Hoe AI je kan helpen betere projectbeslissingen te nemen

Galen Low: In some ways, projects are just a collection of decisions. In fact, decisions are a big part of what drives projects forward. But how can we expect humans to make good decisions when projects are moving at speed, and there's an overwhelming amount of information and data to assess? Definitely, AI can help, but it doesn't always have the context, and making it accountable for the decisions isn't an option anyways.

Meanwhile, sometimes the snap decisions based on gut instinct turn out to be the best ones. So what's the right balance between data-driven, AI-assisted decision-making and just talking it out? Where are there diminishing returns, and more importantly, how do we actually make decisions? To dive into that, I've brought in an AI and knowledge management specialist whose background has included everything from psychology and enterprise estimation to entrepreneurship and product ownership roles at Fortune 50 companies like Meta and ExxonMobil.

Together, we're gonna walk through some specific examples of how AI can support quality decision-making, how traceability can help humans get better at making decisions, how accountability changes with AI in the mix, and what becomes of the role of the project manager if AI is facilitating the decisions that used to cross our plate.

Hope you enjoy the episode.

Welcome to the Digital Project Manager Podcast—the show that helps delivery leaders work smarter, deliver smoother, and lead their teams with confidence in the age of AI. I'm Galen, and every week we dive into real-world strategies, emerging trends, proven frameworks, and the occasional war story from the project front lines.

Whether you're steering massive transformation projects, wrangling AI workflows, or just trying to keep the chaos under control, you're in the right place. Oh, and if you've been liking what you've been hearing from us lately, please consider following us wherever you're listening or watching, and maybe even leaving us a review. All right, let's get into it.

So today we are talking about AI-supported decision-making and why decision quality might be the thing that makes or breaks the success of a project or initiative today. We're gonna be unpacking how disconnected decisions create risk, how project leaders can improve decision quality across complex teams, and how AI can help us make better decisions, and vice versa.

With me today is Cali Collins, co-founder and chief product officer of Optimality. Cali is an experienced product leader whose career has spanned AI infrastructure, knowledge systems, enterprise transformation, and large-scale capital project execution. Before co-founding Optimality, she led product development at organizations including Meta and ExxonMobil, giving her a unique perspective on how decisions get made or don't get made inside both fast-moving tech environments and complex Fortune 50 enterprises.

Today, with Optimality, she has been championing the value of having an AI-native decisions infrastructure platform that helps organizations structure, optimize, and scale decision-making in an era of overwhelming data and competing priorities.

Cali, thanks for being with me here today.

Cali Collins: Absolutely. Thank you so much for having me.

Galen Low: I'm excited to have you on. I'm gonna start out with a bit of a spicy question, and I'll take a running start at it. So many organizations have actually very little structure or discipline around how decisions get made, and sometimes consequential decisions are made in an email chain or verbally on a golf course.

And mostly teams are just too busy reacting to those decisions to pause and think, you know, "Is this a well-informed decision?" They're just kind of, going with it. So my question is this: Given that we as humans can't really explain to one another how decisions get made and why, is it not maybe a little bit delusional to think that AI can suddenly improve organizational decision-making?

Cali Collins: I mean, yeah, absolutely. I think a lot of it comes down to bias. If you ask me why I made that decision, the thing that I tell you maybe isn't actually why I made that decision that way. We all have our own perspectives. You have your perspective of how you would hear me repeating that decision back, and I think we were also talking a little bit in the prep, too.

Everything you do is a decision. You can't document everything as you're doing it. I made a decision to log into this call. Am I gonna go then write that down? 'Cause I have to make a decision on when I'm gonna write down the decision that I just made. Our lives are a series of decisions, and by having something like AI that's automatically capturing it as you're doing it and you don't have to consciously document it, that's where it's gonna help us out.

Galen Low: I really like that because you start out with bias, and I'm somebody who believes in the fact that bias can be good and bad. We do have, you know, unconscious bias that definitely is holding us back as a species, but there is just the bias of knowing what you know. And yeah, we talk a lot about sort of making decisions on gut instinct and whether that's good or bad, but fundamentally, a lot of, at the time it's like, okay, especially, you know, at decision maker, quote unquote, levels, but also anyone, it's yeah, it would be arduous to have to explain every little bit of it, but it is happening there.

You know? I think very few decisions ... Actually, as I say that, I'm like, a proportion of decisions are mostly made with some logic behind them, even if people aren't showing their work. Yeah. Sure as people are listening, they're like, "Yeah, but what about that decision where, you know, that snap decision where, you know, things went wrong?"

Definitely there's that too, but yeah. And I think it's like we kind of brush over it because, yeah, we don't have, A, the time to sort of map it all back and explain every single thing. We have to trust, you know, decision makers to make good decisions. And also, you know, there is a lot that goes into it, and that isn't necessarily, you know, the worst thing in the world.

Cali Collins: That's absolutely true. Yeah, I mean, there's science between what part of your brain do you use to make different types of decisions. It's a different part of your brain that's making those quick, rash decisions that are based on instinct versus those ones that you give some time to think about. So it's interesting.

Galen Low: I really like that. And actually, you know, maybe that's a good place to zoom in because I like the sort of two modes of decision making. One, you know, arguably there's like this sort of, I don't know if it's fight or flight, but it's like the urgency of making a fast decision might bypass, you know, all of the sort of logic and underpinnings that we have.

And then coming back to that sort of like notion of, I don't know, I don't know if safety is the right word, but the state of calm, right? To make an informed decision by evaluating, by analyzing. You know, in a way, like having the information available sort of brings us back down to that level to not make snap decisions because we know we can sort of take our time.

We have, you know, information available to us that can come to us rapidly, and we can make that decision, you know, in an informed way. Maybe let me zoom out even further than that, because what I like about your experience is that, you know, you come from a product background. You are a deeply experienced product specialist who has taken arguably a bit of a left turn away from working for organizations like Meta and ExxonMobil to build basically an AI native decision operating system.

I hope I can characterize it as that.

Cali Collins: Yeah.

Galen Low: So I thought maybe I'd just ask, like why decisions? Like, where have you seen faulty decision making, like crash a project or a strategic plan, and like what are some of those like underlying causes that like compelled you to design a solution to that?

Cali Collins: Everything you do is in decisions. It's a human-wide problem, and I wanted to do something that really would affect how we work on a day-to-day, really make a big impact. And like you were starting to get into, there's those data-driven decisions that you think about, and then those immediate quick decisions.

And you said safety, and I started thinking safety a little different because being from a project background, I'm thinking like HSE safety if somebody's about to get hurt.

Galen Low: Right.

Cali Collins: There are certain decisions if you see some unsafe activity, you gotta jump in there and act. But there's other decisions when you're stepping back that you gotta analyze through and think a little bit more on it.

And just understanding and being in... I did estimating many years. I started in estimating. Always had a product side, too, but estimating was my day-to-day job. And so I had exposure to a lot of senior executives, a lot of leaders. I was at the negotiation table when we were talking through it, and understanding what goes into those big decisions.

And I would see these executives, and people are putting them on a pedestal, and they're just like, "What do you think? What should we do?" And once I had a couple of chances to actually sit with some of these executives one on one, and I was like, "How do you know the right decision? Is it all this data that you have?

When is the diminishing or point of return?" And even these guys are like, "You know what? I'll tell you a secret. Half the time, I don't know. I'm just kind of winging it." They're like, "I'm in this position because I've gotten to it because I've made really good decisions." But at the same time People expect that, and they don't really always have the knowledge and the data to be able to make the right decision.

And I was like, if I could help empower these people that are making these huge changing decisions of do we spend these billions of dollars on this or on this, and there's... it's such an important choice that they're making, how can I help them understand where that, that point of diminishing return is?

Where, where does the data get you? And give it to them quickly so that they can make those quick decisions, rely on their instincts at the same time of having decision data.

Galen Low: I love that moment of vulnerability. It stands to reason that it's, you know, similar for a lot of us, where we're like, "I don't know I'm under pressure to make a decision.

You know, I've made some good decisions before. I have zero guarantee of always making the right decision, but my probability, you know, my you know, my batting average is pretty good, and that's why I've gotten to this role. But I don't know. Sometimes I'm just winging it." And I love that idea of you don't have to.

Not that your decision quality is poor already, but could we have a better, you know... I keep using the word batting average. I'm thinking baseball right now. But you know, could we have a stronger probability of making the right decision? And frankly, even just that anxiety, right, of being like, "Gosh, I'm in this position."

Maybe a bit of imposter syndrome, but definitely pressure. You're talking billions of dollars sometimes, and you're like, "I don't know. I have to make a decision, and, you know, I've got to pull on what I do know. But also, I can support that with other data to inform my decision." And I think that's, you know, the beauty, I guess, of AI right now in terms of its ability to kind of gather information and data together in large swaths and distill it down based on what we need.

And maybe that's a good segue because I love the idea of Optimality. I wondered if maybe you could tell us a little bit about what Optimality does, how it helps, and gosh, maybe some of those use cases that remove some of the risk in decision-making, or at least sort of minimize it.

Cali Collins: Absolutely.

Yeah. We've been working on Optimality for about two and a half years now, and so it's obviously been transforming and really growing into what is Optimality going to be. While the core was always that improving decision-making and it became into optimizing decision-making, which is really where the name Optimality came from, there is an optimization theory mathematically that is a little bit of that diminishing point of return.

What is the optimal decision that should be made? And so we really have embraced that on what we're working on. And the idea is that Optimality sits on a context graph. So when you are collaborating in Optimality, you are using the system to understand, what should I be working on today? What did I do yesterday?

But it's so much bigger than that. It's the actual content and deliverables that you're producing. It includes all of the decisions that you're working on. We actually just released today, a few hours ago, a new decision wizard, where if you do have a deep decision that you need to do, you can get in there, and it walks you through kind of the step-to-step process of doing the decisions.

But while you're working through this wizard, you have Opti Chat next to you, and you can ask Opti. You can say, "Well, I'm thinking between this and this. What do you think, Opti?" And it helps you frame that. And then you get into the next thing, and it's like, "Well, then what are the uncertainties around this decision that I'm trying to make?

Where does my personal bias, or when did we make a decision like this previously, and how did it work out?" You always have the power of Opti right there to chat with to understand, and he's or they are looking into your context graph to really want to understand when have you worked on this decision before?

You're mapping out your organizational and your execution intelligence into the system. It's constantly absorbing all of that information into Opti's brain behind the scenes- Hmm ... to grow smarter and be able to provide you with better insight when you need to know, do I do A or B, or is there something on a option C I haven't even thought about?

Galen Low: I really love, even just to take it all the way back I love how nerdy the name actually is. Not just optimizing decision-making, but there is mathematical theory for where you hit that point of diminishing returns on decision-making. I think that's super cool. And then, yeah, you had schooled me as we were prepping for this on this idea of the context graph being the sort of neural network really for some of the AI tools that we're using.

I believe you used the word decision context, and I'm really sort of getting a sense of it now, right? Of something like Opti's absorbing, you know, all this information along the way. Not like, "Hey, we need to make a decision. Can you please gather all of the stuff and go ask these people?" You know?

That we need to put together advice for a decision maker and that could take days. It's actually just kind of there. I'm wondering even outside of AI, for decision context what does that actually mean in a project environment? And what are teams sometimes missing from their decision-making process that this sort of helps support?

Cali Collins: I think it's just the human brain is great at certain things, and the human brain also has limitations. You cannot be in every single meeting gathering all of the data that everyone's speaking on. Imagine if you could have a delegate that's in these other meetings absorbing this other information too.

That's where the AI side steps in. But in the human side, we're based on collaboration and communication, and that was a big part that took me to Meta because they're trying to build a collaboration platform for the world. People go on there to share and Become part of a community. And that's just such an important part of being a human, which is weird saying that now with this whole AI versus human world.

But being a human, my power is to be able to get together with other people and bounce ideas off of them, and grow and change my thoughts based on the things that other people tell me. That's where you still have to keep the human in the decision-making party. It's got to be where we've all had our own experiences, and we need to be able to collaborate to really come up with what is the best decision that can be made here.

Galen Low: I really like that the sort of interplay between it. The positioning is not, well, humans are bad at making decisions 'cause we don't have the capacity to absorb all this information, and we predict the future very poorly. Which are- Yes. ... arguably all true, but, you know, it's that interplay between the fact that, yeah, we are creatures that dialogue and, you know, have emotion and make decisions based on that, and we have technology and, you know, like today, specifically AI, to be able to do what our brain can't, right?

Pull together all this stuff, conversations that have been had, and distill it back down so that we don't have to sit in front of someone and be like, "Oh, remember that time where we had that 15-minute meeting on a Tuesday back in December, and you had said these exact words?" You know, we just don't have that, nor do we have the culture around that.

But I like this idea that it's like, you know, advising and informing, but not replacing the conversation of how are we gonna make this... Let's, let's talk it out, you know? And I think there's so much that is important about human collaboration where it's just let's talk it out so we can make a good decision.

But let's also support it with things that our brain just cannot do.

Cali Collins: Exactly, and I talk it out with Opti all the time. If there's anything- That's what, yeah ... it probably knows more, being in a startup, there's only a ... It's a small team. And so there are certain times where it's nice to talk it out with the AI, too, and let it bounce ideas off of you.

And it knows what you're working on because Opti's based on a context graph. It knows what you and everybody on your team are doing.

Galen Low: Maybe we can go there because I think, you know, for a lot of folks in my community, AI is, it's here. Let's just say that. AI is here. Yeah. It's a bit forced in a way, right?

It's like we haven't got a choice about using it. And then there's this sort of like also this Big Brother aspect because the nature of the technology is that the more context it has, the more it knows, the better it can do its thing, you know, do our bidding, you know, help us out a little. But, you know, I think folks listening will be like, "Oh my gosh, that sounds really scary that Opti's just listening to everything and taking it all in," and is this, you know, Orwellian in a way?

In the context of decision context and context graphing, is there a way to just choose what we share with AI to help it advise us better? Is it an all or nothing thing? And if it's not an all or nothing thing what are the things that we should be sharing with AI, even if it's not full access to every keystroke that we enter and every conversation we have through our computers?

Yeah, what are the right things to provide to AI to Help it advise us?

Cali Collins: So when we first started building Optimality, we API'd into my emails, we API'd into my chat. We started all of this, and then we quickly realized if we're trying to scale this, that's scary, and people want a little more control over what they share.

This is also going back to the governance side of AI. You don't necessarily want it taking your information, making decisions, and sending emails for you. Because if you're one of those people that gives Opti- or gives the AI access to your password manager and access to send emails and access to your bank account, you just opened up a whole bunch of mess.

And if something secret gets leaked out in that email The person who started that agent, who said, "Yes, you can do this," you are liable for it. You are responsible for it from a governance standpoint legally, even if you never saw that email before it went out. And so that is very, very important to understand and let you stay in control over what you give it and what you allow it to do.

And so that is why we set up where you can choose, at the end of this meeting, I want this transcript in there, or I don't. You can choose, I want this email conversation in there, or I don't. And it allows you to have more control over the good information you give it, too. Because when it's just trained on the garbage of the internet, AI doesn't know what is good and what is not.

And so if you're able to keep it clean by giving it just information that is good, that's going to help give it more power as well. Plus, we keep everything to be human in the loop, and so that way Optimality will generate documents for you. It can draft emails for you. It can status things for you. It can do all of these things, but there's still a person at the end saying, "That is what I want it to do," which is something that we do that's a little bit different than a lot of other companies that are wanting to say, "I can build you an entire marketing team, and they can just run wild and do everything for you and post on your social media and take over that job."

Instead, we think that there's more of a collaboration between humans and AI, how you can work together for good, where if you want to work on marketing, you put on your marketing hat, you have your marketing assistant with you who is a expert in marketing guiding you along the way, but you're still the one making those decisions.

Galen Low: I love that, and I think a lot of my listeners will love that as well. Again, it's that interplay between technology and our humanness. I wanna come back later. I wanna come back to that idea of curating what you sort of share into the context graph, because I think it's really interesting in terms of the way humans make decisions.

But maybe even before we get to that, you kinda framed a, a really good use case. I wondered if we can kind of dive into it. You know, this idea that okay, well, how can AI, how can Optimality, like, how can it help us make better decisions? I like that idea of we've gotta think about our marketing.

We've got a thought partner. We've got Opti we can chat with. Can we dive into that a little bit, just in terms of step by step, like, how it's all working, you know? In your case, Optimality working in the background, gathering information, and then we're like, "Okay. Well, we need to develop whatever campaign, or we have to do our LinkedIn post of the week," and sort of like what that interaction looks like and you know, how it's improved versus if we didn't have AI behind it.

Cali Collins: Right. Yeah, absolutely. So one of the use cases that we're moving into is a business operating system that sits on Optimality. So Optimality you can think of as a platform that enables all different great use cases, kind of like Excel. You can do any sorts of things in Excel, but it's not telling you, you have to use the formulas to make a budget spreadsheet.

It is enabling that ability. We have more structure than that. We have more structure than just Excel as far as that goes, but we also allow a lot of different use cases. So one of the fun ones that we're starting to work on now is a business operating system, and what that will allow you to do is if you want to

A lot of people are trying to start up side hustles, especially in this AI world where I can vibe code an app over the course of It's not a day That's not real. Over the course of a few weeks to make it good, depending on what you're wanting to do with it, and then you have the security aspects of a vibe-coded tool with garbage code.

But that's something different. But if you want to actually build your own company, it's a lot more accessible nowadays. There's a lot of room for people to be entrepreneurs. But maybe you know a lot about that type of app. You don't know about marketing or social media or building a business plan, or maybe you're really good at marketing, but you don't know enough about finances and taxes and all this.

And so what we're trying to build is the ability for you to go into Optimality, you talk to it about what are your goals, what are you wanting to do, and it's going to take that information, it's going to document it, it's gonna create that strategy for you and say, "Is this what you want to do?" You kind of talk back and forth with Opti a bit.

You say, "Yeah, that's actually kinda what I want. Save it." It's gonna map out your plan to get to your goals. And so it's just doing that automatically with the AI. You can go in, you can modify, you can do whatever you want with it. But it's helping you put some structure around building a business. Now, building a business is just one use case.

You can do construction of a refinery. We're doing with Chevron, for example as one of our clients. But working over toward this one I wanna build a business. And as you're going through each of the different steps in there, some of them will say, "Would you like Opti to assist you with your marketing plan?

Would you like Opti to assist you in developing content for this? Would you like Opti to do all these things that agents can do?" And you can work side by side with the AI, making it accessible. There's a lot of people out there that are running small to medium businesses or thinking about doing side hustles, and they're AI curious, but just don't really know where to start.

And how can we make it easy for people without opening Claude and trying to figure out how to run agents and burning through so many tokens, and I am an avid Claude user. I don't speak down of it. I love it. I'm using it every day. But at the same time, it takes a while to get there. And for most average people, that's not super accessible.

So that's just one of the use cases that we're working through.

Galen Low: That's awesome. And first of all, thank you for saying all of that. Originally, I was gonna be like, "Here's where Cali says, 'No shade to vibe coding.'" But I think you've really hit the nail on the head. The barriers to entry, for example, you know, building a business, are much lower, but there's also the hype of being like, you could vibe code an entire product and platform in a day.

No, you can't. But it probably is a lot- easier to do it than it was before, right? So there's middle ground here. And I'm thinking through it, and like you said, you said Excel is kind of like a, a very general tool. Optimality kind of like puts some of this stuff on rails. So if Claude was Excel, Optimality is kind of like, okay, it's a bit more specialized, it's a bit better suited to do a certain task than just like the general chatbot thing.

'Cause I think some folks listening will be like, "Cali, I can do that. I do do that in Claude. Claude helps me with my marketing. I just ask it stuff." But the other thing that and correct me if I'm wrong, but it sounds like work actually gets done in Optimality. Like in other words, it's not just "Let me dip over here to my, you know, chat prompt and get some advice, and then export a document somewhere, and then it's up to me to remember when it's due, or to actually hit publish, or to make these other decisions along the way."

Is that the case? Like actually work is getting done. It's not just a decision optimization platform, it's also like a work/collaboration platform.

Cali Collins: It is absolutely a collaboration platform. So now you're seeing where the real value is. After you become that one person who's trying to start up a company, doesn't really know what to do, and let's put some structure around it, that next big area is these companies that are small businesses and they're growing to the point where you no longer know what exactly what everybody's doing.

And if you can get those feelers into the system where they're collaborating in there, and you say, "This person is working on this area, and these people are doing this," and at the same time, as you're working through it, you're building your context graph. So what is the vision of your company? What is your mission?

What are your principles? That now becomes part of Opti's brain, so when it's making suggestions for this is a, a marketing strategy, these are some different things that you can do, and giving you options, it's based on what you tell it. So it knows what's important to you, what's valuable to you. Are you building a flower shop and you wanna be the biggest flower shop in the world and be monopoly flower shop and take over all the flower shops in the world?

Or is your passion to be the best flower shop for your small town because you think that you have a niche there and you have a passion for that? And it helps to frame having that context behind it. Yes, you could go to Claude and say, "Build me a marketing plan." It's gonna say, "Here's a generic something."

It doesn't understand all the bits and pieces and moving parts of everything that's going on inside of your system, inside of your ecosystem there.

Galen Low: I really like that. And it's the thing you mentioned about as your business grows, visibility sometimes recedes because, you know, we don't always build with these systems.

Traditionally, we're kind of like, we're scrappy at first, and then we're like, "Oh gosh, I have 15, 50 employees now, you know? We need... I don't know what's going on. I don't know some of their middle names." And also are our processes still, you know, are people still doing things right, in alignment with our values, in alignment with our mission?

And I like that sort of idea. I wanted to return to a bit of my devil's advocate take about sort of deciding what we give- To AI, but actually maybe even people too, in order for us to... we think we are curating in a way that helps people make decisions. For example, do they need to know that I had a dentist appointment on last Friday?

You know, maybe I don't need to have full access to my calendar. But also historically you know, I've been in rooms where decisions get made around this notion that we don't have to give everyone the full picture, obfuscating information, leaving things out. And in some ways, A, a lot of decisions kind of do get made in this way where we're like, we haven't surrendered this information.

And then the other side of it being like, okay, well, if we're not giving AI the full context are we not undercutting its abilities? It seems like this sort of balancing point between giving too much information and, yeah, maybe some of those human nuances might confuse it, versus like almost manipulating it to give a certain outcome by not sharing, "Oh, no, don't...

This, this board meeting, don't put the meeting notes in your context graph," because that's gonna reinforce my sort of abilities as a human, or it's going to, you know, keep me in control of AI. I, I don't... I'm like, how do I formulate this as a question? But it's I guess two sides of it, right? One, how do we guard against people sort of abusing AI to lean decisions in a certain way?

And maybe the other thing is like, are there maybe some decisions that need human nuance, that need, you know, more than just a context graph that we should probably, you know, lead in a more human way versus always approaching every decision the same way in terms of okay, here's some context, pull the data, then let's talk it out.

Cali Collins: Yeah. No, I, I think that as far as going into what do you give AI, it doesn't know how to prioritize what's important or not. So if you really do just dump absolutely everything into it, and you don't help say, "This transcript was with this senior leader," and make sure that whatever was said in this one overrides things that were said in others.

Or you need to provide some guardrails or not just give it all of absolutely everything as well because, like I say, it doesn't quite know how to prioritize what's important and what's important to what people, unless you give it that information. I think as far as going with the decisions on for people, yeah, we definitely want the human to make that final ultimate decision.

Can we make better decisions if we're presented with the data that we need? Yes, we need that too. One, when I was at ExxonMobil, I was building a knowledge management center of excellence and an AI native platform for knowledge management and learnings. And one of the big use cases there was you have a product manager who will come in, or project manager on construction, and say, "I've been on five projects.

Every time we've done this, it's failed. We should not do this." Maybe those five projects are literally the only time that instance has ever failed, and every other time it's worked out, and this is the 2%. Now, is it because of that particular person? Maybe. And they're on this project too? Maybe. That's something that AI can help with.

But going back to where we started with that bias, it helps to eliminate some of that. But at the same time, that person's personal awareness is, "This just doesn't seem to work out." They know something that somebody else doesn't. There's some reason why they know that that doesn't work out. And should you just say, "Well, that was the 2%, throw it out, your opinion doesn't matter"?

No. But that fact that that is the 2% gives the room the ability to make a more well-informed decision that you normally wouldn't be able to do.

Galen Low: I really like that because it's not, "Whatever, you're right and you're wrong." There's something to be learned here, and yeah, maybe we can avoid making highly generalized decisions based on, you know, a sample size of one right?

Or a sample size of five people, you know, from one source. But also, there's information there, there's knowledge there that we do need to take into account because, you know, if it happened five times, there's gotta be something there that we should look at. And I guess maybe what I like about this use case is that it can be pre and it can be post, right?

You could be like, "Okay, we need to make this decision. Let's gather all the information we can." Or maybe it's just a we're talking it out as humans, but maybe we should fact check this. Maybe we should apply a bit of critical thinking and see if we can improve the quality of the decision we make, even if it's like, "Yeah, actually, you're right.

Every time we do this type of project, that does happen." It's just that ability to, A, be a bit more data-led, and B, especially when it comes to decision-making, right? Not just listen to the loud, boisterous voices in the room, which often happens 'cause everyone's like, "Okay, well, I don't really wanna mess with that guy, that super senior project manager who's seen it all, and, you know, will fight you if you have any objection."

They will just leave it. But that could also cascade into, you know, years and years of poor decision-making when stakes are incredibly high. You've worked on projects that are very high stakes, very high visibility, you know, big budgets, big personalities, I'm sure, and this is sort of like a way to I guess, just temper that out a little bit.

Because, again, it's that interplay between, you know, humans and technology. We have the ability to not just rely on charismatic orators, right? Who are gonna convince us of a thing, whether it's silly or not, and we can actually apply some critical thinking to it.

Cali Collins: Exactly. And some people are better at rash decision-making.

There are people out there that excel at that. There are people out there that are really bad at that, and they just need time to think on it and ponder through it and map it all out on a whiteboard. We don't all make decisions the same way. We're not all good at making decisions the same way, and this can help normalize that playing field a little bit, too.

Or bring awareness that maybe you're not so good at making rash decisions. Maybe you should step back and think about it for a little longer.

Galen Low: I've approached this from such a technology angle that I didn't even think to ask, but maybe I'll ask now. As you've been building this product, yeah, what have you been learning about human decision-making?

I love that idea that some people are, like, quite good at making rash or snap decisions, and some people are bad, and some people are probably bad and think they're good. W- I mean, maybe let's frame this in maybe on project leaders. Sometimes we are tasked with this responsibility, this rather sort of unsexy responsibility of not always making decisions, but sometimes facilitating them, right?

And we want them to be high-quality decisions, but we also need them to be timely. You know, we're dealing with big personalities, like stakeholders, executive sponsors. Maybe the question is just whether we're leaning on AI or not what are some of the practical habits or, you know, tactics that project leaders can develop to help, I guess, improve the way people involved in their projects make good decisions without sacrificing speed?

Where I'm going with this would you know if somebody who made a snap decision is actually making a really poor decision?

Cali Collins: So a few different spaces on there. I'll step... So I actually have a psychology bachelor out of high school.

Galen Low: I did not know that.

Cali Collins: It's a little bit of why you'll get a little bit of this flavor in there.

I'm also a self-help book addict. Audiobooks now at this point. But I love anything, podcasts, that really just talk about the human psyche and why we think and why we do things the way that we do. Then I went into construction management and got that second bachelor, 'cause I couldn't do anything with the psychology bachelor.

And so that really is what got me into understanding from a construction standpoint just all the decisions that people are making. Why are they making these decisions? Eventually did the MBA as well. So kind of being able to triangulate around all of this different space. Thinking back to what was your original question here, some of the way that we are working through decision-making, how do you know if you're making a- Positive decisions, bad decisions.

It's, you've gotta be able to capture that knowledge management side, that what was the outcome, why did you make that decision as you're making it so that the AI can hopefully go back and reflect on it, 'cause people aren't going to take the time to go back and reflect on it, especially if it didn't work out well.

And why didn't it work out well? Another space that we're really looking into is this kinda AI, the provenance of what's the lineage of data? How does it move through? So when I was at Meta, I built an AI lineage platform to map all of the training data for their AI models across all of Meta. That was my task.

That's what I put on there. And the idea is we built a little bit of that flavor into Optimality. If there is a document you have it create, and what is our marketing plan, as it goes through there, you actually can open it up and say, "This paragraph came from these transcripts. This paragraph came from what this person told me here."

And it knows not just what it should say, but it doesn't wanna be a black box either. It wants to present to you where did these come from. That can be scary for some people, because if it put something in there that you decided and it didn't work out, it now knows that was you- Right ... who made that decision.

But I feel that we as people cannot ... If you're making bad decisions over and over again, you can't change that unless you have awareness of What caused that to happen? Or awareness that you made poor decisions in this specific space, and you can't improve yourself and which is going to improve the company.

We as people need to just... You gotta change for the better, and it takes change, and being able to be vulnerable and understand what are some ways that I can improve, and that's gonna help everybody out. And I think that AI can also surface some of that for people as well, while that might be uncomfortable.

Galen Low: You know, I had asked a question, you know, how can AI help us make better decisions? And when I was thinking about that, I was like, "Data. It's gotta be data." But that whole framing of you know, giving us an opportunity to, A, do a retrospective on decisions we make. Maybe not every single decision we make every day.

I mean, you could if you wanted to, I guess. But also sort of doing that retro, having that traceability, and then using that sort of traceability and visibility to reflect on, you know, getting better and improving. You know, we're not saying AI will eventually make better decisions than humans, and humans won't have to make decisions anymore.

In fact, arguably, the human decision is becoming more and more important. But we do sometimes have that blanket thing, right? We're like, "Oh, yeah, great leaders are the people who just make good decisions." But as we talked about at the beginning, a lot of them are just kind of-- They just feel like they're winging it, and maybe they have a good, you know, sort of success rate.

Cali Collins: Intuition.

Galen Low: Yeah, their intuition. But even though they're feeling like very self-conscious about all of these, you know, decisions that they're making, this is actually an avenue to sort of improve. And then even y- you had mentioned safety earlier as in sort of like, you know, physical safety from harm, and like these notions that we do wanna be able to do a retro on some, some things like that, right?

And we do wanna be able to trace these things. It's not always just, you know, how can I blackmail, you know, Sue from accounting by going into, you know, this like the context graph or whatever and tracing back that it was her. She made the mistake. Let's go cancel her. But actually, we could actually be using it to improve the way we all make decisions, 'cause certainly there aren't any humans that I've met who always make great decisions or even understand why.

Cali Collins: Yes.

Galen Low: I wanted to come back to, you know, this notion of governance and accountability 'cause I think it's a big one. I think right now, as you mentioned, right? It's like it's gonna be the human that's gonna be accountable for a good or bad decision, even if AI pulled the trigger. But I'm curious, like AI does make decisions all the time, right?

Like you mentioned at the top, we are all decision-makers. This isn't just reserved for, you know, a certain echelon of humans, and it certainly isn't reserved for a certain echelon of technology either. So I guess maybe the question is just like what are some of the guardrails that we need to put in place as humans to not just help AI avoid, you know, making a silly mistake, but also reinforcing good behavior or sort of creating a situation where AI is also getting better at making decisions?

And, and would that ever get to the point where we've improved it in such a way that we're like, "Okay, well, now it's good enough that it can make decisions and be accountable for them"? And if that is a possible future, like what does that even look like?

Cali Collins: So there's a lot of communication, a lot of research going into this Kind of space.

I mean, AI is, if we think just large language models, it is a series of decisions. It is deciding what is the next word that I'm going to put based on what is the most likely best next word. All it is is making a series of decisions based on what it was trained to do. And so I think that there's also the side where, yes, you may start out with, "I want it to draft emails for me," and maybe it's not very good to start, but it gets where it's doing a better job of drafting emails or a better job of responding to direct messages in your company's social media.

This is one that has been presented to me at Meta. It says, "We can now send DM responses if somebody messages the company." And I'm looking at some of the r- drafted up responses. I'm like, "Mm, these are not very good." But at a certain point, I can go in and I can edit them and train it, and it's gonna get better.

And it, I probably will for certain low-stakes things. I don't know about that one. But for low stakes, it's probably okay to let it respond to certain things. If you get an email and you're selling T-shirts, and the email is somebody asking, "Do you have this design in this size?" An AI can probably look it up and respond back to that person and let them know.

But there might be another sort of email that you don't want the AI to respond to. So it's really understanding those certain guardrails. There's also, kind of going back to that, where y- there's a lot of research in this area. Is AI going to hit a ceiling of what level of decisions can it make, and what is that going to look like?

Do we need ... I know our personal philosophy is keep the human in the loop and then slowly give over anything that you may feel like that you can give AI, some of those low-stakes types of things responding to those kind of inquiry of emails. But at the same time, keeping it where I'm ultimately responsible because

And not to be scary on the governance side, but I do not wanna pass off my legal liability and my fiduciary duty to Optimality that has access to be able to see what's in my emails to respond out to somebody else and could leak something that could harm the company. And that's not a risk I'm willing to take.

I will keep myself in the loop, and I will let AI do other things to help me optimize my day.

Galen Low: You know, it's so funny 'cause we do have examples from outside of an AI context, or even, I think actually you had shared with me a story about sort of like a medical device that delivers a diagnosis. Again you know, in a certain tolerance range, right?

Not "Hey, you appear to have stage four cancer," "Here you go." But, you know, at a certain level, being able to make that call. And it was interesting what you said about you know, would the software developer be sort of accountable? And I guess in some cases- ... I don't know, maybe talk me through this, because I think the medical device company, like in that scenario, took on the liability for the diagnosis, so to speak.

But in some contexts, that's probably not necessarily the case. It's a possibility, but you know, I don't see, you know, whatever, OpenAI or Anthropic doing that anytime soon. No. So then it comes back to- Mm-hmm ... the human in the loop. But then even you know, like there's other scenarios, right? We were just talking earlier in my community about, you know, training junior project coordinators, right?

And how can we sort of give them feedback and give them safe-to-fail scenarios or situations where they can improve, where the tolerance is high. You know, like answering an email about whatever, yeah, T-shirt size or, you know, a deadline or, you know, clarity on a certain deliverable. Yeah, maybe that's doable.

And even if it's not the best response no planes are crashing, right? It, it's just it will just improve, improve, improve. But there would be some things, right? Please don't, you know, promise a new date without consulting with me. We do this all the time, and there are these tiers of accountability where of course I'd be accountable for my junior project coordinator promising something to a client or a sponsor without checking with people, and that's a coaching moment, but ultimately I have to take, you know, I'm the one who's accountable for it.

And yeah, there will be some repercussions, I guess, and like an action to be taken to improve. But you know, it's not necessarily that different when we're looking at it from the AI lens. Somewhere in there, there's probably a question, but I may have skimmed over that.

Cali Collins: I like that. Nah. I mean, yeah, you think of it as an intern, that you are accountable for the actions of that intern based on what you told it to do.

So yeah, it's definitely the AI can't be accountable, so it doesn't have a sense of accountability. You can't fire it. You can't give it a demotion or whatever- Right. ... you might wanna do because it made a bad decision. But another space that is interesting that you can use AI for is things like contractual compliance.

So, that could be something where you train the AI, "This is my contractual requirements. These are our specs. Now watch what I'm doing and alert me if I potentially could be breaching my contract. Now, you're not doing it for me, but you're watching behind my back to keep me safe and keep me in compliance and keep me in a governance loop", so.

Galen Low: That's a really neat one. Yeah. A, I mean, I love that idea of, you know, providing feedback so that these models are getting better and are becoming more capable. Maybe they won't yet get to the point where they are accountable, right? And like to your point, you can't really fire them or put them on a PIP or anything like that.

But we can, you know, there's coachable moments. And then also leveraging it for what it does best, right? Which is that like I can't hold all our service level agreements in my head all at one time. Wouldn't it be great if something could? Oh, yeah. Like actually that is the case. I wonder if maybe we could just round out by talking about the future.

I kind of alluded to it throughout, but, you know, AI technology is getting better and better at pulling things together and deciding what is important and what is not important. Heck, maybe they are also having that imposter syndrome moment. Every decision they're like, "Oh, is that the right word that will come next?"

And they're like, "Okay, I got it right. Yes. Okay. All right. We'll just keep winging it." But either way, I think, you know, there is a lot of impact on, you know, how decisions are being made, again, in day-to-day work or, you know, the big executive decisions as well, sometimes billions of dollars at stake. I talked about the sort of project manager role as being someone who kind of helps facilitate decision-making and definitely risk management and surfacing, you know, these things.

I guess maybe I'd ask you, you have a project background, you have a product background, and you've been working, you know, in enterprise and tech environments. What happens to the project manager's role if AI becomes better at spotting risks and surfacing dependencies and recommending decisions?

What is the impact to some of the roles that we hold as humans?

Cali Collins: It's even more important. Oh my goodness, yes. The project manager and the project manager, your role is going to be so much more important. So my background being in estimating, doing a lot of project controls. You alluded to it earlier, my job was the one to put together the data and give it to the senior manager to make decisions from it.

That funnel is going to be condensed. So it's not saying we're getting rid of project controls people, but instead you're giving project controls people better roles of not having to just sit there and run around and say, "Did you finish this today? Did you finish this today?" And they can actually spend their time doing things that aren't mind-numbingly boring-

and putting together reports, and they can analyze the data and use their expertise and make things better, and then give that to the product manager. And we have actually noticed in the industry, especially in software, it used to be where you would have, I think, eight to 10 developers for a product manager.

It's actually going down because the developers are getting more efficient at building. You need more product. You need more project management because those are the people that are actually guiding the direction, guiding the decisions. Being able to get data, you're giving so much more data so much faster, so much better information, that they need more of you to be able to make these big decisions from the information.

I know myself, I've just been overloaded in product because my developers are getting faster, and now they're like, "Hey, do you wanna just vibe code up with your ideas and give it to us and then we'll turn it into real code?" I'm like, "Sure." And so I'm learning how to do this, too. But it really does, like you were saying, lowers the barrier of entry to being able to build some of your own stuff, but it just makes- Product and project, anyone making strategic decisions, your role is so much more valuable.

Galen Low: I think that's wonderful.

Honestly, I, I wasn't expecting that optimistic of an answer, but I think you're right. It's a really good point. The velocity of work is increasing, and yeah, sometimes fewer people are getting more things done, but there are certain roles, you know, the product and project and others who are kind of like leading and guiding and facilitating decision-making.

If the work's happening faster, then yeah, maybe we need more of them. I also really like that idea that like what you didn't say is that the data goes directly from AI to a decision-maker now, that you still need to understand it, right? The burden is on you to understand and analyze and assess the data, and then that thing you said earlier, which is like we still make decisions by talking it out.

That is a... Some might say it's a human flaw, but it's just the way we are. We don't really want a machine to sort of tell us how to make our decision. We wanna sort of... There is a human element. There is a persuasive element. There is this sort of double-checking and just, yeah, having the conversation to feel confident that we've arrived at a good decision, and I think that's a really interesting sort of role.

And again, this just, AI just supercharges that.

Cali Collins: It does, absolutely.

Galen Low: Cali, this has been so much fun. Thank you for spending the time with me today. Where can people learn more about you and Optimality?

Cali Collins: So of course, we're on LinkedIn. I'm on LinkedIn. Optimalitypro.com is the website. You can now sign up directly for a free trial for 14 days of Optimality or check out what we're doing.

We keep that live and evergreen, and then as we have more of these exciting vertical use cases building on there, we will have the supporting documentation and all that for those, so you can go check those out too as people are building more and more on top of the Optimality platform, so optimalitypro.com.

Galen Low: Boom. Love that. I will include the links in the show notes to your profile, to Optimality's site. Thank you again for coming on the show. I love your background's so interesting. I didn't even know about that psychology bit. This has been so much fun, really insightful. Decisions, right? They make the world go round.

And, you know, now we've got AI in the mix, but we've also still got the humans in the mix, and I think it's, gosh, what a time to be alive.

Cali Collins: Exactly. It's exciting right now.

Galen Low: Thank you so much, Cali.

Cali Collins: Thank you.

Galen Low: All right, folks. That's it for today's episode of the Digital Project Manager Podcast. If you enjoyed this conversation, make sure to subscribe wherever you're listening. And if you want even more tactical insights, case studies, and playbooks, create a free account with us at thedigitalprojectmanager.com. Until next time, thanks for listening.