# The Six Traps of AI Adoption (and How to Get Out of Them)

Faith Forster, Just Product 2026, 11:00. Automatic transcript of the livestream (Parakeet ASR, unedited, expect misheard names). Timestamps are positions in the talk video.

[0:01] So next up, last one before the lunch break.
[0:10] We have faith.
[0:12] And faith got introduced to me and Thomas by Zimo Netta, who is also here today.
[0:20] But maybe she is not on the stage.
[0:23] So uh Simonetta introduced us because uh Faith was putting together a group of um product people, engineers and designer because she had a question.
[0:36] She and the question was mostly about okay, so what is AI actually changing about product?
[0:44] Right?
[0:44] And she put together a group to essentially create a manifesto that is now out and that she will also share in a bit, called the makers manifesto.
[0:53] And the manifesto is trying to answer two questions.
[0:57] So, AI is here.
[0:59] What remains, and what will completely change.
[1:05] And today she's gonna talk us through the four traps of AI adoption.
[1:10] So please welcome the founder of Discovery.com and the initiator of the Make Us Manifesto, Faith Forster.
[1:34] It's Monday morning, you open your laptop.
[1:38] The engineers or their agents have built three features over the weekend.
[1:43] They're in staging, they look great.
[1:47] No one's quite sure who asked for them.
[1:52] Your Slack is already full.
[1:54] Sales are asking when can they demo them?
[1:57] Your CEO has seen a release video from one of your competitors and is asking why are we not shipping faster?
[2:06] Can I get me?
[2:08] Sorry.
[2:23] AI has made the ability for engineering to build much, much faster.
[2:28] And this has created a lot of pressure on product to keep up.
[2:36] So people are taking shortcuts.
[2:39] Or they're not doing discovery at all and jumping straight to the build.
[2:43] In the excitement of being able to build all these shiny toys that we couldn't quite fit in before, we are forgetting to stop and ask should we?
[2:55] It feels like progress.
[2:57] It isn't necessarily.com.
[3:12] And I've been hearing versions of this Monday morning story all year.
[3:17] Back in February and March, I was looking for some beta testers, and so I reached out to a few product leader WhatsApp groups and ended up speaking to 96 product leaders about how they're using AI.
[3:30] Two things struck me from those conversations.
[3:34] The first the level of understanding was so wildly different, not just to the technology itself, but the implications it would have.
[4:25] Been thrown out.
[4:26] We needed a new one.
[4:28] And so I gathered a global group of 45 cross-disciplinary experts and operators, product people, designers, engineers, marketers, founders, some of whom are in the room, Sebastian Thomas, Martin, you'll see Dave later, Arnie and Simonetta here too.
[4:47] And we created the next version of the Agile Manifesto for an AI age.
[4:54] We called it the Maker's Manifesto.
[4:57] Four values, 16 principles.
[5:01] Today I'm going to talk to you about some of the traps I've seen when businesses are adopting AI and how to use these principles to help you get out of them.
[5:13] When Sebastian and I had a chat about doing this talk, we decided on four traps.
[5:17] I thought that sounds like fun.
[5:20] But a lot like the manifesto, which we desperately tried to get down to 10 principles and ended up with 16 because it just wasn't quite enough.
[5:29] When I started thinking this talk through, four traps also wasn't going to cover it.
[5:35] And so actually, we're going to talk about six traps.
[5:38] Six kinds of thinking.
[5:40] That we skip when we rush to keep up.
[5:45] And the principle that gets you back on track for each one.
[5:51] So if there's one thing you take away from today, it's this.
[5:55] Don't rush to keep up.
[5:58] Rethink instead.
[6:04] So when we're racing to keep up, the first thing we skip thinking about is value.
[6:10] We measure activity instead.
[6:13] Because it's easy to measure, and it proves we're doing stuff right.
[6:18] So that's called token maxing.
[6:20] Competing to burn as many AI tokens as you can.
[6:25] The principle that answers this is number seven.
[6:28] Measure what drives value.
[6:32] What we choose to measure shapes what gets built.
[6:37] You know you're in this trap when there's a leaderboard of who uses the most tokens, and your job is dependent on where you sit in that leaderboard.
[6:47] The percentage of code written by AI is the headline metric on the board deck.
[6:52] Pull requests and velocity have increased three times.
[6:58] But when you ask what additional revenue you've got from that, or what value customers are seeing, there's silence.
[7:25] Leader board that ranked the AI usage between their teams.
[7:33] The whole, not surprisingly, everyone started using tokens, doing whatever they could to chew them up.
[7:39] Their whole year's budget was gone in four months.
[7:45] When Andrew McDonald, the CEO came out and talked about it, he was asked the question, what value did that deliver to your customers?
[7:55] And his answer was this.
[7:57] That link is not there yet.
[8:00] There was no quantifiable value for their customers.
[8:07] Thankfully, he said this publicly because that was the moment that most leaders stopped and started asking, or stopped celebrating token usage, and started asking, what is that actually gonna, what is that buying us?
[8:21] I'm now hearing a very different conversation coming out with executives and board members as well, shifting towards what outcomes are we achieving for that money spent?
[8:32] Which as product people, we always knew is what we should be doing, of course.
[8:39] So the second trap, the abandoned junior.
[8:43] When we're racing to keep up, we skip thinking about what good looks like.
[8:49] We start writing prompts and we see what happens.
[8:53] There's no time to set the AI tools up properly.
[8:56] So when something gets wrong, we just fix it ourselves.
[9:01] The principle that answers this is number six: make context explicit.
[9:06] Good judgment needs clear intent.
[9:09] Direction, trade-off, constraints.
[9:13] Without it, what we optimize for drifts.
[9:17] Or your agents are only good as the context you give them.
[9:24] You know you're in this trap when you hear things like, it doesn't write the code the way we like it.
[9:30] The documents just not very good.
[9:34] So we just do it ourselves.
[9:37] Now I heard every single one of those statements when I started recently on the board of a HR Tech Scale Up.
[9:45] They'd tried AI, hadn't worked for them.
[9:48] And so the justification, that was justification for people going back to how they'd always worked.
[9:54] The AI wasn't the problem.
[9:58] The way we like it had never been written down.
[10:02] It lives in people's heads.
[10:05] So you think about it, you've hired a brilliant graduate, you give them no onboarding, no examples, no feedback, and then by Friday you decide they're not good enough.
[10:18] That's exactly what we're doing to our agents.
[10:22] With AI, the role shifts from building the feature to teaching the agent how to build the feature well.
[10:34] Onboard it, give it feedback, write down what good looks like.
[10:39] As product managers, this means sharing all the rationale behind why we chose to build that feature, as well as the outcome we're trying to achieve by doing so.
[10:51] If you do this, it very quickly becomes a superpower for your whole team.
[11:01] When we're racing to keep up, we skip thinking about adoption.
[11:07] The next fixture is waiting to be built, so we call it done and we move on.
[11:12] The answer to this principle is number nine.
[11:10] Make for adoption.
[11:18] Distribution, adoption, monetization, retention, they don't come after the making.
[11:24] That is what making is now.
[11:29] The build is no longer the constraint, and so we need to stop treating it like it is.
[11:34] We can continue to completely rebuild the feature or the product until customers love it.
[11:43] You know you're in this trap when the fixture goes live and the party pops on Slack.
[11:51] Three months later, no one's actually using it.
[11:56] Launch is the last line on your plan, and nobody knows what's going to happen after.
[12:02] Adoption is someone else's job.
[12:05] Marketing, customer success.
[12:08] Someone outside of the product org.
[12:41] Yeah?
[12:41] Not many people, interesting.
[12:43] They are painful.
[12:45] At Dexter took us nine months to redo our pricing and packaging.
[12:50] At OmniO, the thing I love about this story is they released their first product within three months.
[12:56] And then they spent the next six months refining it.
[13:01] They played around with the packaging mix, they looked at the revenue impact, they played around with the way they surfaced it to different audiences until they got the profitability.
[13:11] So, as well as the subscription uptake, they also increased the average order value and the number of tickets per customer.
[13:19] So the customers were coming back again and again.
[13:24] The great thing about this story is not only did they deliver a product that customers loved, it also had very meaningful commercial benefit for the business.
[13:39] When we're racing to keep up, we skip thinking about what customers are ready for.
[13:46] So we've embraced AI.
[13:48] We're running at the speed of light, we ship feature after feature after feature.
[13:54] And then the problems start.
[13:56] The customers start complaining.
[13:58] They can't keep up.
[14:03] The answer to this problem is principle number eight.
[14:06] Match pace to readiness.
[14:09] Our capacity is accelerating faster than customers can consume it.
[14:14] What they're ready for is where value is won or lost.
[14:20] Your customers have day jobs.
[14:22] Our software is there to help them, not distract them.
[14:29] You know you're in this trap when the customers are closing those 50 what's new pop-ups that are coming up every day.
[14:36] Your sales and support teams find out about new features from customers.
[14:40] The same customer is in three different betas at once.
[14:46] Customers can't find the thing they use every day because it's moved yet again.
[14:55] So John Moratori is a design leader at Intercom Fin.
[14:59] We're gonna go into the Intercom story again.
[15:01] So they did amazing.
[15:03] They went all in on AI-led development 12 months ago because they had to.
[15:07] They're in a do-or-die situation.
[15:09] And they managed to increase productivity over three times.
[15:14] And they got to the point where they shipped nine tier one releases within two months, which is amazing, right?
[15:22] That's the dream.
[15:23] That's what we're all trying to do.
[15:26] Until the customers started complaining.
[15:29] They couldn't keep up.
[15:30] The product didn't make sense.
[15:32] There was too many random things being added in.
[15:36] So yes, humans are the bottleneck, but often
[15:40] Not the humans within your own business.
[15:43] It's your customers.
[15:45] And so now they become the benchmark for the pace that you should be aiming for.
[15:54] When we're racing to keep up, we skip thinking about consequences.
[16:00] We let the AI work it out.
[16:02] It can decide we just hope for the best, or we ignore the gaps.
[16:07] In Australia, we had a joke at school about if you hadn't done your homework, the dog ate it.
[16:12] I hope that translates here.
[16:15] So AI is now doing our homework for us.
[16:19] The principle that answers this is number 16.
[16:22] Humans sign the work.
[16:25] Humans must be answerable for the consequences by design, not by default.
[16:31] You can delegate decisions, but you cannot delegate accountability.
[16:44] An AI generated statistic that nobody can source ends up in the board pack, which happened to Deloitte in Australia and they had to repay half a million dollars to their client.
[16:56] The post-mortem says the root cause was the model that did that because the humans hadn't put the right guardrails in place.
[17:24] When he claimed it, Air Canada said no, they don't pay out that discount after the travel.
[17:31] And so he took him to a tribunal, and Air Canada tried to argue that it wasn't their fault, it was the chatbot's fault.
[17:41] The tribunal was astounded.
[17:44] The comment was this is a remarkable submission.
[17:48] It said the chatbot is still part of your website.
[17:51] Surprise, surprise.
[17:53] And they had to repay the difference.
[17:55] So it was 650 Canadian dollars.
[17:58] It's not a lot in terms of money.
[17:59] But the PR impact was huge.
[18:03] A grieving customer, company trying to blame its chatbot.
[18:11] When we stop to rethink, when we never stop to rethink, we also skip thinking from first principles.
[18:20] What problem are we solving?
[18:22] Who for, and how do we make money.
[18:27] Instead, we just bolt AI on top.
[18:30] Same product, same roadmap, with some AI magic in front of every line item.
[18:38] Two principles answer this one.
[18:41] The first, understand the problem worth solving.
[18:46] We now have this amazing technology available to us.
[18:49] The way we solve those problems can be fundamentally different.
[19:00] Greater capability and capacity is an invitation to deliver on greater ambition.
[19:08] You know you're in this trap when every screen has a sparkle icon.
[19:13] There's a summarized button on things nobody wanted summarized.
[19:17] Your roadmap is the same, but with AI added to every line.
[19:25] The winners in the next few years will be those who go back to first principles and rethink how do we solve this problem now in a fundamentally different way.com have done this.
[19:39] They went in all in on AI early.
[19:43] They completely reimagined not only how they work, but the whole product experience and in time the range of problems that they solve for their customers.
[19:53] And as Martin said, in June, Salesforce agreed to buy them for 3.6 billion.
[19:58] That is a seriously impressive turnaround from do or die in under 12 months.
[20:05] Later today, you're going to hear from Dave.
[20:07] Pendo of completely rethinking how they provide analytics for agent-based products with Novus.
[20:15] And I've done it myself.com.
[20:21] It is now a shared context layer or your team's product brain.
[20:26] It gathers all the different information you need to inform your product decisions, structures it, cites it, make sure it's audible, and refresh and surfaces it in whatever tools your team are using throughout the product lifecycle.
[20:41] That concept barely existed only months ago.
[20:45] It took me six weeks to get from idea to first customers, someone else will.
[21:00] When it's so easy to build, that is a very real.
[21:06] So don't rush to keep up.
[21:09] Rethink.
[21:14] So, six traps, six kinds of thinking we skip when we rush to keep up.
[21:21] Token maxing, the abandoned junior, it's live.
[21:27] The fire hydrant, the AI 8 my homework, and lipstick on legacy, my personal favorite.
[21:35] And for each one, the principle that helps you get back on track.
[21:39] The build is only gonna get faster.
[21:42] So the most valuable thing you can do is not to rush to keep up with engineering, but to step back and help your team rethink how you solve these problems.
[21:55] So I'll leave you with a question that every makers that we talk about at every makers manifesto workshop.
[22:01] What's the smallest step you could take tomorrow to put this to work for you, your team, your organization?
[22:12] There's a few things I've seen teams do with the Makers Manifesto.
[22:15] So you can give it to your agents.
[22:17] When you go to our website, you can download the version for humans and the version for agents.
[22:23] This becomes part of their context about how to build great products.
[22:30] Initial feedback is it has made a difference.
[22:34] Use it as the basis for your next retro.
[22:38] The four values make four good areas for you to consider how you improve.
[22:44] Run a session with your team, pick a value and ask Are we skipping the thinking on this?
[22:51] Or name your trap from today.
[22:54] And take this back and use it to do do things differently.
[23:00] Whatever you choose to do, start here.
[23:03] Download the manifesto, share it with your team, argue about it.
[23:08] That's principle four.
[23:10] 14, sorry, challenge by design.
[23:13] Difficult problems deserve disagreement.
[23:17] And help your organization be one of the winners in this crazy AI world we find ourselves in.
[23:24] I'd love to hear how this changes the game for you.
[23:27] So please do share your story.
[23:29] You can do so via the website or you can find me on LinkedIn as well.
[23:35] Thank you.
[23:41] Thank you so much, Faith.
[23:43] Thank you.
[23:48] Yeah, great.
[23:50] All right.
[23:51] So do you have any questions for Faith?
[23:55] Oh, that like that was quick.
[23:56] That was very quick.
[23:58] It's always just like the first time, it's like very slow, and then it's it's getting quicker here.
[24:04] And please make use of the cube, but please say something before you throw it so we can prepare everybody.
[24:13] Good.
[24:21] Thanks so much for the talk.
[24:23] I'm just wondering whether you could name a couple of tips on how to solve fire hydrant problem potentially, because I see it in our organization that even internal teams sometimes have trouble to keep up with our speed.
[24:40] Yeah, yeah, yeah.
[24:43] So we talk we talk quite a lot about the entire product life cycle was built around the fact that the build took the longest bit, it was the biggest risk.
[24:52] That's shrunk, and so it's put pressure on either end of deciding what to build and then actually gaining adoption with that.
[25:00] It is I personally think of all the principles that we have in Manifesto, this is the biggest mindset shift.
[25:07] If if it's no longer about shipping, if the definition of done is shifted from shipping to adoption, that is a whole different way of working.
[25:17] If you put the question of adoption in the center of everything that you're trying to do, the way you organize your work, the way you organize your teams, everything shifts around that.
[25:28] So in my mind, that's the biggest challenge.
[25:31] It's also the most important within our world today.
[25:36] You can, and we had a comment earlier about we no longer have.
[25:40] Time to do quite as much discovery, we're doing less of it.
[25:43] That's okay if you're building to learn and take the discipline of learning with that.
[25:49] It is a very different discipline.
[25:52] And so you can build, you can build features in MIL products really, really quickly, but don't assume that that's the finished product.
[25:58] Don't assume that's right.
[26:00] You do need to work really really closely with your customers to come around and work out kind of did that work?
[26:06] Have we solved the problem right?
[26:07] Is that the right problem?
[26:08] Is it getting the impacts that we hoped it would?
[26:10] Um, and having that clear understanding of kind of what was the impact we were trying to achieve in the first place.
[26:16] But it is it's a very different approach, and so it's something that um from what I've seen across a number of organizations and friends that I hear from, it's not until the customers start crying out saying stop that people actually kind of go, oh, hang on a second, we need this is not good, we need to think about this a bit more.
[26:35] And it is it comes back to like how you think about your roadmap, how you think about your planning, how you organize your teams, and so it is it is something that has to come from executive level in terms of how we think about the way that we support our customers better in this environment.
[26:49] I don't know if that helps because if you're the PM trying to get a bunch of stuff out because you're being told to, it's not easy conversation to have.
[26:56] That's fun.
[26:57] Thank you for sharing.
[27:00] All right, who's next?
[27:03] Oh, that's uh oh, you can also throw through like back through it, throw it.
[27:09] Yay.
[27:12] Thank you for the talk.
[27:13] Um we heard before that BMs might be in a strange situation where the backlog is empty and the developer asks what to do next.
[27:25] I understand here.
[27:27] Say, stay, wait, think, rethink.
[27:32] But how do you deal with that challenge of uh hey, what should I do next?
[27:38] Yeah, do you have a similar list for the developers then?
[27:44] Uh to kind of pace down, slow down.
[27:48] I'm I'm curious what do you think about it.
[27:51] Yeah, I think I think you need to um so going back to my own leadership roles, I guess, in organizations.
[27:56] If I when I joined and we hadn't quite worked out the product strategy or what we wanted to focus on, the advice I always gave is kind of what is your no regrets moves.
[28:04] So what are the things that you know need to be done and there's like not there's low risk that it can be wrong.
[28:11] Um and so having an understanding of those helps you sort of fill some of those gaps.
[28:15] But I think I think also we need to stop thinking about it as like PMs and designers do all the work up front and then it's handed off to the developers, and so you're kind of rushing to feed them.
[28:27] Uh I think you can get the developers like you can build to learn, you can get the developers involved in creating things quite quickly to work out kind of what is this problem?
[28:38] What are all the different ways we could solve it?
[28:39] Because you can create 10 prototypes now where it was expensive to create one before.
[28:44] Um you can create a working version of the product and get internal users or your you know your advocates or your your close-knit customers to help you.
[28:52] So I think we need to stop thinking about it as this sort of handoff and putting pressure on ourselves to have all the answers before we start talking to developers.
[29:02] Okay, thanks.
[29:05] I think we can take one more question.
[29:11] Preferably in the front, because I really like when the thing is thrown.
[29:21] These lights are super bright.
[29:23] I know.
[29:23] It's crazy, isn't it?
[29:26] The tech folks did a really good job because now you can see us, but for us we can't see you.
[29:34] Which is also sometimes helpful if you're super nervous on stage, as I was this morning.
[29:39] So okay.
[29:41] Otherwise, definitely feel free to grab Faith.
[29:45] Um, because first of all, I want to say so.
[29:48] We are also supporting the Makers Manifesto.
[29:51] I think it's a great thing.
[29:52] Um, and I think this is still not like done.
[29:55] So it's like it's also not something that that the group is pretending that it's done, but it's a living document that we have to improve because who knows where we will be in one, two or three years, and so uh feedback is is very helpful and always appreciated.
[30:12] Um and with that, thank you so much, Faith.
[30:15] Thank you so much.
[30:16] Have a one for you here.
[30:20] Perfect, thank you so much.