AI in GTM at Notion — Flora Liu

Well, first of all, hi everyone. I'm an Well, first of all, hi everyone. I'm an engineer on the product growth team at engineer on the product growth team at engineer on the product growth team at Notion and now working on the GTM Notion and now working on the GTM Notion and now working on the GTM engineering team. A year ago, I would engineering team. A year ago, I would engineering team. A year ago, I would have told you that building a GTM system have told you that building a GTM system have told you that building a GTM system was a marketing ops problem. And today, was a marketing ops problem. And today, was a marketing ops problem. And today, I think it's one of the most interesting I think it's one of the most interesting I think it's one of the most interesting distributed systems problems that I've distributed systems problems that I've distributed systems problems that I've worked on. worked on. worked on. GTM at most companies involve a GTM at most companies involve a GTM at most companies involve a spiderweb of tools like what you see spiderweb of tools like what you see spiderweb of tools like what you see here, and they're stitched together by here, and they're stitched together by here, and they're stitched together by customer notes, proposals, contracts customer notes, proposals, contracts customer notes, proposals, contracts that are passed back and forth. Over the that are passed back and forth. Over the that are passed back and forth. Over the last few months, a small team and last few months, a small team and last few months, a small team and myself, um, we've been trying to turn myself, um, we've been trying to turn myself, um, we've been trying to turn this spider web into a unified GTM this spider web into a unified GTM this spider web into a unified GTM system. We're still in the midst of it, system. We're still in the midst of it, system. We're still in the midst of it, but we've learned a ton that I'd like to but we've learned a ton that I'd like to but we've learned a ton that I'd like to share with you today. share with you today. share with you today. So, this isn't really a new problem. So, this isn't really a new problem. So, this isn't really a new problem. We've been wrestling with pieces of it We've been wrestling with pieces of it We've been wrestling with pieces of it for years. Life cycle messaging, product for years. Life cycle messaging, product for years. Life cycle messaging, product recommendations, recommendations, recommendations, sales automations, customer data, and sales automations, customer data, and sales automations, customer data, and onboarding. But the solutions were onboarding. But the solutions were onboarding. But the solutions were fragmented because the underlying fragmented because the underlying fragmented because the underlying technology forced them to be.

technology forced them to be. technology forced them to be. Then over the winter break, our CEO Ivan Then over the winter break, our CEO Ivan Then over the winter break, our CEO Ivan built spent it building a video game and built spent it building a video game and built spent it building a video game and he came back convinced that software he came back convinced that software he came back convinced that software engineering could be applied to many engineering could be applied to many engineering could be applied to many problems that were previously unwieldy problems that were previously unwieldy problems that were previously unwieldy or too costly. At the same time, our or too costly. At the same time, our or too costly. At the same time, our ability to execute skyrocketed with ability to execute skyrocketed with ability to execute skyrocketed with agentic technology and the breath of agentic technology and the breath of agentic technology and the breath of problems we could solve did too. problems we could solve did too. problems we could solve did too. the costly, time-consuming, and the costly, time-consuming, and the costly, time-consuming, and previously unsolvable spaghetti could previously unsolvable spaghetti could previously unsolvable spaghetti could now be sorted. So, what we found is that now be sorted. So, what we found is that now be sorted. So, what we found is that GTM had become a systems problem. That GTM had become a systems problem. That GTM had become a systems problem. That made us realize we could chip away at made us realize we could chip away at made us realize we could chip away at this holistically. this holistically. this holistically. In case you don't know, notion's In case you don't know, notion's In case you don't know, notion's platform is a collaborative brain for platform is a collaborative brain for platform is a collaborative brain for human and agents to think together. Over human and agents to think together. Over human and agents to think together. Over the years, we've evolved into a context the years, we've evolved into a context the years, we've evolved into a context layer for your company and AI agents can layer for your company and AI agents can layer for your company and AI agents can act on it. Notion's business moves act on it. Notion's business moves act on it. Notion's business moves between self-s serve growth and sales between self-s serve growth and sales between self-s serve growth and sales assist. Um, and customers move between assist. Um, and customers move between assist. Um, and customers move between these two motions all the time.

these two motions all the time. these two motions all the time. The problem is that customers experience The problem is that customers experience The problem is that customers experience one journey, but internally it is one journey, but internally it is one journey, but internally it is supported by disconnected systems that supported by disconnected systems that supported by disconnected systems that don't actually talk to each other very don't actually talk to each other very don't actually talk to each other very well. These processes were rife with well. These processes were rife with well. These processes were rife with human error and put a lot of cognitive human error and put a lot of cognitive human error and put a lot of cognitive burden on our teams. For example, sales burden on our teams. For example, sales burden on our teams. For example, sales reps are probably not the strongest at reps are probably not the strongest at reps are probably not the strongest at managing systems, but their strength is managing systems, but their strength is managing systems, but their strength is in sussing out human signals during the in sussing out human signals during the in sussing out human signals during the buying process. So, every time they had buying process. So, every time they had buying process. So, every time they had to context switch between after a call, to context switch between after a call, to context switch between after a call, doing research, drafting follow-up, they doing research, drafting follow-up, they doing research, drafting follow-up, they were spending less time with our were spending less time with our were spending less time with our customers and customer problems. customers and customer problems. customers and customer problems. Most companies have separate systems for Most companies have separate systems for Most companies have separate systems for sales assist and productled growth. And sales assist and productled growth. And sales assist and productled growth. And this is actually also true for us. this is actually also true for us. this is actually also true for us. Marketing run runs on one set of tools, Marketing run runs on one set of tools, Marketing run runs on one set of tools, sales on another, customer ops on a sales on another, customer ops on a sales on another, customer ops on a third. But all of them are looking at a third. But all of them are looking at a third. But all of them are looking at a customer independently and making customer independently and making customer independently and making decisions separately. So what we set out decisions separately. So what we set out decisions separately. So what we set out to build is a single decisioning system to build is a single decisioning system to build is a single decisioning system that spans self-s serve growth and sales that spans self-s serve growth and sales that spans self-s serve growth and sales assist. and it can help the customer assist. and it can help the customer assist. and it can help the customer decide the next step so that everything decide the next step so that everything decide the next step so that everything is cohesive.

is cohesive. is cohesive. Our vision is for this system to be Our vision is for this system to be Our vision is for this system to be programmable, programmable, programmable, proactive and continuous. proactive and continuous. proactive and continuous. When we started, we were faced with some When we started, we were faced with some When we started, we were faced with some challenges. There was so no single challenges. There was so no single challenges. There was so no single source of truth. So customer data was source of truth. So customer data was source of truth. So customer data was spread across Salesforce, Gong, spread across Salesforce, Gong, spread across Salesforce, Gong, Outreach, Zoom Info and many more. Outreach, Zoom Info and many more. Outreach, Zoom Info and many more. Product usage lived in Snowflake and a Product usage lived in Snowflake and a Product usage lived in Snowflake and a decade of the most important context decade of the most important context decade of the most important context lived in notes and meeting docs. Yes, lived in notes and meeting docs. Yes, lived in notes and meeting docs. Yes, our sales reps do use notion for that our sales reps do use notion for that our sales reps do use notion for that too. Notion employees were actively too. Notion employees were actively too. Notion employees were actively using MCPs and their own agents to solve using MCPs and their own agents to solve using MCPs and their own agents to solve problems already, but they were problems already, but they were problems already, but they were innovating within their own departments. innovating within their own departments. innovating within their own departments. So it was single player mode or you So it was single player mode or you So it was single player mode or you could say here single department mode. could say here single department mode. could say here single department mode. Marketing built tools for tool uh Marketing built tools for tool uh Marketing built tools for tool uh marketing sales built tools for sales marketing sales built tools for sales marketing sales built tools for sales and each tool served a tiny slice of and each tool served a tiny slice of and each tool served a tiny slice of that customer journey and it would make that customer journey and it would make that customer journey and it would make it really hard to create something that it really hard to create something that it really hard to create something that was more holistic. was more holistic. was more holistic. When we tried to automate across all of When we tried to automate across all of When we tried to automate across all of that we hit some roadblocks. First data that we hit some roadblocks. First data that we hit some roadblocks. First data quality conflicting systems of records quality conflicting systems of records quality conflicting systems of records wrong contacts tied to different wrong contacts tied to different wrong contacts tied to different accounts. One bad mapping was enough to accounts. One bad mapping was enough to accounts. One bad mapping was enough to lose trust for sales rep.

lose trust for sales rep. lose trust for sales rep. Secondly, data latency. Every vendor Secondly, data latency. Every vendor Secondly, data latency. Every vendor added a hop and this lag was causing us added a hop and this lag was causing us added a hop and this lag was causing us to act on stale data and that meant we to act on stale data and that meant we to act on stale data and that meant we were automating on yesterday's world. were automating on yesterday's world. were automating on yesterday's world. Third, and this is a big one, structured Third, and this is a big one, structured Third, and this is a big one, structured and unstructured data. The most and unstructured data. The most and unstructured data. The most important facts about a customer were important facts about a customer were important facts about a customer were left in notes like the champion just left in notes like the champion just left in notes like the champion just left or don't contact this customer left or don't contact this customer left or don't contact this customer again. Um or they're blocked illegal. again. Um or they're blocked illegal. again. Um or they're blocked illegal. And so these are exactly the types of And so these are exactly the types of And so these are exactly the types of notes that help sales rep move forward notes that help sales rep move forward notes that help sales rep move forward and decide what to do next. And if an and decide what to do next. And if an and decide what to do next. And if an automation couldn't see it or process automation couldn't see it or process automation couldn't see it or process it, it could do something it, it could do something it, it could do something catastrophically wrong. catastrophically wrong. catastrophically wrong. So our project team consisted of CX, So our project team consisted of CX, So our project team consisted of CX, RevOps, product, engineering, sales. And RevOps, product, engineering, sales. And RevOps, product, engineering, sales. And after brainstorming together, we all after brainstorming together, we all after brainstorming together, we all kept finding the same patterns kept finding the same patterns kept finding the same patterns underneath that complexity. underneath that complexity. underneath that complexity. Whoops. Oh. Every workflow could be Whoops. Oh. Every workflow could be Whoops. Oh. Every workflow could be reduced to four questions. What do we reduced to four questions. What do we reduced to four questions. What do we know about the customer? What should know about the customer? What should know about the customer? What should happen next? How do we execute that happen next? How do we execute that happen next? How do we execute that safely? And did it work? That became our safely? And did it work? That became our safely? And did it work? That became our architecture.

architecture. architecture. So the system has four layers. Know a So the system has four layers. Know a So the system has four layers. Know a context layer we can trust about every context layer we can trust about every context layer we can trust about every customer. Decide, choose the single next customer. Decide, choose the single next customer. Decide, choose the single next best step for them. Third, fire act and best step for them. Third, fire act and best step for them. Third, fire act and fire a concrete action that could be a fire a concrete action that could be a fire a concrete action that could be a life cycle email um an inapp nudge or a life cycle email um an inapp nudge or a life cycle email um an inapp nudge or a task handed to a rep and then learn task handed to a rep and then learn task handed to a rep and then learn watch what happened and feed it back watch what happened and feed it back watch what happened and feed it back into the decisioning so that it's a into the decisioning so that it's a into the decisioning so that it's a loop. loop. loop. But this architecture is missing But this architecture is missing But this architecture is missing something important. The most important part is that humans The most important part is that humans and agents are operating on the same and agents are operating on the same and agents are operating on the same loop. Concretely, this means that the loop. Concretely, this means that the loop. Concretely, this means that the context needs to be displayed so that context needs to be displayed so that context needs to be displayed so that humans and agents can read and operate humans and agents can read and operate humans and agents can read and operate on it together. on it together. on it together. So you can see that they're working in So you can see that they're working in So you can see that they're working in the same system, but they might have the same system, but they might have the same system, but they might have different roles. Agents do the different roles. Agents do the different roles. Agents do the repetitive work at scale like gathering repetitive work at scale like gathering repetitive work at scale like gathering context, researching, drafting context, researching, drafting context, researching, drafting recommendations, and writing artifacts.

recommendations, and writing artifacts. recommendations, and writing artifacts. Humans provide the judgment, adding Humans provide the judgment, adding Humans provide the judgment, adding nuance, deciding what to do next, and if nuance, deciding what to do next, and if nuance, deciding what to do next, and if a recommendation is correct. and owning a recommendation is correct. and owning a recommendation is correct. and owning the customer relationship. the customer relationship. the customer relationship. We found that instead of building an AI We found that instead of building an AI We found that instead of building an AI layer on top of our business, we layer on top of our business, we layer on top of our business, we designed our architecture so that the designed our architecture so that the designed our architecture so that the agent can operate as another operator agent can operate as another operator agent can operate as another operator within the same system as humans. within the same system as humans. within the same system as humans. Before we built the system, we made some Before we built the system, we made some Before we built the system, we made some choices about how we were going to choices about how we were going to choices about how we were going to implement this. Firstly, we deliberately implement this. Firstly, we deliberately implement this. Firstly, we deliberately chose not to let an agent talk directly chose not to let an agent talk directly chose not to let an agent talk directly to a customer. For sales assist to a customer. For sales assist to a customer. For sales assist workflows, humans stay in the loop by workflows, humans stay in the loop by workflows, humans stay in the loop by default and approve anything the agents default and approve anything the agents default and approve anything the agents do. The agents do the busy work. That do. The agents do the busy work. That do. The agents do the busy work. That decision that decision also has a decision that decision also has a decision that decision also has a security dimension too. If a prospect security dimension too. If a prospect security dimension too. If a prospect fills out a contact sales form online, fills out a contact sales form online, fills out a contact sales form online, we treat that as untrusted user input. we treat that as untrusted user input. we treat that as untrusted user input. And so trust boundaries don't break And so trust boundaries don't break And so trust boundaries don't break down, especially because there is an down, especially because there is an down, especially because there is an agent in the middle. agent in the middle. agent in the middle. Secondly, routing and eligibility became Secondly, routing and eligibility became Secondly, routing and eligibility became a first class primitive. Eligibility a first class primitive. Eligibility a first class primitive. Eligibility used to be scattered in all over the used to be scattered in all over the used to be scattered in all over the place. We had one check or rule in an place. We had one check or rule in an place. We had one check or rule in an email tool, another in sales and we email tool, another in sales and we email tool, another in sales and we pulled that all into one place so that pulled that all into one place so that pulled that all into one place so that these rules can be consumed across our these rules can be consumed across our these rules can be consumed across our codebase uh product sales engineering codebase uh product sales engineering codebase uh product sales engineering and these are like customer and these are like customer and these are like customer segmentations or signal signal segmentations or signal signal segmentations or signal signal definitions and then a single classifier definitions and then a single classifier definitions and then a single classifier will route what the customer should do

will route what the customer should do will route what the customer should do and this will actually prevent double and this will actually prevent double and this will actually prevent double sends from our system and create very sends from our system and create very sends from our system and create very cohesive communication across cohesive communication across cohesive communication across And last but not least, we decided that And last but not least, we decided that And last but not least, we decided that it was very important to own the contact it was very important to own the contact it was very important to own the contact layer and we decided to rent everything layer and we decided to rent everything layer and we decided to rent everything else. Since we are a lean team, we will else. Since we are a lean team, we will else. Since we are a lean team, we will not build our own email vendor or not build our own email vendor or not build our own email vendor or enrichment services like Clay. We use enrichment services like Clay. We use enrichment services like Clay. We use Clay and we believe that we understood Clay and we believe that we understood Clay and we believe that we understood our customers the best. So we will not our customers the best. So we will not our customers the best. So we will not um give that away. um give that away. um give that away. So let's get into what we built. So let's get into what we built. So let's get into what we built. The first step was to gather a The first step was to gather a The first step was to gather a consolidated view of all of our consolidated view of all of our consolidated view of all of our customers. customers. customers. Snowflake, which is our data warehouse, Snowflake, which is our data warehouse, Snowflake, which is our data warehouse, is where we compute this truth. We is where we compute this truth. We is where we compute this truth. We ingest data from all the vendors in our ingest data from all the vendors in our ingest data from all the vendors in our GTM stack to Snowflake. We run daily GTM stack to Snowflake. We run daily GTM stack to Snowflake. We run daily transforms and in some cases real time transforms and in some cases real time transforms and in some cases real time to produce a small set of modeled to produce a small set of modeled to produce a small set of modeled versioned entities and these are versioned entities and these are versioned entities and these are accounts, contacts, workspaces, accounts, contacts, workspaces, accounts, contacts, workspaces, eligibility, and facts. And this also eligibility, and facts. And this also eligibility, and facts. And this also has clear ownership of what teams or uh has clear ownership of what teams or uh has clear ownership of what teams or uh tools they come from and like tools they come from and like tools they come from and like timestamps.

timestamps. timestamps. Dynamob is our key value store and it's Dynamob is our key value store and it's Dynamob is our key value store and it's where we compute our truth or serve our where we compute our truth or serve our where we compute our truth or serve our truth. We publish a denormalized key truth. We publish a denormalized key truth. We publish a denormalized key addressable profile that agents can addressable profile that agents can addressable profile that agents can quickly query in milliseconds with no quickly query in milliseconds with no quickly query in milliseconds with no joins. We also persist agent uh joins. We also persist agent uh joins. We also persist agent uh persisted uh or generated artifacts and persisted uh or generated artifacts and persisted uh or generated artifacts and these are research snippets, summarized these are research snippets, summarized these are research snippets, summarized notes, rolling summaries and these notes, rolling summaries and these notes, rolling summaries and these unstructured data are also keyed by the unstructured data are also keyed by the unstructured data are also keyed by the same ids so that downstream systems can same ids so that downstream systems can same ids so that downstream systems can read all of this in one shot. read all of this in one shot. read all of this in one shot. So this data was normalized and of So this data was normalized and of So this data was normalized and of course we brought it into notion so that course we brought it into notion so that course we brought it into notion so that we could work with structured and we could work with structured and we could work with structured and unstructured data at the same time. So unstructured data at the same time. So unstructured data at the same time. So some of the data I showed you in the some of the data I showed you in the some of the data I showed you in the boxes earlier, there's like product boxes earlier, there's like product boxes earlier, there's like product usage data, there's activity log from usage data, there's activity log from usage data, there's activity log from across our vendor stack, and then we across our vendor stack, and then we across our vendor stack, and then we also have like unstructured data that also have like unstructured data that also have like unstructured data that like I mentioned that is most important like I mentioned that is most important like I mentioned that is most important for sales context with research reports for sales context with research reports for sales context with research reports and notes. and notes. and notes. And this turned out to be powerful for And this turned out to be powerful for And this turned out to be powerful for two reasons. First, our internal GTM two reasons. First, our internal GTM two reasons. First, our internal GTM teams didn't need to jump between many teams didn't need to jump between many teams didn't need to jump between many tools anymore. They could use notion tools anymore. They could use notion tools anymore. They could use notion itself, a tool they were already using itself, a tool they were already using itself, a tool they were already using to explore context, investigate to explore context, investigate to explore context, investigate investigate accounts, answer questions, investigate accounts, answer questions, investigate accounts, answer questions, and they could even take actions like and they could even take actions like and they could even take actions like sending to Nooks or um sending to sending to Nooks or um sending to sending to Nooks or um sending to outreach. This is a tool that they were outreach. This is a tool that they were outreach. This is a tool that they were very familiar with, and they didn't need very familiar with, and they didn't need very familiar with, and they didn't need to open seven tabs anymore. Secondly, to open seven tabs anymore. Secondly, to open seven tabs anymore. Secondly, because we weren't building an AI layer,

because we weren't building an AI layer, because we weren't building an AI layer, um humans, workflows, and agents could um humans, workflows, and agents could um humans, workflows, and agents could all operate on the same source of truth. all operate on the same source of truth. all operate on the same source of truth. In a very literal sense, we are using In a very literal sense, we are using In a very literal sense, we are using notion to grow notion. The next primitive we decided that we The next primitive we decided that we needed was a way to turn customer events needed was a way to turn customer events needed was a way to turn customer events into actions. The unit here is a signal. into actions. The unit here is a signal. into actions. The unit here is a signal. A signal is a single customer event A signal is a single customer event A signal is a single customer event that's important enough to change what that's important enough to change what that's important enough to change what should happen next for a customer. Some should happen next for a customer. Some should happen next for a customer. Some are userdriven like a customer hitting are userdriven like a customer hitting are userdriven like a customer hitting their a AI limit or maybe they reached their a AI limit or maybe they reached their a AI limit or maybe they reached out to contact contact sales. But some out to contact contact sales. But some out to contact contact sales. But some of them are not user initiated at all of them are not user initiated at all of them are not user initiated at all which are these external signal examples which are these external signal examples which are these external signal examples I listed here like company raising I listed here like company raising I listed here like company raising funding, hiring signals or shift in funding, hiring signals or shift in funding, hiring signals or shift in their tech stack. Those external signals their tech stack. Those external signals their tech stack. Those external signals are what allowed us to be proactive are what allowed us to be proactive are what allowed us to be proactive instead of reactive. instead of reactive. instead of reactive. So this is the signal service that So this is the signal service that So this is the signal service that watches the customer profile, decides watches the customer profile, decides watches the customer profile, decides whether a single action is available, whether a single action is available, whether a single action is available, decides who should own that action, and decides who should own that action, and decides who should own that action, and then it emits a concrete task following then it emits a concrete task following then it emits a concrete task following the architecture I described before. And the architecture I described before. And the architecture I described before. And this task could be for a human or an this task could be for a human or an this task could be for a human or an agent. If there's a task for sales rep, agent. If there's a task for sales rep, agent. If there's a task for sales rep, that actually just lands in their notion that actually just lands in their notion that actually just lands in their notion database and they can quickly view it database and they can quickly view it database and they can quickly view it and act on it.

and act on it. and act on it. What's what's interesting about the way What's what's interesting about the way What's what's interesting about the way we built our GTM systems is that if we built our GTM systems is that if we built our GTM systems is that if there is no signal about a customer, the there is no signal about a customer, the there is no signal about a customer, the marketing component of our system kicks marketing component of our system kicks marketing component of our system kicks in. We have a predictive engine that in. We have a predictive engine that in. We have a predictive engine that will recommend product features most will recommend product features most will recommend product features most relevant for that customer and it will relevant for that customer and it will relevant for that customer and it will send out life cycle emails and inapp send out life cycle emails and inapp send out life cycle emails and inapp nudges or multi-channel communication to nudges or multi-channel communication to nudges or multi-channel communication to drive a customer towards adoption drive a customer towards adoption drive a customer towards adoption automatically. So diving deep into a small slice of So diving deep into a small slice of what happens when we decide what action what happens when we decide what action what happens when we decide what action should be emitted. Um this is for like should be emitted. Um this is for like should be emitted. Um this is for like the sales workflow and we shadowed our the sales workflow and we shadowed our the sales workflow and we shadowed our best reps to capture something that was best reps to capture something that was best reps to capture something that was the most repetitive part of our job and the most repetitive part of our job and the most repetitive part of our job and encoded it as a durable multi- aent encoded it as a durable multi- aent encoded it as a durable multi- aent workflow. Every signal becomes a workflow. Every signal becomes a workflow. Every signal becomes a workflow on temporal which is something workflow on temporal which is something workflow on temporal which is something we rent and a single run will touch we rent and a single run will touch we rent and a single run will touch enrichment web search draft generation enrichment web search draft generation enrichment web search draft generation and more. Each of these is a network and more. Each of these is a network and more. Each of these is a network call that could fail or rate limit. And call that could fail or rate limit. And call that could fail or rate limit. And so temporal lets us focus on writing the so temporal lets us focus on writing the so temporal lets us focus on writing the sequential logic for our GTM use cases sequential logic for our GTM use cases sequential logic for our GTM use cases while it will handle the retries, ddupes while it will handle the retries, ddupes while it will handle the retries, ddupes um and handling and going back to um and handling and going back to um and handling and going back to exactly where failures left off. and one exactly where failures left off. and one exactly where failures left off. and one malformed transcript can't take down the malformed transcript can't take down the malformed transcript can't take down the whole batch which was really important whole batch which was really important whole batch which was really important to us. So an example of a cold outbound to us. So an example of a cold outbound to us. So an example of a cold outbound signal for us will have a research sub signal for us will have a research sub signal for us will have a research sub agent do concurrent researches then

agent do concurrent researches then agent do concurrent researches then it'll draft like an email and those it'll draft like an email and those it'll draft like an email and those emails uh there should be three of them emails uh there should be three of them emails uh there should be three of them so they're scored and then a review so they're scored and then a review so they're scored and then a review agent will pick the highest scoring one agent will pick the highest scoring one agent will pick the highest scoring one and make any updates if necessary. And and make any updates if necessary. And and make any updates if necessary. And this also operates on a loop um so that this also operates on a loop um so that this also operates on a loop um so that the email drafts are improved. And then the email drafts are improved. And then the email drafts are improved. And then when it's ready, this email draft will when it's ready, this email draft will when it's ready, this email draft will land in the sales task that is available land in the sales task that is available land in the sales task that is available for for them to act on. Um for more for for them to act on. Um for more for for them to act on. Um for more reactive signals after a follow-up call, reactive signals after a follow-up call, reactive signals after a follow-up call, the Gong transcript will come in. Our the Gong transcript will come in. Our the Gong transcript will come in. Our agent will parse the transcript and then agent will parse the transcript and then agent will parse the transcript and then um again it will extract the critical um again it will extract the critical um again it will extract the critical sales medpic data uh metrics economic sales medpic data uh metrics economic sales medpic data uh metrics economic buyer decision criteria plan and buyer decision criteria plan and buyer decision criteria plan and champion and draft a grounded followup champion and draft a grounded followup champion and draft a grounded followup for that. Every LLM step is traced so for that. Every LLM step is traced so for that. Every LLM step is traced so that we can evaluate quality and improve that we can evaluate quality and improve that we can evaluate quality and improve over time. over time. over time. The third layer is what turns this The third layer is what turns this The third layer is what turns this automation into a system that automation into a system that automation into a system that self-improves. Every action is a self-improves. Every action is a self-improves. Every action is a decision log and every outcome threads decision log and every outcome threads decision log and every outcome threads back to the decision that caused it. So back to the decision that caused it. So back to the decision that caused it. So the naive version of this is a data the naive version of this is a data the naive version of this is a data analyst coming in and trying to analyst coming in and trying to analyst coming in and trying to understand if the output of this could understand if the output of this could understand if the output of this could be better. The rebuilt version of this be better. The rebuilt version of this be better. The rebuilt version of this is wiring our engagement history back is wiring our engagement history back is wiring our engagement history back into the decision layer so that the into the decision layer so that the into the decision layer so that the system decides whether or not to system decides whether or not to system decides whether or not to continue a thread, advance to the next continue a thread, advance to the next continue a thread, advance to the next step or pivot. The system will continue step or pivot. The system will continue step or pivot. The system will continue to do that with the life cycle message to do that with the life cycle message to do that with the life cycle message performance history as well. So these

performance history as well. So these performance history as well. So these verification loops are really critical verification loops are really critical verification loops are really critical so that the system can self-heal and so that the system can self-heal and so that the system can self-heal and continuously improve. continuously improve. continuously improve. Let's see how an agent and human work Let's see how an agent and human work Let's see how an agent and human work together in this shared customer view. together in this shared customer view. together in this shared customer view. In the customer view, a rep can come In the customer view, a rep can come In the customer view, a rep can come here and see the product usage, the here and see the product usage, the here and see the product usage, the recent activity and get an answer using recent activity and get an answer using recent activity and get an answer using that same data. They used to find all of that same data. They used to find all of that same data. They used to find all of this across many different tabs and now this across many different tabs and now this across many different tabs and now they can just come here each day. The they can just come here each day. The they can just come here each day. The rep can ask an agent and the agent will rep can ask an agent and the agent will rep can ask an agent and the agent will reply uh querying our context layer. We reply uh querying our context layer. We reply uh querying our context layer. We can also use notion custom agents which can also use notion custom agents which can also use notion custom agents which are sharable across companies to access are sharable across companies to access are sharable across companies to access this data context for recurring this data context for recurring this data context for recurring automated workflows. automated workflows. automated workflows. In the task view, a rep starts their day In the task view, a rep starts their day In the task view, a rep starts their day with an already prioritized task box and with an already prioritized task box and with an already prioritized task box and they already know how to move forward they already know how to move forward they already know how to move forward with accounts and contacts. And the with accounts and contacts. And the with accounts and contacts. And the email draft for an outreach task is email draft for an outreach task is email draft for an outreach task is already pre-ressearched and available already pre-ressearched and available already pre-ressearched and available for them to review. The human is still for them to review. The human is still for them to review. The human is still in the loop and actually adds their own in the loop and actually adds their own in the loop and actually adds their own judgment and taste um and sales secret judgment and taste um and sales secret judgment and taste um and sales secret sauce, but they're no longer starting sauce, but they're no longer starting sauce, but they're no longer starting from a blank sta slate. And this does from a blank sta slate. And this does from a blank sta slate. And this does more than one help one rep be more than one help one rep be more than one help one rep be productive. Um our goal is actually to productive. Um our goal is actually to productive. Um our goal is actually to raise the floor for the entire team. So raise the floor for the entire team. So raise the floor for the entire team. So a Neil sales rep coming in, they can a Neil sales rep coming in, they can a Neil sales rep coming in, they can learn the notion sales process, learn the notion sales process, learn the notion sales process, understand what signals are important to understand what signals are important to understand what signals are important to look for, um, understand which playbooks look for, um, understand which playbooks look for, um, understand which playbooks are effective, and basically know what

are effective, and basically know what are effective, and basically know what good followup looks like. Reps who can good followup looks like. Reps who can good followup looks like. Reps who can are ramping can still learn from the are ramping can still learn from the are ramping can still learn from the patterns of the strongest reps without patterns of the strongest reps without patterns of the strongest reps without needing every lesson to be passed down needing every lesson to be passed down needing every lesson to be passed down manually. manually. manually. So, one of the questions that we came So, one of the questions that we came So, one of the questions that we came across along every step of the way is a across along every step of the way is a across along every step of the way is a classic question. Do we build or buy? classic question. Do we build or buy? classic question. Do we build or buy? And it's very tempting and trendy to say And it's very tempting and trendy to say And it's very tempting and trendy to say build everything. But what we found is build everything. But what we found is build everything. But what we found is that there are still key areas to build that there are still key areas to build that there are still key areas to build and rent access to at every single and rent access to at every single and rent access to at every single layer. Internal agents are actually layer. Internal agents are actually layer. Internal agents are actually cheaper and faster to build than most cheaper and faster to build than most cheaper and faster to build than most people assume. And so since we have the people assume. And so since we have the people assume. And so since we have the most data on our con on our data model most data on our con on our data model most data on our con on our data model um we build it there first and then we um we build it there first and then we um we build it there first and then we uh rented the generalizable parts later. uh rented the generalizable parts later. uh rented the generalizable parts later. So for us the build versus buy as a per So for us the build versus buy as a per So for us the build versus buy as a per layer decision. We will not build a lot layer decision. We will not build a lot layer decision. We will not build a lot of these tools like orchestration, of these tools like orchestration, of these tools like orchestration, email, CRM. Um vendors do that really email, CRM. Um vendors do that really email, CRM. Um vendors do that really well. We refuse to outsource the context well. We refuse to outsource the context well. We refuse to outsource the context layer because that's where our edge is.

layer because that's where our edge is. layer because that's where our edge is. a generic tool can't capture all of our a generic tool can't capture all of our a generic tool can't capture all of our esoteric data models or workflows and we esoteric data models or workflows and we esoteric data models or workflows and we do not want that context layer to be do not want that context layer to be do not want that context layer to be something we can't um debug something we can't um debug something we can't um debug and so as I mentioned before that and so as I mentioned before that and so as I mentioned before that context layer is a notion it's built off context layer is a notion it's built off context layer is a notion it's built off of plain markdown a language that agents of plain markdown a language that agents of plain markdown a language that agents are fluent in and we have databases and are fluent in and we have databases and are fluent in and we have databases and hierarchies that they can navigate hierarchies that they can navigate hierarchies that they can navigate easily at the same time this is well easily at the same time this is well easily at the same time this is well designed for human um so this is what designed for human um so this is what designed for human um so this is what lets our engineers, agents, and GTM work lets our engineers, agents, and GTM work lets our engineers, agents, and GTM work off the same context. And this has all off the same context. And this has all off the same context. And this has all the data synced across sources. Ultimately, the reason to see if we can Ultimately, the reason to see if we can do all of this is to see if we could get do all of this is to see if we could get do all of this is to see if we could get a better throughput on deals. It's very a better throughput on deals. It's very a better throughput on deals. It's very early for us. We're still building this early for us. We're still building this early for us. We're still building this out, but the initial signs are out, but the initial signs are out, but the initial signs are promising. In the last 13 weeks, we are promising. In the last 13 weeks, we are promising. In the last 13 weeks, we are already seeing enterprise reps have already seeing enterprise reps have already seeing enterprise reps have increased qualification or qualified increased qualification or qualified increased qualification or qualified opportunities. And on the life cycle opportunities. And on the life cycle opportunities. And on the life cycle marketing side, users who received marketing side, users who received marketing side, users who received contextaware recommendations were 63% contextaware recommendations were 63% contextaware recommendations were 63% more likely to take the next step. This more likely to take the next step. This more likely to take the next step. This is the early days with a lot more is the early days with a lot more is the early days with a lot more features we want to build, but our features we want to build, but our features we want to build, but our thesis that us building a single system thesis that us building a single system thesis that us building a single system on no, decide, act, and learn seems to on no, decide, act, and learn seems to on no, decide, act, and learn seems to be right so far.

be right so far. be right so far. A few key takeaways. um from entering A few key takeaways. um from entering A few key takeaways. um from entering entering this world as an engineer in entering this world as an engineer in entering this world as an engineer in the last six months is that before you the last six months is that before you the last six months is that before you build shadow your best human. I talked build shadow your best human. I talked build shadow your best human. I talked to many sales reps and when I opened uh to many sales reps and when I opened uh to many sales reps and when I opened uh when they opened their computers I saw when they opened their computers I saw when they opened their computers I saw how many tabs and tools they were how many tabs and tools they were how many tabs and tools they were navigating between and that was a chaos navigating between and that was a chaos navigating between and that was a chaos but it was also the spec and so if you but it was also the spec and so if you but it was also the spec and so if you encode a mediocre process you get a encode a mediocre process you get a encode a mediocre process you get a mediocre agent. Start with the most mediocre agent. Start with the most mediocre agent. Start with the most legible workflow. That's the one that's legible workflow. That's the one that's legible workflow. That's the one that's documented and repeated. And let humans documented and repeated. And let humans documented and repeated. And let humans stay in the loop on where there are stay in the loop on where there are stay in the loop on where there are risky possibilities. risky possibilities. risky possibilities. Model GTM as primitives, entities, Model GTM as primitives, entities, Model GTM as primitives, entities, context, triggers, actions, eligibility context, triggers, actions, eligibility context, triggers, actions, eligibility rules, and the alien world becomes a rules, and the alien world becomes a rules, and the alien world becomes a system you can engineer. Last but not system you can engineer. Last but not system you can engineer. Last but not least, be headless by default and design least, be headless by default and design least, be headless by default and design for agents as operators and not just for agents as operators and not just for agents as operators and not just co-pilots. If humans and agents can't co-pilots. If humans and agents can't co-pilots. If humans and agents can't read from the same substrate, you're read from the same substrate, you're read from the same substrate, you're basically building two systems that will basically building two systems that will basically building two systems that will eventually drift apart. For us, that eventually drift apart. For us, that eventually drift apart. For us, that layer is notion.

layer is notion. layer is notion. Right now, humans are the primary Right now, humans are the primary Right now, humans are the primary consumer of GTM data and agents are consumer of GTM data and agents are consumer of GTM data and agents are helping at the edges. Soon, agents will helping at the edges. Soon, agents will helping at the edges. Soon, agents will become primary first class consumers become primary first class consumers become primary first class consumers within the system, moving from drafting within the system, moving from drafting within the system, moving from drafting to acting within guard rails by creating to acting within guard rails by creating to acting within guard rails by creating the best context and substrate for the best context and substrate for the best context and substrate for humans and agents to collaborate humans and agents to collaborate humans and agents to collaborate together. Now, you're setting up your together. Now, you're setting up your together. Now, you're setting up your team to sprint faster. team to sprint faster. team to sprint faster. Um, yeah, and feel free to contact me if Um, yeah, and feel free to contact me if Um, yeah, and feel free to contact me if you guys want to ask more questions.