Why AI Can't Be a Chief of Staff with Sawyer Middeleer

0:09Sawyer Middeleer: Hello and welcome to the Chief of Staff podcast. I'm Scott Amenta, founder of the Chief of Staff Network, and this is where we sit down with the operators shaping how modern organizations actually run. Every company has bought the AI seeds. However, very few have actually changed how they work. That gap is where my guest lives. Sawyer Middeleer was the first hire in Chief of staff at Omni, an AI startup where he ended up shipping production code with CloudCode straight from customer feedback. Today, he runs Revy Systems, helping chiefs of staff and other operators bring AI into their organizations without drowning in slop.

0:41Sawyer Middeleer: In this episode, we get into building things you'll throw away in 12 months, deciding what AI should and shouldn't touch, and why an AI will never be a chief of staff. Let's get into it. Sawyer, welcome to the podcast. Nice to have you here. Thanks.

0:56Scott Amenta: Good to be here.

0:57Sawyer Middeleer: We had a bit of back and forth rescheduling this, but I'm glad we finally made it work between my sick kids and the storm that rolled through last week in the Northeast. So I'm glad to be having this conversation with you.

1:11Scott Amenta: Same here. Good to be back in New York City. Dry and with good internet as well.

1:17Sawyer Middeleer: For the topic that we'll be talking about today, I think internet is probably key for most of the systems that chiefs of staff find themselves operating in today. So let's jump right into your background. Give us the lay of the land, some context on your past as a chief of staff, also in biz ops, and where you find yourself today.

1:37Scott Amenta: So my background is, I would say, eclectic and pretty typical of a lot of self-described generalists and chiefs of staff. We'll go back all the way. But yeah, my, I guess, early career was in operations, was in strategy and ops for a big e-commerce company, went into biz ops, doing very similar things at, again, startups. So in the like, you know, 100 to and then eventually scaling up into the several hundred range of headcount. And then a few years ago, I joined as the very first employee of this AI software company back in 2023 called Omni, where I was the chief of staff slash kind of do everything person.

2:20Scott Amenta: The company grew over time, raised a pre-seed round and then a seed. And I kind of remained the only non-engineer for a long time until we eventually hired somebody to lead up sales. And so my experience as a chief of staff was very much in the early stage, zero to one growth, building out those original systems, doing a little bit of everything from operations to actually go to market sales and marketing. And then eventually, and I know this is something we'll talk about a little bit more, I'd say in my last six months there, when I was a heavy user of Cloud Code, when that first came out, I was actually shipping features in our product to production.

2:58Scott Amenta: And so my job kind of expanded a little bit once again and did some work that I think you could consider nowadays, like Ford employed engineering with our clients. So a little bit of everything. And then in the last few months, I've gone solo. So founder of Revvy Systems, where I basically do AI operations in a fractional sense or consulting sense for startups, midsize companies of various industries. Yeah, it's a little bit about myself. And yeah, so eclectic, I would say.

3:28Sawyer Middeleer: It's pretty unusual to see a chief of staff pushing code, production code for that matter, to kind of live systems. Was that something that you were advocating for yourself internally? Was it something that you just kind of found yourself doing by the sheer nature of being a startup and you wearing lots of hats? How did that kind of come through? And do you have a technical background or was cloud code like your first foray into coding?

3:51Scott Amenta: I have a somewhat technical background, not in a formal sense. I didn't study computer science or anything like that in school. I was never a developer. But I taught myself to code when I was doing BizOps. I was really interested in that and wanted to go deeper into really data analytics. And I saw a lot of value in building out web apps and automations as well. So building with a lot of the early automation tools, things like N8N and so on. So I was not shipping code initially. That wasn't really my job there. But it was always an interest. I guess. And I guess you could say I was an amateur at it for a long time.

4:30Scott Amenta: And what actually prompted it inside the company as chief of staff shipping code was really more a company need. So like I mentioned, I was doing a lot of embedded work with our customers. So our product was an enterprise sales research and automation tool. And what we ended up doing for a lot of our customers as the product evolved was building out their CRM automations, research automations, basically enabling their account executives and solution engineers and folks like that. And so my job, because I was really the primary customer facing person, became a bit of like a forward deployed engineer in a sense, just over time, it evolved that way.

5:09Scott Amenta: And what ended up happening is that because I was so close with our customers and our users, I was the one getting the most direct feedback about additional features they needed or what was broken, what could be changed. And so once we had tools like Cloud Code, and it became very easy for me to, with a little bit of intuition, like a little bit of sense of what code should look like and how things should work. And then I could just start to say, hey, Cloud, like I was just talking to customer XYZ. They mentioned that they really need this feature. And I think it's probably not too difficult to implement.

5:43Scott Amenta: For example, it wouldn't require any database changes. Maybe it was just something on the front end. Can you implement it? And then yeah, Claude could implement it. I could test it in our sandbox development environment. And it turned into the kind of thing where I was able to actually shift code alongside the engineering team and within a well-scoped way. So I wasn't working on the backend systems or anything like that, but it just became this really cool feedback loop where customer needs something, we can ship it within hours just because I'm the one there to do it and then Claude can or do it there with Claude rather and just have it reviewed by one of the engineers on the team before it goes out.

6:21Sawyer Middeleer: It's quite a cool setup and I think one that a lot of companies aspire to small and large you know anyone on the team that is like especially the ones that are client facing can kind of take that feedback and implement almost near real time obviously with the guardrails that you described maybe even more at larger enterprise. I'm curious So you became a power user for Cloud Code. What was the culture like in the rest of the organization? Were you also helping to lead the rest of the company in adoption? And I know we're going to get to this. This is now basically your full-time job. So was that something that started as a chief of staff before you moved on to Revy?

7:02Scott Amenta: Yeah, it actually really was. So our company was very engineering-led internally. So when we had, we brought on folks who were on the sales and marketing side later, I was the one to help them onboard onto these tools and feel comfortable using them. This was back in 2024, really, when these things were very foreign to people. 2025, eventually, you know, cloud code became more widespread. And a lot of my work actually was embedded within customer organizations. So kind of playing this, like, like I mentioned, sort of forward deployed solutions engineer, rev ops kind of role. And so I was really close with sales and marketing leaders and the more frontline ICs who are the end users of our product and enabling them on the AI tools and helping drive adoption there.

7:45Scott Amenta: And that's really where I saw the need that ended up kind of becoming the kernel of Revy Systems, where that's a lot of what I do now.

7:54Sawyer Middeleer: And who was that stakeholder? Was it chiefs of staff? Was it COOs? Was it CTOs or chief product officers? Who were you working for at that time? And what's the focus from your lens at Revy?

8:05Scott Amenta: Yeah, initially it was VPs of sales often or VPs of marketing. Basically, it was this kind of persona who... And this was more mature organizations. So our customers tended to be larger companies of a few hundred to potentially a few thousand people, where the teams we were working with tended to be specialized on the enterprise side. So they weren't really interested in AI SDRs, which were really common at the time. And basically these tools where you could just say, hey, I want to sell my product to this persona. And it would just go send out hundreds or thousands of emails. These were much more strategic folks.

8:40Scott Amenta: And so the ways that they needed to use AI were much more consultative and they wanted to go deep, not broad. And a lot of the resources at the time really didn't... Resources meaning guides that you could read online or the automations that you could build with the tools just really weren't conducive to that easily. There was basically a real need to say, hey, like this kind of research that we're doing or the kind of prioritization and strategy work that we're doing around choosing which accounts are we going after? How do we position our product or offering for them? How do we map all of the different solutions that we might sell to what we think they actually need and will purchase from us?

9:19Scott Amenta: Like that was really the problem set that these personas were trying to solve. and I could help them bring in AI strategically to assist with that kind of not automate the whole pipeline from like spamming out emails and things, but more like saying, Hey, you have this universe of an addressable market and we can have AI in just a ton of data about that market, as well as context about your solution and your company internally, and then focus what your people do. Right. So at that point, bringing human judgment to sort of filtered information that the AI has given you, the synthesized information, and actually go do something with it.

9:54Scott Amenta: And then over time, I started because I was a chief of staff. When I started Revue Systems, my network was largely chiefs of staff. And I was talking to folks who were in that strategic operator role, and they were bringing to me a lot of the same problems, not just sales and marketing, but in other industries, it was really similar things that I started to hear, which was, look, we know we don't want just automation and AI run rampant. That's not helpful for us. It's creating a ton of slop, right? It's creating a ton of slop. It's creating a lot of headache. It's just not doing good things for us.

10:25Scott Amenta: And often these were chiefs of staff who are saying, well, our vendors had sold us these AI tools. We're using Cloud a little bit, or we're using ChatGPT a little bit. But we know we're not really doing a good job with this. And we need a way to think about it and actually use these tools to actually get value out of them. Because for one reason or another, they were breaking.

10:46Sawyer Middeleer: You've said before to me that teams need to accept that they're building things that won't exist in 12 months. And even though software has come and gone for the last 20 years, that still feels like a pretty foreign concept for teams that are used to setting up durable systems. And even for startups that should be nimble and flexible, wanting to implement things that obviously last more than 12 months. What does that change in mindset mean? And as the chief of staff, have you found them to be the change makers within companies to adopt a more flexible and nimble mindset?

11:22Scott Amenta: The rate at which everything is changing, I think, really does throw people for a loop.

11:26Sawyer Middeleer: I mean, there's a new model every 48 hours.

11:29Scott Amenta: Yeah, there's a new model every 48 hours. Think of it this way. We had, at the beginning of this year, I remember it was January or February that OpenClaw was taking off. At the time, it wasn't OpenClaw, and it took that name. And it created this firestorm around, first of all, it was going to change everything. And then there were all of these security issues because you were giving this AI model access to all of your personal information and it was vulnerable. And now in the last week, OpenAI and Manus and before that, XAI have released their own basically personal agent tools, which are the more polished version of that.

12:11Scott Amenta: And so this thing that became first, it was at first a toy and an experiment, and then it was massively adopted, but was this huge security vulnerability that a lot of chiefs of staff, of course, could not touch and put into their organization for real. So it's now something mainstream. And that's good. And that's the way that these tools are evolving. And so the form factors are changing, not just the models. And I think that's really the important thing is to understand that, like, yeah, the models are getting smarter. So there are things that if you prompt, you know, Claude Cowork or whatever tool today to build for you, it could do that it just would not be capable of 12 months ago.

12:49And so when you're thinking about what tools do we purchase from vendors?

12:55Scott Amenta: What tools do we build ourselves? What do we think we're capable of supporting and maintaining ourselves? How do we want to work with these tools? Like all of these sorts of questions that impact like the operational workflows and planning and budgeting around headcount and all of this stuff, training and learning and development, like the form factors of the models are changing. The capabilities of the models are changing as they get smarter and bigger and all of that. And I think it's really fair to assume that in some ways, the way we work will change every six months to 12 months. And that's, I think, fine.

13:31Scott Amenta: And that's just kind of the world we're in. And I mean that in a narrow sense, not because I think a lot of people are still doing their annual planning, their financial modeling, their board decks, kind of the same way that they did before, but they're just using different tools now. And they're accessing information in different ways. And they're perhaps taking on scopes of work that are larger than what they could have imagined before. And I think that those are the ways really that work is changing.

13:57Sawyer Middeleer: If you were to bucket some of these things, what do you think belongs in the short-term bucket versus what still deserves long-term design? Are companies still doing five-year annual plans, or is that just completely crazy at this point? Is it even possible to plan that far ahead?

14:14Scott Amenta: There are absolutely things that you can plan a long time into the future. You can approach it from a very first principles perspective, which I think is important. I mean, it's always been important for chiefs of staff to think about strategy in a first principles way because we're getting better models. There are things that you will probably be able to do with technology 12 months from now that today you cannot do with technology. I think that's true, right? But that doesn't mean that the way your customers and your stakeholders interact with your service or product are going to be necessarily that much different.

14:51Scott Amenta: There are industries where, yeah, of course, if you're telling me that you're a 10-person software startup, you're maybe in for a bumpy road in terms of what features today are impossible, what features today exist in other products. That's always really been the case for small tech startups. Whereas if what you're doing is financial services or healthcare, education, or even just advisory consulting work, the value you provide to your customers today is probably pretty similar to what it was maybe two years ago with some technology changes around the edges. And what's really changed are the rhythms of business that you kind of run day to day to deliver that.

15:29Scott Amenta: So there are probably pieces of work that used to take somebody a week to do last year that today can be accomplished in an hour. There are probably things that somebody would pay$10 ,000 for a year ago that today they say, well, hey, wait a second, we can just prompt this and chat GPT? Why are we paying$10 ,000 for it? And in some cases, that might be the case. In other cases, you have some differentiators. And so I think thinking about your strategic positioning and kind of doing that bottoms-up analysis is still important just because the things that are durable will become pretty clear and the things that are temporary or not defensible because of technology will also become pretty clear.

16:13Sawyer Middeleer: You've seen kind of these implementations, small and large at different companies at this point through Revy. I'm curious, as you come across different work cultures, how do you handle the fear of failed experiments culturally? Do you start with one particular group? If you're working with the chief of staff, do you go directly through ops or focus on the office of the CEO? Or are you looking for, going back to first principles, what are the biggest problems in the organization that AI can chip away at? And let's start there first. How do you kind of work your way down through that transformation project?

16:48Scott Amenta: I really like starting from the chief of staff in the office of the CEO for this, just because I think that person often has incredible visibility into the actual problems that need to be solved and has view into the scope of them, all of the interconnected pieces. Because bringing AI into an organization and embedding it into how you actually deliver your product or service for real. Like not just prompting a little bit of ChatGPT, but really going deep to say, there are chunks of how we do work today. For example, that might be our advisory work. That might be the way that a consumer of our software interacts with the tool that before were manual or before were static and rules-based kind of old school automation that today we can put an LLM in and make it much more capable or automate a much larger chunk of it.

17:42Scott Amenta: The chief of staff has great visibility into that. Like they're probably not the product leader, of course, they're not the engineering leader, but they have this 360 view and they really understand who the different stakeholders are and who the dynamic or how the dynamics work within the organization that they can give you the roadmap. And in my experience, they've working with chiefs of staff and strategic operators has really accelerated how quickly I can kind of get the lay of the land and understand here's where the opportunity is. Here's where the early adopters are. Here's where there may be laggards and resistance.

18:13Scott Amenta: And I guess when I think about these types of projects, they're really transformation projects. They're not just like, hey, we're going to buy a software and implement it. It's we're going to change the way we operate enabled by technology. And I think that's a very important distinction to make when you're purchasing and using AI tools for work, which are just a different class of product than, hey, we're just going to replace Salesforce with HubSpot. That's not the same kind of change.

18:38Sawyer Middeleer: I mean, I think at this stage, most companies have basically bought the seats, right? There's some new tech layer, whether it's just the LLMs or other ragged systems based around them, or an adoption of the AI++ from the tools that they already had, whether that was Notion or Asana or these other products. but what what does the gap between like we have clawed and we work differently actually look like inside of a company right like those are two very different things and i think the the baseline which i think a lot of organizations were going off of which was just purely adoption right like are people actually using the tools is no longer that interesting right because we know that people are using the tools what we don't know is are they actually using the tools to like fundamentally change the way that they work or the way that their team operates.

19:31Sawyer Middeleer: And that's very different, as you said, than just kind of a regular prompt on a day-to-day basis versus let's redesign this process from the ground up.

19:39Scott Amenta: There are a lot of different ways, I think, that you could approach this. But the way that I think about it is, first of all, start from assessing the current state of your business. Somebody inevitably is using an AI tool, whether it's built into one of their vendors, Slack, I saw I got a notification today, just released skills. So somebody's using Slack skills to do their work, or somebody's using Notion skills, or somebody's whatever. These are all in different places. And it's up to the user to choose which ones they want to adopt. But somebody's using something somewhere. And I think in order to actually...

20:15Scott Amenta: You're saying go from... Okay, yeah, we've bought the seats. We have the tools. But how do we show that we're actually getting ROI from them and how do we actually change the way that we work. It starts from really understanding the current state. Like what are the processes now that we do? How do our people take the inputs of our business, whether that's users or data or whatever it might be, and transform those into solutions such as customer advisory or software or widgets that you put on a shelf? like there are of course a ton of these different processes that you can map out and some of them are going to be very easily automatable because they're operational they're low value there's a really clear kind of pathway to turning a customer support ticket into a resolved customer support you know request and a happy customer right like we know when somebody sends us this kind of message we really we have an sop we have like a macro somewhere there's a database that says, hey, this request comes in, we know how to resolve it, and then that will lead to a good result.

21:25Scott Amenta: Well, if we can say for each of these tickets that come in and then get resolved, we can put some ROI behind that. Either some likelihood that person will come back. There's some purpose to doing this. Well, then suddenly you have the basis of an ROI calculation. You have value associated with a process. And then you can map out what is our... How do we currently do this? So we have a person, let's say, or likely what you have is you're purchasing an AI product already from Gorgeous or Zendesk or Intercom or one of these tools, right? And it's resolving the ticket. Or maybe some human is taking the thing and putting it in chat GPT and then using that to resolve the ticket.

22:09Scott Amenta: How long does this take us now? And if we were to automate it, whether in terms of time savings or capacity gain or quality improvement, how can we actually move the needle? And how would we have to work differently in order for that to happen? And then you can sort of do this across the entire business in a way. So you might want to identify the process that you know you can either speed up, make better, or get some capacity gain by saying, hey, somebody is currently doing this. We can actually move them to something more valuable, more strategic. And what would have to happen for that to be the case?

22:47Scott Amenta: And I think this is where a lot of people will get tripped up because they'll say, hey, we have a person who's currently handling these support tickets. We can just replace them with a bot. And then suddenly, that'll be cheaper because the bot costs five cents to do this and the person costs five dollars to do this and the person is just you know kind of taking in the request and then responding to the customer and closing the ticket how hard could it be and it turns out that this is actually not always that easy because the human has a lot of context in their head they have a lot of judgment in their head about what a good response looks like whereas people often don't necessarily see the complexity in turning the inputs into the outputs.

23:29Scott Amenta: And so I think there's, you have to be really intentional about understanding where's their opportunity to do this. And then what has to be true? Like what are each of the different points along this process? Where is the person pulling the information from? What memories do they have or what intuition do they have that they're bringing in to this task in order to do it well? And can we inject that into the AI process in the same way. And I think by doing that, and if you can then sort of tell that story and map out that process really accurately and completely, then you'll have a really good chance to transform that process using AI.

24:10Scott Amenta: And because you have the foundational sort of metrics or the KPI that is underneath that, you'll sort of get the clear ROI by doing that.

24:19Sawyer Middeleer: Do you think it's that kind of understanding of what the ROI might be or what the KPIs that we're, you know, benchmarking the differences between the fully human version of this process versus the, let's call it like AI empowered version of this process? Is that the separation between a company that is living in that gap of we've adopted the tools, or sorry, we've bought the tools and we've adopted the tools versus we've actually changed, again, the way we work? Is the ROI the first thing that people should be looking at?

24:57Scott Amenta: Yeah. I mean, look, there's a huge gap between what AI can do and what AI can do well and what you should be using AI for within an organization.

25:04Sawyer Middeleer: And that's two-gap identified, yeah.

25:08Scott Amenta: And this is really important because in talking about this mapping out the process earlier, I think something that I've learned to do through a lot of trial and error, but I think you can bring some intentionality to this and some rigor to doing this is, okay, you choose a process, right? So it's resolving the customer support ticket, or it's constructing a board deck, or it's truing up our books against budget and projections, whatever it is, There are all these different processes that you can identify. And some of them should be automated because what the human does is low leverage, you could say.

25:44Scott Amenta: So somebody spends an hour to do this. How much value does it create? Can you tie that person's time to ROI? And you can't always. There are certain things that, of course, you can because it's easily metriced against. So like customer support, one hour in brings you X number of tickets. and you can sort of tie that to other things. But then, you know, doing a strategic plan, what ROI is there from spending an extra hour on a strategic plan? I don't know. I don't think anybody can measure that. So ROI directly is only one piece of it. The other piece of it is understanding, you know, when you have that understanding of what an AI can do well versus what a human can do well, should the person actually do this themselves?

26:29Scott Amenta: And you can easily be the judge of quality there. You can either go by the qualitative results or the quantitative results. I think this is where you can bring in a lot of experimentation to say, hey, we have some humans doing this process. Let's say it's triaging bugs in our software product versus having an AI do it. And you can measure them against each other and see qualitatively and quantitatively as well. Where's the performance coming from? But yeah, I think ROI is part of it, but also just kind of an intuition of where relative skills lie and where people's time is best spent. I think a huge failure mode for bringing in AI in this way.

27:10Scott Amenta: And I've seen teams go much too far with this, where they say, hey, our cloud code can do all of these different things. So we're just going to apply it to everything. It does it faster than a human, but it turns out that it does it worse. and when you do that suddenly you erode whatever you know value you have i mean i think we've all seen our favorite brands or favorite products kind of devolve into slop right maybe there are more bugs than there used to be maybe the marketing is less compelling than it used to be maybe they're out of touch with their customers in a way and i think that's kind of a failure case that's really easy to fall into when you simply apply a brute force roi number to everything in the sense of time saved or hey volume right because time saved and there's things that you should spend a lot of time on and there are things that should be quality over quantity this is why i keep coming back to the map out the process and map out the value chain piece of things which is where i think a chief of staff is really well equipped to understand that whole thing end to end and doing that really forces you to think about what does good actually look like What does success actually look like?

28:19Scott Amenta: What's the difference between what an AI is producing versus what a human is producing? And where those things look virtually the same, that's probably where an AI is going to be well spent. And where the human just has that special something, you need to keep the person in there.

28:33Sawyer Middeleer: This has quickly become an excellent lesson in change management, even though we haven't used those words yet.

28:39Scott Amenta: So the funny thing is, I don't come from an enterprise change management background, but But I've quickly, quickly had to learn because that's really what this is, right? We're talking about a different way of working. We're talking about not just bringing in a different sort of technology that you have to learn and have to interact with differently that's capable of different things, but it's also a different way of working with each other and a different way of not just like the communication and the coordination, but the org charts are different. Like we have, for example, we have so many roles now inside of small startups where the change management is easy that are these like hybrid sales and engineer or marketer and engineer because AI kind of takes somebody and gives them these technical superpowers in a way that they didn't have before.

29:30Scott Amenta: And this kind of role, I think, naturally is diffusing a bit slower into bigger companies where they have more structure and more risk associated with trying things out that might not work so well. But in a small company sense, everything is changing about how they're structured and how they operate. And yeah, naturally, that becomes a change management problem.

29:55Sawyer Middeleer: That's a really interesting insight. And not one that I've heard put so bluntly, but you're actually a living, walking example of that yourself in some ways. Chief of staff pushing code now becomes the connoisseur of other roles blending into those engineering positions that are equally as dangerous from a product evolution standpoint, right?

30:18Scott Amenta: I wrote something a couple months ago. I was surprised how much it ruffled feathers, but I made the case basically that every role is becoming engineering. And I think I said it a little bluntly, but the thesis that at the time was when you have LLMs and agents that are responsible for doing more parts of more jobs, and these jobs are becoming more and more technical. So if you're a salesperson today, like let's say you're the very frontline kind of salesperson. Yeah, you might just be writing emails one-to-one, picking up the phone, dialing all the time. But chances are you're not doing that.

30:57Scott Amenta: Chances are you're prompting cloud code. You're building automations into your work. You have what is quickly becoming little pieces of internal software and tools and automations that help you become more effective in your work. Well, and then what happens? then you have more and more of these automations. They're going to break. They have to integrate with each other. You need to quality test them. You need to maintain them. And these are certainly costs and trade-offs of doing it this way. But when you do figure out how to systematize these things, and when the salesperson, for example, does become a better engineer, that's when you get this kind of 10x productivity boost in that individual, which I think is super interesting.

31:39Scott Amenta: And so when I say that every job is becoming engineering, which I think is true, it's marketing, it's sales, it's customer support, it's product management, it's operations, certainly. You have to start to think like an engineer, not just an operator, not just a go-to-market person or whatever it is. You can't lose that. I don't think you can be a chief of staff who only suddenly thinks like an engineer. You can't put an engineer in a chief of staff seat and expect them to be successful. that that's you know a recipe for disaster but if you are a chief of staff you should think more like an engineer or or get comfortable thinking more like an engineer in a system sense in an architecture sense in a like judgment sense about you know how should software systems work not that you have to know how to code one of them or not that you're going to suddenly become an engineering manager or software architect.

32:33Scott Amenta: But it's just natural when the tools that you're using to do your work and so much of the day-to-day operations is owned by software and not people.

32:42Sawyer Middeleer: And so much of that software is now no longer being bought off the shelf, right? It's being built internally. As you said, the kind of small working tools that might be suitable for an individual or a small team or an entire chunk of an organization. I've got two questions for what you just what you just described the first is a is an easy one who was most ruffled the engineers or the the other side of the equation do you think it was the other side of the equation actually

33:10Scott Amenta: not wanting to become more technical i think it was also just a resistance to a perceived kind of lack of humanity in what happens when a interpersonal job becomes more more engineering focus, like I would saying. But I think this is where, and again, I'll go back to what I said earlier. There are things that AIs should do and there are things that people should do. And there's a real art and science to this and a real balance where you can't push. You can't, for example, there are so many companies, so many startups, not to name many, that are shilling these AI chief of staff products. An AI cannot be a chief of staff.

33:50Scott Amenta: You cannot actually build a software version of a chief of staff and expect it to be useful. right? At the very best, it's going to help you, you know, triage your inbox and track your meetings and whatever. And I think there's a lot of value in that for sure, right? But there's so many things that a good chief of staff does. There's so many things that a good salesperson does in the go-to-market context that an AI absolutely cannot do. And so I think it's more about like, hey, how do we blend these things together? Take the best of both worlds, not replace one with the other.

34:19Sawyer Middeleer: And the person often sitting in that seat that does have, say, more interactivity with customers, partners, or human beings versus code has the insight, has the context to understand more clearly where that line should be drawn. But I think we can both acknowledge that that line is also a moving target. So as AI gets better, as the systems evolve, as the models change, as the tools around them change as well, that person needs to understand like, okay, well, six months ago, this was a terrible idea to throw at AI. But today, actually, we can systematize or automate more of this process than we had before.

35:00Sawyer Middeleer: And that only comes by trial and error, by testing, and going back to where we started, the willingness to throw things away, not just code, but throw former ideas of what the systems couldn't do away as well. I think that's the hardest part is like the chain that change is happening so quickly that it's not just uh you know engineers being willing to throw code away it's it's people being willing to throw their

35:26Scott Amenta: ideas of what the technology should be allowed to do yeah no totally like there are things that i just remember a year ago wanting to do with ai or people asking to do with ai things that involved Like, you know, hey, I want to take every stake, you know, for example, every investor of ours, I want to map out all of my investors networks, I want to filter that down to the individuals who I should ask for introductions to, so that, you know, I can maybe raise more funding or get more customers or whatever this kind of like network based strategy of, you know, whatever it was. and these things were complex and like you could absolutely build it 100 but you really had to understand how to engineer it well now you can literally prompt you could probably just take the transcript of what i said shove it into astra and it will build the whole thing for you so yeah the the bar is definitely moving and like there are things also that we're discovering you should just buy and not build it yourself because like there's a real cost to deciding, Hey, I have this like superpower of being able to code any software or, you know, automate anything to say like, well, okay, but there, there's a really good value in having other people maintain it for you.

36:49Scott Amenta: And like a good example of this that I found is like anything that requires integrations, if you're not an engineer and you haven't had to build integrations into other software before. So such as like a custom connection from maybe Notion into some bespoke tool that you use, that kind of sucks, right? And every single time that they change their system, you might have to update your thing too. And a lot of people are realizing like, there's a good reason why I pay Notion to do this integration for me and not build it myself.

37:22Sawyer Middeleer: When you work with chiefs of staff, either they come to you or you've built a relationship through an existing client. I'm assuming very often that's happening when the CEO has handed them an AI transformation mandate, saying, do more AI. Why aren't we doing more AI? And the chief of staff says, well, what does that mean? But what do you think is usually missing from that mandate?

37:47Scott Amenta: Where is that gap? It's funny, actually, a lot of it is, yes, the CEO, the founders say, hey, we think we should automate this. It's like, we need to build out this function. Often where I come in again is sort of on all the customer facing stuff. So it's customer support or sales or marketing. And they tell the chief of staff, hey, we should be growing more, but I don't want to hire any more headcount. That cliche of like, do more with less or scale without adding headcount is real. And it makes sense because like, I would say the most tangible value from using AI well, and or at least this is with market.

38:26Scott Amenta: And I think it's very real is revenue per headcount. If you're a startup and you're trying to raise money, or if you're a startup and you're trying to become profitable in scale, right? So a startup is broadly, not just like a five-person company, but this is true of several hundred person kind of mature startups as well. Being able to show that we've taken our team from, let's say, 250K revenue per full-time headcount and grown that to 500K is massive. That's really significant and that's a clear outcome of using automation really well. But very often in the chief of staff, what that looks like is you get a request that lands on your plate.

39:06Scott Amenta: hey, we have higher targets this year for our marketing demand gen. We don't want to hire any more marketers. So can we use AI to do this? And often it's like, wait, what? Because again, like I was saying, you can't just plug in AI and expect it to do a good job. You really have to understand the process. You have to understand where's the value coming from? What's unique about how we do marketing in this case that really brings customers to us? How do we use AI to drive that and enable that versus, you know, just kind of create another spam cannon. And I would say, honestly, chiefs of staff have the toolkit for this.

39:44Scott Amenta: It's not anything different from any kind of transformation or change, you know, process that they've done before. It's, you have to understand, you go talk to your stakeholders, understand what they, you know, where they get value from, understand what actually matters, what understand what doesn't matter. How do different stakeholders come into play here. So for example, bringing in AI systems, yes, like the goal is to automate marketing, let's say. But if you have engineers, your engineers are going to be involved. If you have data analytics folks, those folks are going to be involved. So is IT.

40:18Scott Amenta: And so it becomes a cross-functional project. And again, yeah, chiefs of staff are great at doing this kind of thing. And what's really important is just to bring in the mental model of like, where is AI going to be most useful for us? And how can I kind of guide this conversation so that it becomes a little bit more strategic than can we use AI to automate marketing? How do we do that? Because the answer is yes. It's just very hard to do it well. It's very easy to automate, very hard to automate well.

40:45Sawyer Middeleer: The other factor there obviously is like this sense of urgency combined with this sense of fear, both of like missing out on opportunities or being overcome by competitors who may be moving faster because of X, Y, and Z tool, or are just publicizing that they're moving faster because of X, Y, and Z tool, even though it might be a totally different story behind the scenes. There's a lot of this built up agita of, are we doing things the right way? Are we doing them the fast way? Is there a combination of those two things that we need to be more focused on?

41:21Scott Amenta: This is such a good point. My natural inclination is to be the systems person, thinking and systems thinking and you know what do we automate how do the processes fit together like all of this but there's such a core cultural component here like we have making sure that everyone is kind of sharing a belief in the future of the organization that you're building or at least that you know there is kind of a co-curring north star that everyone's moving towards and that there's a clear story of why that's the north star and what people stand to gain from it because nobody wants to be in a position of feeling like their job is six months away from being automated out of the picture.

42:02Scott Amenta: And in reality, the data backs up the opposite narrative in the first place, where companies that use AI and use it well often actually need more headcount. There's obviously the nuance and the where the headcount is needed and all of that. But AI is largely complementary with people in kind of a big sense, at least right now. And I think that's a good thing. And that's not that surprising because you realize that you can do a lot more when you use these systems. Well, there's so many functions within a business, so many kind of projects and ways to delight customers that just get tabled today because we don't have capacity.

42:45Scott Amenta: We don't have time. What AI does when you use it well is it adds the capacity. And it enables people to level up what they're doing as long as they have the willingness and training to do so. And I think there's a huge cultural component in bringing that kind of change about versus being the story on the other side, which is, hey, we just cut a bunch of headcount because we didn't bother. Or we adopted these tools and they're actually not bringing in ROI. And so we have to pivot. That's obviously not a good story, but it's entirely avoidable.

43:17Sawyer Middeleer: My theory is that organizations that have always fostered more grassroots ideation from the broadest parts of the organization and have actually used those ideas to bring them to market, those organizations probably have the best chance of success. Because as those employees free time, they're given the opportunity to turn those ideas now actually into real life things. And as you said, like if that capacity was never there in the first place, the idea exists, but the ability to go build it doesn't. And so, but, but the organizations that are purely top down, you know, they, they will suffer more, I think, because like they don't have that culture, that spirit, that entrepreneurship of people, you know, just wanting to see kind of their customers succeed.

44:05Sawyer Middeleer: And as you said, finding creative ways to delight them.

44:08Scott Amenta: I've seen this over LinkedIn in many different instances where people, I see, you know, people who I know their job changes and their job changes are going from like, you know, account manager to AI solutions architect, or they're basically moving like the job title changes are surprising. And I think in hindsight though, it's really not that surprising because to what you said, like when you have this spirit of entrepreneurship and you've like organizations that encourage people to learn and build and solve their own problems using these tools without putting in so many like, you know, just friendly, unnecessary guardrails and constraints.

44:48Scott Amenta: You wind up empowering people, I think, to like, think of their work differently, self-identify differently as builders and not just, you know, what their job description was before. And imagine, hey, like I can be an account manager that builds solutions for other account managers within my team because this is the kind of thing I love to do. Or, hey, I used to be able to have a book of 10 accounts, let's say, and I could do those well. But then I built automations for myself and I built solutions for myself. Now I can take on 15 accounts and do a better job. That's incredible, right? And that should be a reason to put the foot on the gas and try to grow bigger and not go the other side and say, hey, look, this person's 50 % more effective so we can cut out 50 % of the team, there's probably opportunities to put folks in the positions to be successful and really put gas under the fire and try to grow faster, which is where I think a lot of the opportunity lies right now.

45:45Sawyer Middeleer: Where do you think the chief of staff role goes? What does that job become? Is it AI operator? Is that even a title yet?

45:52Scott Amenta: I'm trying to make it a title. No, I think chiefs of staff are going to stay chiefs of staff. And I think a lot of the titles themselves might not change so much. But like the job descriptions themselves will like, I think a chief of staff is today already expected to know quite a bit about how automation systems work, both at the 10 ,000 foot view and kind of on the ground floor, you know, nitty gritty of how to implement them and how to how to use them. But that's, again, not too different from how chiefs of staff have always been needing to understand the 10 ,000 foot view and the ground floor view and kind of be comfortable in both of those spaces.

46:27Scott Amenta: is just that the tool set's a little different. And I also feel like engineering, product, data science, data engineering are more kind of core stakeholders now than perhaps they used to be outside of non-software organizations. And so that might be a change for folks as well, being suddenly in rooms with different kinds of conversations going on that are more technical in nature. But I don't know, chiefs of staff are really good at adapting. And so I don't think you have to change the makeup of someone. so much.

46:57Sawyer Middeleer: That's one thing we've got going for us. Sawyer, you're joining us at Chief of Staff Connect New York this October. Tell us a little bit about what your session is going to be about. I'm excited for it.

47:08Scott Amenta: So I'm going to be talking about building AI flywheels and basically making AI work for your organization. And so my focus is really going to be around context systems and kind of building out this mental, helping chiefs of staff understand how these tools work and think and get a little bit more tactical about how to actually implement them in their team and understand what they're doing and answer some of these questions for themselves that I think a lot of people are wondering, like, hey, we bought the seats. How do we make this valuable? Where do we identify the processes that deserve automation and the ones that don't?

47:46Scott Amenta: Or, hey, how do we actually kind of level up from, yeah, we're using the tools. It's maybe making us 15 % more effective. of like, how do we take that to the next step? That's really what I'm going to be focusing on.

47:58Sawyer Middeleer: Awesome. We've identified a few different gaps that are worth closing on this call. So I think your session will definitely be one for the books to help people advance in this journey. Hope so. Looking forward to it. Sawyer, thanks so much for joining us today.

48:12Scott Amenta: And I'm looking forward to seeing you in New York. Likewise. Thank you.

48:16Sawyer Middeleer: Thanks again to Sawyer for the conversation today. Two things I'm taking away. First, map the process before you automate it, because the context and judgment in a person's head is exactly what breaks when you hand it to AI. Second, it's easy to automate, hard to automate well. Faster isn't necessarily the goal if the work gets worse. You can catch Sawyer live at Chief of Staff Connect, New York, this October 22nd and 23rd, where he's going deep on building AI flywheels inside your org. If you want more conversations like this one, join the Chief of Staff Network, where a few thousand verified Chiefs of Staff across 25 cities with in-person events, programs like Launchpad and the Chief of AI Fellowship, and a membership built for operators doing this job every day.

48:57Sawyer Middeleer: Find us at chiefofstaff.network. If you enjoyed this episode, share it with another operator you know if you think they'll get value out of it. Thanks for listening to the Chief of Staff podcast. We'll see you next week.

49:21Bye.