Subscribe
Share
Share
Embed
We follow the journey of Faith Forster as she creates an AI native tech startup & product.
Hey. Welcome back to Building Out Loud with me, Randy, and
Faith Forster:Faith Foster.
Randy Silver:And we are following Faith's journey as she's building Discoveree and doing lots of interesting stuff. And we're probably gonna get a little more philosophical today and a little more abstract. But just a quick check-in first. What's happened in the last week? Faith, where are you at today?
Faith Forster:As you know, I've spoken to a lot of people. I've got a mountain of feedback that I'm now busy using to update the product. So the last week has been heads down vibrating. I showed you last week the competitor page. We're moving it more towards a strategic advisor.
Faith Forster:I'm now taking that throughout the whole product experience. And I have to say it is going far beyond what I anticipated when I first started working on this, because where I was originally covering two core ideas one was largely automating or doing the legwork for the core product workflow, and then also adding on the link to business goals and understanding the implications of decisions we're making. I'm now also adding into that some agents that will help think through and shape and apply the product strategy as well so that that will be embedded into all the decision making as well. It is super exciting. I've had some input from a few different people on what some of those considerations are.
Faith Forster:I just need to do it, and hopefully, have something to share next week.
Randy Silver:Fantastic. Okay, we'll look forward to that. But you've also had some philosophical ideas, some fundamental wonder of the world. Why do we do product this way? What is the fundamental nature of the approach type things going on?
Randy Silver:Where's this coming from?
Faith Forster:It's been really interesting. As you know, I've now spoken to over 70 product leaders across The UK and Europe as input into this product design. A lot of this also came from the lunch I had a week or so ago with Antonoska, founder of Lovell. So he brought together 12 AI native product leaders and also AI experts to think about the future of product development and where is this heading? What does this mean?
Faith Forster:And there was some really interesting discussion that came out of that that has made me stop and rethink how we're approaching what we're supporting through the product. So things like, if it takes days to build a fixture, what's a roadmap in that context? Like, do you need epics and tasks? Or is it actually just a series of things that you're working through? There was a really interesting comment from Rags, who is the founder of floato.ai.
Faith Forster:They're creating AI personas based on existing customer research to help designers in the moment in Figma apply feedback from their customer segment. And he's a career Google PM. And so he's done PRDs his whole life. And he was saying like, it didn't work. We're no longer defining pictures for engineers.
Faith Forster:We're speccing it for an agent. And an agent needs different guidance. It's not just the success criteria. It's actually thinking through what is the behaviour? What's the tone?
Faith Forster:What's the feeling? What does good look like? What does bad look like? And giving those sort of guardrails, but also an outcome for the agent to then apply. And so the way we share that understanding needs to be different.
Faith Forster:So he's created a template which he calls the product experience document instead. Covers quite different things. As I've been thinking about how this should work and how we support this within Discoveree, I've really had to face into the question of, do I design this product for how product teams are working today? Because there are still a lot, most, who are still working under a traditional product life cycle, feeding things into Jira and then engineers pick them up and, they may or may not be using agents to code. So do you design this product for how people have always understood this process works?
Faith Forster:Or do you design it for where I think it's heading? Even though that's not really defined yet, it has been a really interesting reflection, I guess. It's been some quite specific things I've tested with some people. So even the concept of a Product Team is changing. It may or may not continue to be a PM, a designer, and devs.
Faith Forster:And so I've had a couple of examples who are changing the nature of those relationships already. And so even designing the workspaces around Product Teams doesn't really make sense for them. And so, I've reframed it from a Product Team, a workspace for a Product Team, to a Product Domain instead. I've split out the idea of this is a space within the product that people are working on as opposed to the role or the people who form the team around them.
Randy Silver:Let's definitely talk about Teams, but you started off talking about PRDs and epics and tasks and artifacts. And let's talk about that for a couple of minutes and then we'll come back to the people who are doing it. My day job, I'm a consultant and a coach. I work with companies and people of all different shapes and sizes, all different levels of maturity on this kind of stuff. And it's always fascinating because they sometimes ask me to come in and say, what is the right way of doing this?
Randy Silver:What is the framework that we should be doing? And I teach people everything from how to write a user story to how to do a road map and things like that. And yet, there are some basic techniques to doing this stuff. But what I really hate is the people who think the artifact is the point, that the artifact is the thing. And fundamentally for me, I don't care if it's a PRD, an epic, a story, a road map, whatever it might be.
Randy Silver:It's there for a reason. It's to help promote people to have better discussions, to create understanding. Our mutual friend Monica Turska, had this going years ago, the entire job of what we do is help people make better decisions faster. That's it. People spend ages and ages putting a map up.
Randy Silver:We used to do it physically on a wall, and they're out of date five minutes after you finish them. And the point wasn't the the physical thing itself. The point is the understanding that everyone in the room had so that you can make better decisions. Yeah. So I don't know what the right artifacts are.
Randy Silver:I mean, some people swore by PRDs and some people swore by canvases. It doesn't really make a bit of difference to me what you use so long as it works. And it's fascinating to see what the next evolution is going to be as the tools change. Sorry, I'm ranting now.
Faith Forster:No, no, completely agree. I've always been someone who, you know, before I even got into product, was a management consultant and management consultants lived by their methodologies. Always been skeptical of methodologies. Like it's just a means to an end goal. And really it's about the outcome.
Faith Forster:What are you trying to achieve? And then what's the best way to go about achieving that? Any environment I've gone into, I've always looked at what's working, what's not, what do we want to change? What's the biggest unlocks for this product org? And then worked on a process of framework, whatever it is we need to unlock that.
Faith Forster:But it's always been with a clear purpose in mind of what we're trying to achieve. But it's slightly different when you're then building software, right? Cause at least traditionally software has hard coded this stuff. And I don't think that's necessary anymore, and that's not what we're doing with Discoveree. But Discoveree is an opinionated it has a strong point of view in terms of what good looks like.
Faith Forster:It's also flexible and customisable. And so it's sort of getting that balance right and understanding inflection points, where do we need to put forward a point of view as a starting point, and where do we need to allow users to then adjust that to suit wherever they're at in an environment of quite strong fundamental shifts about how we think about and do this.
Randy Silver:One of the more basic questions you was not only what do we need to be using in terms of these artifacts, but who are these artifacts for? Who are we defining the features for? Are we defining it for people, our dev partners, or are we doing it for agents? For my other podcast, the product experience, had a discussion a couple months back with Kasha Forbes, and she was talking about machines as customers. You're building sites for agents to consume because they will be acting on behalf of people, but you're talking about the building side of this.
Randy Silver:So where are you on this right now? Who are you defining features for? Or is it a definitive deterministic answer on this one?
Faith Forster:No. I really don't think it is, actually. That's exactly how I've ended up thinking about building it out at the moment is so within an opportunity, you define the problem statement, which should be pretty clear based on the feedback. From there, you can come up with solution ideas. So if there's an agent that will help you or the team can do it themselves.
Faith Forster:The way I've built it now is we take them through a two step process where they then scope out multiple ideas and think about what could we do here and what impact would that have on our goals. And then you choose which one to take forward, having compared a number of different options. And then there's another series of agents who then help you work through that solution design. But the first question we ask with the solution design is who's the audience for this? Is it an AI agent?
Faith Forster:Is it an engineer? Is it a stakeholder? The way I've formatted it is the one that's creating the solution design for an engineer creates a PID document. You can change what you want covered as part of that PID. Likewise, if you're designing this for an AI agent, it covers more of those behavioural, feeling, tone, what's good look like, what's bad look like type questions.
Faith Forster:And then for a stakeholder, it's more why should we do this? Who are our customers? What sort of investment are we talking about? What return do we think we're gonna get from it?
Randy Silver:You're still calling it a PRD there, but is it the Google style tome of a PRD, or is this a one pager? Is it the same format as what we might have thought of a PRD in the past?
Faith Forster:I fed into the agent a few different examples of PRDs because I do it like this. So it has the typical framework you might expect of problem solving, who's the audience or the personas, what does success look like, what's the success criteria, what's the scope, what are the edge cases, what are the risks? I think probably the more important point is an organisation can choose to change that. They can go into the prompt for that agent and modify what they want that agent to produce as part of that sort of scoping document or design document. The more important point is being really, really clear on there are potentially different audiences for different organizations that are at different levels of thinking or maturity about how they do product development now.
Faith Forster:I guess being able to help organizations meet where they are now, but also support as they progress towards maybe using agents more.
Randy Silver:It sounds like the idea of minimum viable bureaucracy. What is the smallest thing that you can create that creates the understanding that everyone needs so that they can work together, they can prioritize effectively, they can collaborate and communicate? It also sounds like it goes back to the agile manifesto's statements about understanding over documentation.
Faith Forster:Agents in the way that we use them in products, they are often one to one interactions. It's one user talking to the agent and getting a response. A really important part of the way we think about and build products is collaboration and alignment and getting lots of people involved in making those choices. And so that's one of the things I've also had to think about. There are agents who will ask questions and prompt you through the thinking to then shape what this picture might look like.
Faith Forster:But actually, that can't just be one person. How do you get the team involved? And how do you get it to the point where it's then shareable and you can get input back from other people? Those are some of things I've had to think through and build into the way it interacts.
Randy Silver:We've long talked about how do we make sure we have enough viewpoints in the room? How do we have a diverse enough team in terms of the way they're thinking to consider everything that needs to be considered? And when you're doing this alone with an agent, the elements of bias or a lack of perspective can come up really, really quickly. One of the things I was always taught was if you're all together at meeting and you look down and everyone's wearing the same type of shoes, you don't have a nearly enough diversity as a very quick way of looking at. But when you're doing this alone or in a small group with agents, how do you make sure that you've got enough of a diversity viewpoint represented and brought to this?
Randy Silver:How have you dealt with this so far?
Faith Forster:It's a really important fundamental question actually because that's historically been the whole premise behind a product team. It's having the different perspectives that you need to understand and manage the various risks involved in building or investing in whatever it is you want to build. That concept is shifting. We won't necessarily need full time people representing each of those perspectives going forward. It would be seriously limiting to not have the different perspectives represented in some way.
Faith Forster:So the way that I've supported it so far is two ways. It's one through at mentioning and being able to pull in the right people, and the agent can do that as well as the users. But then the second thing is flagging where you should then share and get input from a wider set of people as well and enabling that to happen too.
Randy Silver:But you just mentioned bringing in the right people a couple of times, you also mentioned, and I'm sorry, forgot the name of the guy you had lunch with that's doing artificial personas. Is the agent always going to bring in actual people or is it gonna bring in artificial people as well? Where do you see this going?
Faith Forster:Interesting question. I've assumed real people. The product has different personas supporting the users at different points depending on what they're doing at that time, but my thinking was get real people involved.
Randy Silver:No. I'm still I'm still very much there. I think people are agents are gonna be interesting and artificial personas will be interesting, and they are useful up to a point, but they will never surprise you as much as a real person will. A real person is gonna be your user or your customer, then you absolutely need to eventually go to them or have people who will think like them. This has also always been the problem of building something with only the people in the ivory tower and not getting out of the building, getting out of the head space.
Randy Silver:We make assumptions very, very quickly. We lose perspective very, very quickly, and I don't know that agents are gonna be any better at that.
Faith Forster:Well, and we have biases, and by nature, agents will have those biases built in as well. And so you do need that challenge. It's rarely about finding the right answer. It's not very often there is a clear right answer. There's a thousand ways you could solve any particular problem.
Faith Forster:It's more around getting the insights from your customers and your users and the different perspectives and aligning people around a decision on how you're going to proceed. And I think that's an inherently human thing to agree to a decision or not.
Randy Silver:How we build and how we deliver is fundamentally changing, and that's what's really fascinating. Something I learned from you a couple years ago was a two by two that you were using at decks around trying to help get people to launch things as quickly as possible if the cost was low enough and the risk was low enough. So how do we slice things down to the minimum viable learning increment with a low cost and low risk? But the underlying infrastructure, the underlying assumptions that we have about how long does it take to build something, how much does it cost to build something, what is the risk of trying something, they're changing incredibly. I still work with some companies that are extremely old school that barely use agents or AI at all, and it's fascinating to see in their early stages of evolution on this.
Randy Silver:And then I work with other people who are moving at the speed of thought, and it's fascinating to see. But the philosophy doesn't change very much. It's just how they use the tools.
Faith Forster:Yeah. I think one of the risks we run is thinking that it's quick to build, so I'll just build it without properly understanding what's the problem you're trying to solve. Is this the right problem? Have we thought about it in the right way? Are we designing the solution in the best possible way to solve that problem?
Faith Forster:I think that process of making decisions on what to build is the next bottleneck in that there's plenty of examples of feature heavy products that haven't really solved the problem very well. And actually that creates more problems. The more features you have, the more difficult it becomes to do anything. And so I think there's a real risk that people will think it won't take a lot to build, I'll just build it anyway, without really having understood whether it's the right thing to build. And so I think that relationship in terms of understanding and managing those risks is going to change.
Faith Forster:I don't know what to yet. I think there'll be some hard lessons learned around, oh, we'll just build it because it's quick. And actually, that's not the right thing to do.
Randy Silver:Probably or possibly apocryphal story about the ceramics class where a professor decided to try something and gave one class the objective of create one beautiful object by the end of this semester and the other one was create as many as you can, just do volume. And the ones who did the most, it was the application of the craft that led them to create higher quality, better objects by the end of the semester. So, yeah, it is an interesting thing. Strong opinion, weakly held, just build it as an experiment, learn from it quickly. Maybe that is useful in some cases or in lots of cases.
Randy Silver:In other cases where there is high risk, we do need to be really careful and do things differently. And I don't know that there is a single right answer for this. I think it depends on how seriously you're taking the thing that you build, how critical it is, what's the level of risk associated with it.
Faith Forster:That was actually one of the things that came out of the discussion we had for the lunch with Anton from Lovable is that everything's a prototype now. It's so quick to create something. There's a mind shift that has to come around with that, that it's okay if we don't use this. We can throw it away. It doesn't matter.
Faith Forster:There's always been a balance in terms of the amount of time we invest in Discoveree or working out what's the right thing to build versus actually building. In the same way we have plenty of waste in the build itself, there's plenty of examples of teams investing far too much in Discoveree. And particularly in, I'd say, more corporate environments where you've got senior leaders making decisions on what to build, like people get a bit wedded to certain ideas. I think there is a big buying shift around, yeah, we can build this, but equally, we might throw it away.
Randy Silver:Actually, something that popped up while you were saying that in my head, and when everything is a prototype, that works really well when you're starting from scratch, when you don't have a whole lot going on or where you have a nice platform and you can try new things on it. But there's a lot of old school stuff, a lot of very old infrastructure that you can't prototype on. It's hard to change. It's creaky. It requires a lot of love and care.
Randy Silver:Do you think this philosophy works everywhere or is this only in places that either have an incredibly robust, flexible architecture or starting from scratch?
Faith Forster:I think tech businesses who are using legacy stacks are gonna, at some point, have to face into the question of do they rebuild it? Do they start again for an AI native architecture and world? Because I think the longer they leave that decision, the weaker a competitive position they're going put themselves in. However, there are vastly different types of technology and the complexity that sits behind those can make that quite difficult to do. If I think about, I've worked in a few different payments businesses, the payments infrastructure behind that.
Randy Silver:There are lots of companies that want to sweat the asset because it's out there, it's valuable and rebuilding is, you know, it's cheaper than it would have been, but it's still expensive. And it's what you've got is trusted and tested and tried and true and all that.
Faith Forster:There's a lot of challenges that come with this. Anything that's regulated will be challenging, but I think there is a very real competitive threat as well from AI native businesses that at some point, legacy businesses are gonna have to face into.
Randy Silver:So you would go back to what Facebook did of just buying their competitors. Can the established companies buy up the AIs, do they need to build from scratch? I don't know. I don't have an answer to this, and this is probably a whole different discussion.
Faith Forster:It is very interesting. No one has the answers to this at the moment, but it has become very real for me in the last week, just thinking through as a result of that lunch I had and the questions that came up and thinking about what does that mean for how I build Discoveree, given that there's a lot of businesses who are not there yet in terms of AI agents building up their products versus some who that's all they do. And how do you support or cater people where they're at and where they're heading? How do you get that balance right?
Randy Silver:Well, I think we've done a good exploration of this today. I'm curious. Before we sign off, what's on for this week? What are we gonna check-in with you about next week?
Faith Forster:The priority at the moment is absolutely getting the beta version ready. We've got product con this week, so there's a few distractions, but, yeah, heads down getting the product towards next version and then getting some feedback on that.
Randy Silver:Fantastic. Well, I look forward to hearing about that, seeing if the conference is good and what you learn, and we'll check-in again next week.