# How AI Is Transforming Governance and Workflow Automation in the Enterprise - with Tsavo Knott of Pieces

- **Show:** The AI in Business Podcast
- **Published:** 9 Jul 2026
- **Source:** https://podcast.emerj.com/how-ai-is-transforming-governance-and-workflow-automation-in-the-enterprise-with-tsavo-knott-of-pieces
- **Transcript:** https://podcasts.opennash.com/shows/ai-in-business/how-ai-is-transforming-governance-and-workflow-automation-in-the-enterprise-with-00696b/
- **Provider:** OpenNash Podcasts (machine-generated)

## In brief

Tsavo Knott of Pieces argues that AI is turning every employee into a generalist who manages a set of agents. That speeds work up but scatters context across tools, notes apps and personal systems. His proposed fix is one shared 'context substrate' that records work as it happens and serves it to both people and AI agents. He says the companies that record everything will become AI native fastest.

The tension is speed against memory. People move across more kinds of work than before, yet the reasons behind decisions live in disconnected places, and employees must now learn to pull from shared context themselves. Knott and host Daniel Faggella tie the fix to 'context routers', the overloaded people who already field questions from everyone. Concrete examples include a tax-credit review done in hours and a sales call needing an answer fast.

## Key insights

1. **Everyone is becoming a manager of agents** Knott says AI lets individual contributors cover several domains by delegating to sub-agents, so they act like project managers over agents. Context switching is a side effect of doing more across more areas.
2. **Fragmented context is the core operational bottleneck** Teams use different note takers, connectors, markdown files and homemade dashboards, which Knott calls high entropy with little standardization. People can't tell where information lives or how to share it.
3. **Capture context at the bottom, grant access above** Knott says frontline staff such as QA and support must record their work for managers to query it later. Recording at that level also sets natural boundaries for who can see what.
4. **Always-on records turned weeks of work into hours** Pieces used its own captured history to sort a year of one person's work into research and development projects. It repeated that for 22 more colleagues, cutting a one-to-two-week task to a few hours, even though nobody had taken notes.
5. **Finding a document is not knowing how it was made** Knott contrasts enterprise search tools like Glean, which find the finished file, with capturing the decisions and conversations behind it. He says the 'work in progress' history is what lets teams avoid repeating mistakes and hand context to agents.

## Questions this episode answers

### What is a 'context substrate' and why do companies supposedly need one?

Knott describes it as one shared layer that captures and stores all the work context in a company, so any tool such as Cursor, Claude Code or Codex can draw on it. He argues this is needed because today's context is scattered across many tools and personal systems.

### What does Knott mean by a 'context router'?

A context router is someone, often a product manager, executive or director of technology, who spends their day gathering and passing along information between teams. Knott says they feel the pain most and are the best people to push for a shared system.

### How did Pieces use captured context for its research and development tax credit?

Knott used AI sub-agents in Cursor to review his last twelve months, listing projects, time spent and what reached production. He then repeated it for 22 more employees, turning a one-to-two-week job into a few hours.

### How is this different from enterprise search tools like Glean?

Knott says Glean-style search finds the finished file but not how it was made. His approach also keeps the decisions, conversations and iterations behind the file, so people and agents can reuse that reasoning.

### What should a leader do first to get started?

Knott advises finding where people still copy and paste context by hand, for example from a note taker into an AI tool. He also says to ask the people routing the most context what would make their job easier, since that shows where capture is missing.

## Facts

- **Tsavo Knott** [0:03:37]: "everyone is becoming a PM in a sense, right? A PM over a series of agents, a series of specialized or generalized agents to accomplish tasks."
- **Tsavo Knott** [0:05:59]: "One of the general partners over at Y Combinator, he recently said, like, the the core to becoming AI native period is to record everything"
- **Narrator** [0:00:12]: "Pieces is an AI memory platform that captures and structures context across the tools organizations already use, giving teams shared visibility into institutional knowledge, decisions, and context, so nothing important gets lost."
- **Tsavo Knott** [0:09:28]: "They put up Jira tickets, they're tracking issues in GitHub. They're like, hey, you know, these are the things we need to accomplish."
- **Tsavo Knott** [0:14:05]: "you need one universal context substrate across everything that you're doing that, you know, independent of what you're using, be it cursor, codex, cloud, independent of of any system you're using, they can tap into it."
- **Tsavo Knott** [0:14:33]: "So it's almost like an everything CRM, if you will."
- **Tsavo Knott** [0:16:41]: "chief operating officer, CEO, director of technology, chief innovation officer. They're kind of the in in innate routers of context in the company"
- **Tsavo Knott** [0:17:45]: "how many people have we seen this quarter? Come in on this version of Windows having this type of issue with native audio, and I need to be able to access that information very quickly."
- **Tsavo Knott** [0:23:55]: "I just in cursor, uh, used a variety of sub agents to say, Hey, look at my last twelve months. What are all the projects that I worked on? How much time did I spend on them?"
- **Tsavo Knott** [0:24:26]: "Now I need you to do this twenty two more times across every individual in the company to get a comprehensive snapshot of what in the last twelve months was research and development."
- **Tsavo Knott** [0:24:26]: "And I just turned a process that would have maybe taken a week or two across everyone's time into a couple of hours, right?"
- **Tsavo Knott** [0:24:53]: "no one in the company consciously took notes about what they did and if it went to production and how much time they took on it over the last year"
- **Tsavo Knott** [0:25:19]: "They need to be able to say, when was the last time I worked on this project? What were all the architectural decisions we made, what were the comms we had"
- **Tsavo Knott** [0:27:02]: "And they have a very specific metric, which is if this customer is on the phone for over twenty minutes, the probability that they become a customer drops below 40%, right?"
- **Tsavo Knott** [0:30:38]: "So glean, you know, you might have heard of Glean here and there, but it's kind of this, you know, enterprise search. It'll find you the artifact."
- **Tsavo Knott** [0:34:43]: "They say they literally pull up the notes, they copy it, they paste it, and then they say, I need to do XYZ, create some action items from these"
- **Tsavo Knott** [0:36:09]: "I think the companies that will move to being AI native the fastest are the companies that record everything, right? They're the companies that have no details lost, no details missed."

## Transcript

**[0:00:12] Narrator:** Welcome everyone to the Emerj AI in Business podcast. Today's guest is Tsavo Knott, CEO and co-founder at Pieces. Pieces is an AI memory platform that captures and structures context across the tools organizations already use, giving teams shared visibility into institutional knowledge, decisions, and context, so nothing important gets lost. Tsavo joins Daniel Faggella, Emerj CEO and head of usage, on today's episode to examine how AI is reshaping individual contributor roles and the operational realities that come with teams working across broader surfaces of work.

**[0:00:45] Narrator:** He explains how fragmented context scattered across tools and personal systems is slowing execution, and why enterprises need a unified substrate for capturing and retrieving the day-to-day details that drive decisions. Today's episode is sponsored by Pieces. According to Edison research, 79% of Americans listen to online audio monthly, an estimated 228 million people. Podcasts offer a rare opportunity to capture twenty plus minutes of attention with VP plus leaders in America's largest enterprises. Emerj reaches one million listeners every year.

**[0:01:18] Narrator:** Learn more about how these conversations drive pipeline for AI solution providers. Download our media kit at Emerj dot com slash ad one. That's e n e r j dot com slash ad and the number one. Now the conversation with Tsavo.

**[0:01:50] Daniel Faggella:** So Tsavo, welcome back to the show.

**[0:01:52] Tsavo Knott:** Dan, thanks for having me. How you doing?

**[0:01:54] Daniel Faggella:** Doing well, man. I'm glad to be able to to dive in. AI has moved exceedingly fast in these last eighteen months or so, since we've had a more long form conversation. I I kinda wanna open up with a bigger picture dynamic around sort of teams, AI productivity, communication, and let let you take this where you'd like to take it. But I'm gonna frame it this way, like when you look at teams that are using AI now to sort of do knowledge work that involves a lot of context.

**[0:02:20] Daniel Faggella:** Whether it's in marketing, whether it's in supply chain, whether it's in development and engineering, what do you see as the most common current hurdles with kind of the LLM AI tools we're using today?

**[0:02:31] Tsavo Knott:** Yeah, you know, Dan, we've touched on this before, maybe in a prior episode. But I think one of the the biggest hurdles that we're seeing out there is this transition of the old individual contributor into the new individual contributor. Where they're not particularly working on a specific thing, um, and they're a specialized, you know, contributor to the overall effort. But they're becoming more of a Swiss Army knife, and I'm sure that you've talked about this concept of, you know, the Swiss Army knife individual within a company starting to wear more hats, solve different problems across a variety of surface areas.

**[0:03:03] Tsavo Knott:** And I think the biggest thing is that with AI and with agents, people have the capability now to not only solve a particular task in their domain that they're perhaps comfortable with faster and with a higher level of, you know, efficiency. But they're also able to switch over and cross over to other different domains, right? So I think one of the biggest things is that people are moving away from, and they'll have to move away from this concept of me myself and my work to me myself and the delegation of work to a variety of sub agents or agents to accomplish a variety of tasks.

**[0:03:37] Tsavo Knott:** And so I think the probably the largest hurdle that's quietly going on in companies right now is everyone is becoming a PM in a sense, right? A PM over a series of agents, a series of specialized or generalized agents to accomplish tasks. And so I think, you know, that's the biggest thing. Is context switching is a side effect of this, you know, broader capability that individuals now have, uh, in this world of of AI. So I think that's the biggest thing.

**[0:04:05] Tsavo Knott:** Is just people are doing more across more surface areas.

**[0:04:08] Daniel Faggella:** Yeah, there's there's a few. L Let me put so you're you're touching on. I actually haven't heard it framed this way, but I'm gonna I'm gonna put the col the related and collected ideas on the table, see what your response is. You're talking about old individual contributor model to new individual contributor model. I would agree there's a sea change happening and many companies are already there, but they haven't labeled it, they're just noticing work shift. A few ways that I've almost been like mentally conceiving of this and that that have come up in other conversations with a data robot, a Google cloud, uh, other leaders we've had on.

**[0:04:36] Daniel Faggella:** One is sort of like this notion that departmentally, so company-wide, maybe there's a big meta-brain, but it's definitely made of sub-brain components, not unlike the brains you and I have, where, you know, sales department, whatever. Everybody working in sales is working to close their deals, but they're also working to add context into a bigger system. So it could be so now when the sales manager's like, okay, well, where do we stand after the last call with the, uh, the Walmart account?

**[0:05:00] Daniel Faggella:** They can just ask the system and like eighty five, maybe ninety five of the time get all the detail they need, including the exact parts of the conversation of what came up, just kind of snapped in front of them. Um, so there's like a meta brain. In addition, there's a meta brain, not just about details, but about best practices. So the way that we handle this specific objection has altered over the last two, three months. And, you know, our Gong data shows this.

**[0:05:24] Daniel Faggella:** But also, if we wanna pull transcripts, we can just ask the meta system and we can kind of see how it's going. So the best practices, the overall context, it's going into a shared brain. So part of being an individual contributor is how well are you feeding to this system? And on some level, how well are you taking chunks like fistfuls of the current work you do yourself and giving it to that system? So you're doing more of the stuff that the meta brain can't already handle better than you.

**[0:05:50] Daniel Faggella:** It this this is almost the dynamic that I'm putting a bit of a finger on. But you're super close to this and intentionally building a product around it. Like, what would you tweak about that?

**[0:05:59] Tsavo Knott:** Yeah, I think the, you know, I think there's this contribution of context to this shared layer, right? And the shared layer will be the future of an AI native individual, an AI native team, and inevitably an AI native company. It's interesting. One of the general partners over at Y Combinator, he recently said, like, the the core to becoming AI native period is to record everything, which kind of touches on this idea of, hey, these systems need to have all context, and that context must be centralized.

**[0:06:29] Tsavo Knott:** Now I wanna touch on one last thing, which I think is is interesting when it comes to the shift of roles, and that is the top individual performers are gonna also need to start taking initiative to tap into this collective context. So, for example, you have a a sales team and they're putting up the context of how this call went with their, you know, customer Walmart. Then maybe you have product, uh, team members that wanna tap in and say, Hey, I'm gonna take the initiative to ask the shared context system.

**[0:06:58] Tsavo Knott:** Maybe it's an agentic system, maybe it's pieces, right? But m I wanna ask, how did that call go? And is is there anything that we need to get a jump on from a product side to tighten that deal up, right? Or to, you know, to to get ahead of where the ball is going with this customer? And before what we would see is maybe there's a product manager, there's a a sales manager. Maybe they have a discussion, they're reflecting, and they're like, Hey, the customer is asking for XYZ.

**[0:07:24] Tsavo Knott:** Now we've got to break this down into timelines, you know, we've got to delegate this out to the team. And they were kind of doing this up and over, uh, type of context sharing, right? And I feel like now that context sharing is is becoming flatter. And it has the ability for anyone on product to tap into the exact details of what's going on with sales for the intent of accelerating, you know, the the progress with a particular customer.

**[0:07:49] Tsavo Knott:** So I think that's an interesting component too. Is it's not only the sharing of context, but it's now the initiative and the instinct to tap into that context as it pertains to your role. And this is a muscle that not a lot of people have, but I think the top performers and really, you know, individual contributors in companies will need to lean into.

**[0:08:08] Daniel Faggella:** Yeah, so it's it's almost like, I mean, I I. I personally can't remember work before the internet. I'm sure some of our listeners can. Uh, you definitely can, but I, I can, I can presume there was a time where people were mailing things and were looking for information through other people and through other stuff. When actually, like, no, the internet was good enough to get it done, you know, but you're not, you're not like getting used to just being able to couple keystrokes and then move on.

**[0:08:36] Daniel Faggella:** You know, you're still in sort of the old mode, but then eventually it had to become the norm. Anybody who insists on pen and paper and, you know, bringing stuff to the mailbox, uh, is gonna have a very hard time finding, um, a knowledge work in America today. Similarly, we could imagine X number of years in the future, probably much fewer than what it took to get America on the internet. Uh, it will, it will become increasingly challenging.

**[0:08:58] Daniel Faggella:** You say, well, you know, I I don't know, I don't talk much to like the bigger AI thing, I just kind of do my. It's like, how would you do your work? You you would have you'd need context from supply chain and finance before you make this kind of decision. Like, you're gonna wait till next week to make a meeting. Our competitors aren't doing that, so you're gonna have to work off of that n newer sorts of stuff.

**[0:09:18] Daniel Faggella:** So again, as you had said, it's kind of a muscle, cause people aren't used to this being the way. But to your point, if you want to be a standout contributor, like, you've got to embed as deeply as you can now.

**[0:09:28] Tsavo Knott:** Oh yeah, and and I'll say this, um, I think we actually have a good amount of systems that people can share context, right? So, like, you know, we have, say, for example, again, you're on the product team and you know, the sales team and the the the managers, the interface between the two. They put up Jira tickets, they're tracking issues in GitHub. They're like, hey, you know, these are the things we need to accomplish. Maybe it's tagged for the Walmart, you know, conversation or something like that.

**[0:09:53] Tsavo Knott:** I think that, you know, we have lots of places that we can put context. I think one of the challenges that we're starting to see right now is that this context is living in many different places, right? So you have a lot of systems, both global systems, but also bespoke systems. There's not a ton of process, you know, maybe people are using Obsidian and Notion, and you know, all these kind of like one-off tools. Um, and you have, you know, a variety of note takers, a variety of different systems that you're inherently adopting on your own to to keep track of this with the intent to share and retrieve later.

**[0:10:25] Tsavo Knott:** And I think that's causing a a large amount of fragmentation of context across an org. And I think that's one of the other, you know, components. When it comes to what are the things that teams are struggling with, what are the things that individuals are struggling with, is, you know, we're recording a whole bunch of stuff, but it's in a bunch of different places. And I think that creates challenges when it comes to all right. Where is this thing?

**[0:10:47] Tsavo Knott:** How do I share it? How do I access it, you know, as efficiently as as possible? So, yeah, I wanted to touch on that as a as a pain point as well.

**[0:10:55] Daniel Faggella:** Totally well, and and I know we're gonna touch on it, but there's clearly security elements of this, governance components of this, right? Maybe every single conversation in finance shouldn't just be accessible to, you know, anybody under the sun or moon who wants to know whatever. And maybe every bit of context on, you know, a customer maybe isn't something that everybody in the company should necessarily know. The problems that this hospital system or this bank are dealing with right now that we're trying to solve, you know, m may or may not be something that would be right to sort of share at a broader context than it needs to be, et cetera.

**[0:11:25] Daniel Faggella:** So, yeah, that that's, that's certainly part of the mix. The things you're highlighting here, adapting to what it is to be a new individual contributor, which is, you know, recording all the context and then embedding and drawing from it as much as possible. Feeding into the meta brain in a way that's productive and then taking the stuff that shouldn't be on your plate, cause the meta brain can do it. And just chucking it up there. So you can focus on the places where where you can still be as as valuable as possible.

**[0:11:49] Daniel Faggella:** But then also to your point, okay, well, what tools are gonna do that? What's gonna get recorded? Where who gets to get access to what? Like, like, Yeah, sure, we record all those meetings, but that's in Otter instead of Firefly, instead of this, instead of that, right? Literally, I mean, we see this stuff all the time. So I guess with that said, we can get into kind of the new wave of capabilities. Now you guys have started to kind of move in this direction, and and there's kind of a whole panoply of kind of, like, what needs to evolve.

**[0:12:14] Daniel Faggella:** So, in the face of all of that, this new context heavy man and machine coordinating, you know, everybody's a PM world, what are the capabilities emerging to address that specifically?

**[0:12:25] Tsavo Knott:** Yeah, so I think I think it is, it's interesting, right? So today, what we see is, you know, when you're using a tool like Cursor or Claude Code or Codex, or, you know, some of these other tools out there, you know, uh, cloud co-work. I think what we're seeing is we're connecting it again to these bespoke systems, right? So you, you add a Notion connector, a Google workspace connector, a three sixty five connector. And then maybe you've got a whole bunch of markdown files that, you know, you instructed Claude to record documents about, you know, this project and that project and that initiative, and this is happening at scale.

**[0:12:58] Tsavo Knott:** Some people have these connectors, other people have, you know, different connectors. Some people take and organize information in one way and others in another way, and it's becoming very messy. And then also, by the way, you add this ability to vibe code a dashboard or vibe code an app, or vibe code a pipeline, and you've just got again, like times one zero bespoke microsystems, if you will. And so, yeah, there there's almost this like big bang going on right now with with AI, and it feels very, very chaotic.

**[0:13:29] Tsavo Knott:** It feels like high entropy, low levels of systems, low levels of standardization, low levels of consistency across how everyone in a company does their work. And so one of the things that I think we need to solve very quickly as everyone starts to do more things across a larger surface area, and, you know, generally across a larger volume of context. What we need to do is we need to. One, standardize the way that information is captured, standardize the rate and the timing in which information is captured, and standardize the ways in which it's fed to individuals, but also their agents.

**[0:14:05] Tsavo Knott:** And what I mean by this simply is you need one universal context substrate across everything that you're doing that, you know, independent of what you're using, be it cursor, codex, cloud, independent of of any system you're using, they can tap into it. And so I think this concept of this universal context substrate across all of the work you've done, all the work your team has done, and all the work that's being done within a company, that's that next level, right?

**[0:14:33] Tsavo Knott:** So it's almost like an everything CRM, if you will. And so I think until we have that standardized, and until we're capturing the level of detail that we hold in our brain individually, we're not gonna get to this efficiency in context transfer between humans and agents, humans and humans, and agents and agents. I think that's the biggest thing that we're trying to tackle here.

**[0:14:56] Daniel Faggella:** yeah, I mean, well, so this is the change management piece, right? Because I'm sure we're gonna get into the use cases of like, hey, Tsavo, like, you know what, what are we building now? Like, what? What could I ask in AI when I'm in the middle of X that could help me with Q? And I'm sure we are gonna get into some examples and make this tangible for people. But to your point, to unlock all the use cases we could talk about, and then 1 trillion times more, we we'd first need to sort of be s grounded in the same.

**[0:15:19] Daniel Faggella:** Again, substrate's a fine way to phrase it for for all this context, just to address that change management piece. Cause I wanna make this actionable for listeners here. You know, I'm imagining a company, maybe it's a scale up, maybe it's a mid sized firm, maybe it's an enterprise firm. And they're saying, man, I can resonate with this. I I sort of think we sh, you know, here in marketing, we need to know what's going on with inventory, we need to know what's going on with sales.

**[0:15:40] Daniel Faggella:** And the fact that I gotta bother salespeople, that's silly. I mean, we, this has gotta stop, you know, I. We should be able to make these kind of decisions with this context layer. Great, what's it gonna look like? To get the champions together in the room to decide on the tech we're gonna go with, the cases w like all the things we wanna record, maybe maybe some things we don't want to, if there are some of those things, right?

**[0:16:00] Daniel Faggella:** So what are all the places we're gonna record? And not, um, who does it? Is it just everybody's got? We have seven duplicates of the same recording somewhere, or does that get handled somehow? And then where does that go? And then s centrally, we'd have to decide on governance considerations around topics and users and all this. There there would be like, there'd be like a handful of pretty robust sessions to decide on the things that you articulated within any company bigger than a dozen humans.

**[0:16:25] Daniel Faggella:** And so how do you see this go well? Like when when people start to get context really flowing, I know that there's some companies that are, you know, using pieces here that are on the cutting edge already, you know, and maybe some of them are kind of a template for what this looks like when it goes well. How have you seen change management work?

**[0:16:41] Tsavo Knott:** Yeah, so it's it's interesting. Like for us, we see a very clear pattern inside of the companies that are adopting, you know, some of this, this work that we're doing in this like universal context space, right? And what we've seen in this pattern is under resourced kind of individuals, right? Product managers, people who are interfacing with a lot of different team members, a lot of different role types. So, for example, the executives, right? So, chief operating officer, CEO, director of technology, chief innovation officer.

**[0:17:11] Tsavo Knott:** They're kind of the in in innate routers of context in the company, and they are looking for, hey, I need to get the context around this question answered very quickly. I need to be able to pass that off to an agent to organize it, summarize it, create a, you know, executive brief to share with, you know, leadership, or to share with the with the team. And so they're suffering right now because their team is able to move very quickly across a broad surface of of problems, but they need to access this information in a way that was not possible before.

**[0:17:45] Tsavo Knott:** So what I mean by that, like tangibly, is say, for example, you know, in our company, we are doing iteration on product and we need to say, uh, as a as a product manager, I need to say, how many people have we seen this quarter? Come in on this version of Windows having this type of issue with native audio, and I need to be able to access that information very quickly. Like, we're on a call with three of the engineers working on native audio, I need to be able to tap into that historical context, roll it up and share it.

**[0:18:15] Tsavo Knott:** So in that case, I'm a context router, right? And there's a lot of these these context router individuals in the company that are struggling with this right now. So I think that's where it starts. But I'll say this. One last component is for them to be able to access context and route it. We need to be able to record it at the lowest level. So, for example, for me to be able to ask that question of how many support issues have we seen on this platform related to this problem?

**[0:18:43] Tsavo Knott:** I need to have everyone on my QA team and my support team record these conversations they're having, record the interactions with users, record the cycles that they're doing out there on the front lines for me to be able to tap into it. And this, what I would say, is kind of like naturally sets up your boundaries of context sharing. So I think that maybe maybe touches on a couple of the components, but I think that's the biggest thing.

**[0:19:08] Tsavo Knott:** Is we need to record at the right individual contributor level, but we need to provide access at the right management. Or, you know, the individual who's routing this context, I think that's the yeah,

**[0:19:20] Daniel Faggella:** Have you seen these? So this this context router role? There's probably people listening right now who are like, I'm that like, that's my life, uh, and and and presumably, there are things other than forwarding emails and copying and pasting Slack messages that they could be spending their entire day on. But but nonetheless, they find themselves where they are. So is it, is it often those people who are the catalyst to, Hey, let's get everybody together, like, let's figure out the system here.

**[0:19:46] Daniel Faggella:** Like, cause they know you guys always need this, you guys always need this, you guys are confused about this stuff every day, and it's all sitting here already, you know? And and so they sort of have that knowledge of, like, where the walking paths already are, uh, and

**[0:19:59] Tsavo Knott:** that's right.

**[0:19:59] Daniel Faggella:** then they can say, let's get together and make this. Like, are are they the right person to spark this dialogue?

**[0:20:04] Tsavo Knott:** absolutely. And I think those are the champions in companies that are saying the biggest pain point I'm having right now is everything around me is moving a thousand miles an hour. And I ask someone for something and they're like, Hey, I, you know, I I think it's over here. I got to figure this out, whatever else. And and their whole context sphere is exploding as well, cause they're doing ten x the amount of work. So, like, I think I did this a couple weeks ago, but I know that, you know, Tsavo's asking for it right now.

**[0:20:29] Tsavo Knott:** I got to go dig it up. And there's this just natural back pressure occurring of, like, just in time context delivery, right? So when I ask you for something, you know, your world, Daniel is is maxed out, right? You guys are are just touching across so many, uh, episodes, so many initiatives you guys are doing more than ever over there at Emerj. It's very hard for you to tap into and say, what are all the talking points?

**[0:20:52] Tsavo Knott:** And and what's been the evolution of our relationship with Tsavo and what they're building over there at Pieces over the last year and a half? And how should we tee up these these next episodes? Like, for you to go dig that up? It's at least a two to three day, you know, kind of initiative with a couple of your team members. That's where we want to take that process and just reduce it down to a couple of hours, right?

**[0:21:13] Tsavo Knott:** You should just be able to tap into that. And so we see these champions, for example, a director of technology, they're interfacing with CIO, uh, CIO, the CEO, the CTO, you know, the all their engineering teams, all their security teams. They're the perfect example of this context routing individual. So I think from there, they become the champions of, hey, we need all of our context centralized, available to be resurfaced, historically captured and and indexed. And then the next natural step is they don't have enough context.

**[0:21:46] Tsavo Knott:** And so what they say is everyone needs to have the same system of ingesting, storing, and resurfacing context. And that's where you naturally land in this world of one underlying system across everyone's work that, again, captures, organizes, and resurfaces context, ambiently, in the same standardized way across the entire company. I think that's where we land. Uh, if you fast forward a bit,

**[0:22:11] Daniel Faggella:** Yeah, so, and we're gonna get into really a a fistful of additional concrete use cases that you've seen be useful in the real world as you guys have developed product and, you know, gotten people to roll on here. But I, I will just say kind of off the jump, I I feel like if a playbook is developed for what you had just mentioned, that whole getting on the same page, figuring out the context, connection, et cetera, et cetera.

**[0:22:33] Daniel Faggella:** Like if a playbook gets developed for that. Everybody and their mother knows how to talk to their AI individually, and and everybody knows everybody else is using it in wacky, wild ways across the company. But I think this is everybody's issue right now, on some level, right? S to some degree or another, this is everybody's problem, and and and really does need to be smoothed out from a like a playbook standpoint. As you've started to see some of this, you know, c come come to bear, there's probably been some use cases that consistently, maybe in certain kinds of departments.

**[0:23:01] Daniel Faggella:** We see a lot of sales teams doing this, or a lot of, I don't know, procurement folks doing that, or whatever the case may be. What what for you are the capabilities that people are using the most with pieces? What are you having to build towards because people are asking for it? Like, I think looking at market pull is really helpful, cause the people listening are g same kind of users. Talk about those exciting areas of the present and the near future for this exact kind of stuff.

**[0:23:22] Tsavo Knott:** yeah, so I'll, I'll, I'll talk about one kind of concrete use case, which was which is extremely powerful. So, uh, we're coming up on research and development tax credit seasons, you know, so here in in the United States, and R and D tax credits are are interesting. So over here at Pieces, you know, the short of it is is we're working on this similar type of solution, this universal context substrate. And it is not only the context that you've done recently, but it's the context, it's it's really every moment across every team member's day for the last X amount of time, right?

**[0:23:55] Tsavo Knott:** Since installation. And so recently, you know, we're like, Okay, we've got to do the R and D tax credit. And I just in cursor, uh, used a variety of sub agents to say, Hey, look at my last twelve months. What are all the projects that I worked on? How much time did I spend on them? And then what went to production and what did not go to production? AKA, what do we categorize, you know, time, expenditure, project details, you know, whatever it might be, over the last twelve months as research and development.

**[0:24:26] Tsavo Knott:** And then we just extrapolated that out and I said, Okay, great, you've done this for me. Now I need you to do this twenty two more times across every individual in the company to get a comprehensive snapshot of what in the last twelve months was research and development. So that's one powerful way to say, Hey. For over a year now, we have had this universal context substrate. And I just turned a process that would have maybe taken a week or two across everyone's time into a couple of hours, right?

**[0:24:53] Tsavo Knott:** And that's what we mean by being able to tap into this, like, long-term universal context. Now, the funniest thing about that is no one in the company consciously took notes about what they did and if it went to production and how much time they took on it over the last year, right? Like, no one did that. And that's the funny thing about context is oftentimes you don't realize you should have done that until you realized you didn't.

**[0:25:19] Tsavo Knott:** And now you're, you're in this massive pain point. So that's one example. Um, another example that we've seen out in the wild quite extensively is onboarding. So people are jumping around, you know, they're able to knock out a a series of sprints on a project, and then they circle back to that project a couple of weeks later, right? A month later. And maybe they're onboarding a new team member or their role is switching. They need to be able to say, when was the last time I worked on this project?

**[0:25:45] Tsavo Knott:** What were all the architectural decisions we made, what were the comms we had, what were the initiatives, what were the goals, what were the KPIs? And how do I roll that up so I can hand it off to a team member who's joining the project? So we see that all the time of, you know, someone's coming into a sales team. What's the playbook? You know, how do we? How do we talk to customers, what are the things to touch on, what are the highlights, what are those SOPs, right?

**[0:26:09] Tsavo Knott:** Standard operating procedures. That's another component that's just now available at your fingertips if this context is captured, organized and resurfaced. And then you kind of get the, you know, okay. There's maybe once, once a year that I need to look back at at some deep stuff, or once every couple of months I need to onboard a new team member to the every single day use cases. And what I mean by that is we have a good example of a customer story that we we heard out in the wild recently.

**[0:26:37] Tsavo Knott:** Which is when an SDR is on a call and they're selling insurance today, one of the things that they have to do is the. They're all in a room, everyone's in a room, and they're on a call with a customer. And the customer says, You know, hey, I live in Alaska, I need to do this policy, you know, do you think you can do this? Here's a little bit about myself. And so, you know, the the SDR will look up in the CRM, like Alaska policy, you know, other customers, whatever else.

**[0:27:02] Tsavo Knott:** They don't find anything and they're still on the phone with the customer. So then what they do is they walk over to their colleague's desk and they're like, Hey, I can't find anything in the CRM. Do you know anyone on the team who's done a policy like this in Alaska recently? And they're like, No, you know, I don't know personally, but I think maybe talk to John. So then they walk over to John's desk. And they have a very specific metric, which is if this customer is on the phone for over twenty minutes, the probability that they become a customer drops below 40%, right?

**[0:27:31] Tsavo Knott:** It's it's effectively not happening. And so that type of scenario right there should be. I'm on the call with a customer. I ask a query inside of whatever agent I'm using, right? Co-work, Cursor, Pieces itself, whatever it might be. But it taps into this context substrate, which is what pieces is is plugged into. And you just simply ask, When was the last time someone on my team did an insurance policy for this lookalike customer in Alaska?

**[0:27:58] Tsavo Knott:** And it shouldn't be this twenty minutes. Try try to get the information here, go talk to someone there, you know, whole thing. And you definitely should not risk that customer's probability of having a successful interaction with your company. It should be I need to access this information. I can't remember exactly where, I don't exactly know who, but this information's out there. I should be able to access it within, you know, two to three minutes, just right here with my my agent.

**[0:28:24] Tsavo Knott:** Those are concrete every single day use cases where, you know, you're reducing time to information. Therefore, you're increasing customer, you know, customer enjoyment, customer success, and you're and you're closing more deals, right? So those are some of the the big ones that I've seen, uh, but I'll pause there. I mean, any commentary on some of that?

**[0:28:43] Daniel Faggella:** yeah, I think, well, I I think the I know this exists somewhere, but like, who even would know sort of thing I is is gonna be common to almost everybody listening in. Well, I think I think there's maybe two cases. If I even think about my own life and work, and and uh, even the people we've interviewed. There's categories where you it's sort of like, Okay, I know this person or department has it, but now I gotta ask somebody that that happens.

**[0:29:07] Daniel Faggella:** There's also, gosh, somewhere this exists, you know? If I throw the keywords into Google Drive and I, you know, I sh. Surely, if we dig through enough of the spreadsheets that have these two words, cause I think this should be in there somewhere, the one of these has gotta be it. There's, there's those like this has to exist somewhere, or I know these people have it. Scenarios that come up all the time, and it sounds like maybe the places where you run your face into that repeatedly could be starting points for, you know, adoption, kind of getting rolling.

**[0:29:40] Daniel Faggella:** Um, from what you're articulating, cause I think that that example with the, you know, insurance example, every, every sales team, marketing team, whatever has their equivalent of that in a certain circumstance. Are you seeing those as spearheads of adoption? Like, are there any interesting patterns about, like, where people dive first? So, of course, everybody that sells a platform, it's like, use it everywhere. But we all know like they tend to get started in little clusters, like lighthouses that get them excited enough to get rolling.

**[0:30:05] Daniel Faggella:** And then you can expand. Anything interesting coming from from your, uh, user behavior experience there.

**[0:30:12] Tsavo Knott:** Well, the other interesting thing is, you, you, it's, it's so funny, right, you know? So you're like, Okay, I think this bit of information is somewhere. And that bit of information could be a Google Doc, it could be, uh, you know, an overview deck, it could be, you know, uh, a policy that you did, like, like the example we shared earlier. But the the funny thing is okay, once you find that thing, the next natural question is, how did this thing come to be right?

**[0:30:38] Tsavo Knott:** And so there's a lot of other solutions out there, for example, that help you find things. So glean, you know, you might have heard of Glean here and there, but it's kind of this, you know, enterprise search. It'll find you the artifact. But what the artifact does not hold in and of itself is how that thing got made, the decisions that went into it, the comms that were behind it, the iteration that went into it. And and that work in progress journey is the other value add component here.

**[0:31:04] Tsavo Knott:** And oftentimes, you know, they say you can copy the the thing itself, but you cannot copy how it was made. And when you go to make another document, right? When you go to write up a new insurance policy, or when you go to create a new enterprise overview deck, or create a new, you know, outline for a podcast. You want to tap into those granular decisions that were made before that you want to remember again, so you don't like make the same mistake twice.

**[0:31:32] Tsavo Knott:** Or you capture that nuanced psychology that made the previous, you know, XYZ initiative successful. And so you have the thing itself, but also everything that went into the thing. And those are the two sides of context that we're talking about here. So I think it's important both on a on a retrieval of that artifact, but also on a retrieval of the the background, you know, the work in progress behind that artifact. And I think those are the the points that really stand out to people, those are the the things that are saying like, Okay, I'm a leader.

**[0:32:05] Tsavo Knott:** I have to come up with strategy, I have to come up with, you know, ways to articulate to my team, to my agent. This is how we make things, and this is why we make them the way we make them. I think sharing that context is also very, very painful these days, because, you know, inherently within a team, you're all kind of a part of the same culture, especially for people that have been there for a while, they get it.

**[0:32:28] Tsavo Knott:** But then, as soon as you need to transfer this over to a cloud agent or transfer it over to a cursor agent, it's like explaining the whole, you know, all these nuances from ground fr, you know, just from scratch, right? And that's also very, very painful. So I think for us, what we see out there is the people that are responsible for holding all of this knowledge in their head. They're, you know, they're the go-to person.

**[0:32:51] Tsavo Knott:** Everyone on the team comes to them, they're like, how do we do this again? Where is it? Why do we do it this way? Those are the people that really benefit from this universal long-term context.

**[0:33:02] Daniel Faggella:** Yeah, yeah, again, the people that are feeling it already, uh, today, the con, the context router person who would like to have a couple minutes to do something other than, um, so, uh, I guess bringing bringing ourselves, we got a we got a pretty concrete vision here of, like, the future we're moving towards some of the core barriers that get us there, uh, and some of the use cases that we can enable. I guess, you know, kind of parting notes, speaking to the departmental leaders, executive leadership, uh, or context routers who have some influence over what the heck tech adopt we adopt.

**[0:33:33] Daniel Faggella:** When it comes to getting started on this journey and and sort of moving towards this persistent shared memory, going from having what most people probably today. It's this department uses Otter mostly, this department uses Firefly mostly. These folks should all be using Claude for what they do, but there's only two of them doing it. You know, the these, like these, like these people, I don't even know. They're all using an LLM, but it's probably a different one.

**[0:33:56] Daniel Faggella:** This person's, this salesperson's got a Gemini gem, this salesperson's got a, you know, GPT. This is the fragmented scenario we find ourselves in today. When it comes from concrete steps to go from there to maybe one or two rings out, where a lot more unified, a lot less manual, human context routing, a lot more superpowers for each individual user. What would you advise for, kind of, again, a couple tangible steps forward?

**[0:34:18] Tsavo Knott:** I think the first thing is you have to identify the the you have to identify, like the things that you're still manually doing, right, when it comes to providing context to these agents, right? So the amount of times I hop on a call and I'm like, you know, okay, what, what tools are you using? And they say, you know, I'm using Otter, Firefly or Gemini note taking, and and I'm like, Okay, you know, how do you get that context from the note taker to Claude Code, right?

**[0:34:43] Tsavo Knott:** Or the note taker to co-worker, whatever it is? They say they literally pull up the notes, they copy it, they paste it, and then they say, I need to do XYZ, create some action items from these, you know, these conversations, and so on. And so I think the first thing is like, identify those points where context capture and context transfer is still manual, right? And it might be a note taker from calls, it might be, you know, documentation for Jira tickets or things like that.

**[0:35:09] Tsavo Knott:** But once you identify those like manual context transfers that are still going on, you have to think about, think to yourself, like, okay, that because it is manual, it's not automatic, it's not ambient. That means that we're perhaps only capturing, and therefore only able to transfer five of context, ten of context, twenty of context. And if you only have five of the bigger picture, or twenty of the bigger picture, then you're only gonna get, you know, a a a certain level of quality and a certain level of output from the tools you're already adopting.

**[0:35:43] Tsavo Knott:** So I think the first thing is like, figure out what context you're already capturing, how's it getting stored, how's it getting transferred, and then try to think about what con context you need but is not being captured. And then I think lastly is like, you've got tools that can do things right, like all these agent systems out there, they can do things. But I would say the focus that is is gonna be more and more forefront is systems that know things, right?

**[0:36:09] Tsavo Knott:** So again, back to this concept of this everything CRM, how does that everything CRM get populated? Is it manual? Is it automatic? And I think, like, those are the questions you want to ask yourself. Because at the end of the day, I I think the companies that will move to being AI native the fastest are the companies that record everything, right? They're the companies that have no details lost, no details missed. Have the work in progress journey, can access it, can transfer it between people and agents, um, as fluidly as possible.

**[0:36:40] Tsavo Knott:** And so I think, you know, the guidance here is hard because it's fuzzy, like we're all running around with a a bunch of agents,

**[0:36:46] Daniel Faggella:** Trying to figure out our own bespoke scenarios? Yeah, but you said anything at all about, like, a first couple steps, like, here's who's gotta champion this, here's who's gotta be in the room, here's the first two decisions you just gotta run with. You know, like anything like that, I'm sure will resonate on some level.

**[0:37:00] Tsavo Knott:** yeah, and you know, I think back to that point of like, you know, go to the people, you know. If you're leadership, go to the people who are routing the most context in your company, right? Go to the PMs, go to the the the folks putting the playbooks together. Put, you know, kind of looking at the SOPs, trying to think about processes and standardizations, and just figure out like, how are they feeling? You know, just check in with them, cause odds are they're feeling like they're developing severe ADHD and and and a case of, like, overwhelming, you know, stress from just the the volume of things hitting their desk every single day.

**[0:37:36] Tsavo Knott:** And I think, you know, ask the question, how can we make your job easier, right? And if that question comes down to an answer that's even remotely your job becomes easier. If you can tap into knowledge that's guaranteed available, then that's like an example of, hey, we need to unify and and centralize our context, right? That's an indicator that you need some type of tool that's just running in the background because your individual contributors are not recording enough, right?

**[0:38:03] Tsavo Knott:** And I think if they're feeling stressed or they're feeling like there's a bottleneck of information access, then that's an indicator that you're not capturing enough context across the company. And I think that's what I'm trying to say.

**[0:38:14] Daniel Faggella:** Yeah, yeah, and and maybe what parts to, you know? Getting everybody to use the the the same tool in the same way on the same day might not happen. It's gonna roll out through departments or whatever, but m maybe figuring out where that shockwave needs to emanate from. You could just refer to the humans that are doing the job of what to say I should be doing and just say where, you know, and then and they'll.

**[0:38:34] Daniel Faggella:** They'll be able to maybe point the finger of where their biggest headaches actually live. And probably almost everybody tuned in has got one or two people that they're like, Yeah, they'd know, uh, you know, I'd be lying if I said I didn't have one of my own. So Tsavo, I appreciate us getting getting to some actionability here in addition to the grand vision. I think we will live to see a future very soon where this does become the norm, but it's gonna be a journey.

**[0:38:56] Daniel Faggella:** And at least we got a little bit of some guidance on getting there. So I know that's all we have for time, but thanks so much for being with us again.

**[0:39:00] Tsavo Knott:** 100%, I mean, Daniel, as always, great time, and, you know, I love kicking it on some of this stuff. And of course, you know, we're we're very, very involved with the space. So if anyone out there listening, uh, is interested, let's let's get together, it'll be awesome.

**[0:39:29] Narrator:** Wrapping up today's episode, I think there are three key takeaways from our conversation with Tsavo. First, the shift from traditional individual contributor roles toward AI native contributors requires teams to capture and share context consistently so work can move across border surfaces without friction. Second, fragmented context scattered across tools and personal systems is now a core operational bottleneck, and enterprises need a unified substrate to make information accessible to both humans and agents. And finally, the emerging context router role highlights how leaders must enable rapid retrieval of institutional knowledge so decisions, coordination, and execution can keep pace with the speed of modern AI driven workflows.

**[0:40:09] Narrator:** According to Edison research, 79% of Americans listen to online audio monthly, an estimated 228 million people. Podcasts offer a rare opportunity to capture twenty plus minutes of attention with VP plus leaders in America's largest enterprises. Emerj reaches one million listeners every year. Learn more about how these conversations drive pipeline for AI solution providers. Download our media kit at Emerj dot com slash ad one. That's e n e r j dot com slash ad and the number one.

**[0:40:38] Narrator:** For further executive level analysis and to join our network of leaders delivering workflow impact with AI, visit Emerj dot com. On behalf of the team at Emerj, we'll see you on the next episode.