Is product dead?
It's one of the questions Product organisations constantly come up against, but so are the ones around whether we need super specialised product roles like AI Product Managers, Product Marketing Managers or GTM Product Managers.
In our work with organisations like the BBC, the Economist, and Central Government, we find more than ever, that good product management and product mindsets wins out over domain-specific knowledge and expertise.
Here’s why:
For most organisations, we’re building AI-enabled products and services. Not AI-native products. Given that, we’re often looking for ways that AI can deliver better or more efficiently for users we already serve. We’re not necessarily shifting the paradigm on what we’re delivering.
So, in the context of AI-enabled products, AI is a tool. It’s a way to deliver more value from the data you hold. AI doesn’t have value in and of itself. AI as a tool should sit alongside other tools, like service design, customer experience design, or marketing. Choosing the right tool for the job is part of a good product manager’s skillset.
In AI-enabled products, data = value
If AI doesn’t deliver value in of itself, then where is the value held? That is in your data. That data might be:
- Customer data, such as behaviours, preferences or needs. When we use AI tools, that data might be able to produce more personalised experiences that save them time and increase uptake and conversion.
- Data about inputs, outputs and outcomes. That data might allow us to predict future results or automate simple decisions.
- Past analysis or content. That data might allow us to generate new analysis or content for similar areas.
AI tools allow us to unlock that data to deliver differentiated, more valuable or more efficient experiences and insights. Let’s look at some examples:
Using procurement data to save money for schools
For the Department for Education, we explored how to scale a procurement-support service for schools using AI. We used data about past procurements to run experiments into how to save human-time in the procurement process. The fact that DfE had ‘gold-standard’ past procurements meant that it was possible to train AI on what good might look like in certain scenarios. Consequently, we were able to show that we could shave as much as 2 weeks from a typical journey. Without that data though, an AI product would be inaccurate or unguided.
Using past analysis to ‘predict’ the future
For the Economist Intelligence Unit, we used the article database of past analysis to quickly generate ‘Scenario Pathways’. These are potential future scenarios, based on past events and analysis of those events. Experimentation in this space helped the EIU understand that their data was an important differentiator for them, and it was important how they managed and protected that going forward.
These examples illustrate why data is where the value lies when delivering AI-enabled products or services, and why applying product management techniques to data is vital to leverage that value.
Product management and data
Domain specific knowledge, and job-titles that focus on just one tool, can lead us down the wrong path. When working with organisations considering how to get more from their data, we encourage them to apply broad product mindsets. For example:
- Understanding value. Product people start with asking questions. Why collect this data? What value are you trying to deliver with it? How do you tie that back to users and business goals? Understanding all of that means your data is able to deliver on the potential value with limited effort and/or investment.
- Continuous improvement. How can you continuously improve the data you have? Do you have a roadmap of improvement opportunities ranked by value or need? Are there regular pain points or issues created by your data that can be addressed at source?
- Continuous discovery. Are you applying discovery approaches to your data? Do you know what needs it’s serving well, and what it can do better? Do you have a backlog of opportunities to validate or understand how your data might be able to meet customer’s unmet needs? If you’re supplying data to other teams, do you know how they consume it and whether it’s meeting their needs?
- Availability, scalability, reliability. Are you making your data usable for consumers? There’s no point exploring the value of AI if the data it will rely on isn’t available outside of the AI experimentation arena.
- Operations and support. Products require support; often humans at the end of a phone or email inbox who can answer queries or reset a password. Data also requires ongoing support to manage feedback, queries and requests. Every day, we add more data to our data sets, and this requires active support. How do we add to that data without unknowingly introducing poor quality or malicious data?
Andrew West-Moore, currently leading product approaches to machine learning (ML) and Generative AI at the BBC, has found the most valuable thing is understanding what problems you're trying to tackle, what the user needs are, and whether ML/AI aids as a tool. We shouldn't default that AI as the answer. Understanding the impact you're looking to have enables you to work out the investment and effort you should expend.
Applying product thinking to your data will enable you to deliver greater value to users and customers. Avoid the trap that AI = value. Instead, explore what additional value can be delivered from the data that helps differentiate or support your organisation and its goals.
Transform is currently working with organisations like the BBC to do just that. If you’re interested, get in touch. Equally, if you want to experiment with AI for your organisation, ask us about our AI Lab.