologies are also likely to increase drone deliveries, especially in suburban and rural areas, but not so much in highly populated, dense urban areas due to regulatory challenges, reliability, and safety concerns.
Q: What is the most important advice you have for product managers about how to successfully adopt AI and ML technology?
Paranjape: As AI advances rapidly, the key to success is a willingness to “go deeper.” The most effective product managers embrace the large amount of information they now have access to and leverage AI and ML to identify actionable insights from detailed datasets. For instance, they can create expansive lists of pain points organized by themes, test and validate hypotheses, and analyze results from surveys and focus groups. This allows them to work backward from the user when envisioning product prototypes that achieve a higher level of differentiation. If a long-standing pain point hasn’t been solved with existing solutions, there is greater potential that new technologies can make a difference.
The future of AI and ML in product development
The benefits associated with AI and ML make widespread adoption of the technologies across industries inevitable. That doesn’t mean there won’t be challenges arising from their use. When product managers understand the pros and cons of AI/ML and leverage the technology responsibly, rich customer insights are discovered, creativity in problem-solving is enhanced, the product development process is streamlined, and the “human touch” that allows companies to appeal more strongly to their target market is retained.
