Agents that listen, think and act for you

Julia Nimchinski:
And now we welcome Joel Klinn, Head of Sales at Hook, and Sophie is about to join us as well, right?

Joe Klin:
She should…

Julia Nimchinski:
So it isn’t… Yeah. What’s the latest and greatest, Joe?

Joe Klin:
So much. We actually… we’re hosting, we host an award show in London tonight for customer… customer leaders, account management leaders, etc. So we’ve got a big award show this evening that we’re all going to, which should be good fun. What about you? What’s the latest and greatest?

Julia Nimchinski:
Amazing! We are, you know, in the forefront of GTM identification, so can’t wait to get into your session.

Joe Klin:
Nice. Nice. Sophie will be joining any second now, but I’m taking the first part of it, so you can just start if that’s better, Julia, it’s up to you.

Julia Nimchinski:
I think she’s here. One second, live TV, and yep. Amazing, she’s rejoining. Welcome to the show! How are you doing, Sophie?

Sophie Cornish:
Not too bad, thanks. Thanks so much for having us.

Julia Nimchinski:
Our pleasure. Let’s get into it.

Joe Klin:
Awesome. So, hello everyone, I’m Joe, head of our sales team at Hook. Sophie has just joined. Sophie, you might want to do a quick intro?

Sophie Cornish:
Yeah, hi everybody, Sophie, Account Executive here at Hook.

Joe Klin:
Awesome. And I think, Julie, everyone can see the screen okay? Yeah, great. So, I think just a quick place to start.
The topic of the conversation for us today is really around, I guess a trend we’re seeing in the market, and probably everyone on the call is going through, which is systems of record, like your typical CRM, the value base when we think about AI products is now moving quite sharply to systems of action.
We’re going to talk a little bit about how we think about that at Hook, but also show you some of the different ways that we’re seeing that play out.
And as a quick intro, I think, as well, just to set the scene for everybody as to what Hook is, we’re, a CS platform, and we leverage agents to basically listen, think, and act on behalf of your CSM, and CS leadership teams.
We do this in a couple of ways, so we ingest a huge amount of data, both structured and unstructured, to build predictive health scores, using machine learning. And then we combine that with agents that are basically looking across your data, your books of business, your customer calls, your emails, to spot for risk, to spot for opportunity.
And then to take action on that, right? And to take action on that at scale.
And I think that nicely leads into the next slide, which is around… the… I guess the trend that we’re seeing in a lot of the conversations we have at the moment, which are… systems of record, be that your CRM, and even systems of intelligence, are just becoming, embarrassingly easy to build, almost, right?
Like, anybody can go and put together some sort of claud agent that can look at a Gong call, figure out where the risk might be in that account, or help draft an email off the back of a call for the follow-up to the prospect or the customer.
We’re seeing so much of our customer base starting to, like, interact with AI tools, interact with agents, build things like that themselves, which is really exciting. And it’s becoming really easy for people to do, which, again, is having a huge impact on the companies that we work with.
And probably everyone on the call by now is… I mean, you’ve probably done this yourself, but you’re seeing across SaaS a huge proliferation of all of these businesses that are now adding AI agents on top of platforms, and it’s just becoming more and more ever-present across every conversation that you have.
And what we are seeing at Hook, and what we think is an interesting trend, and the bit that… where the value in this new AI world is sort of landing, is the systems of record are easy, the systems of intelligence, spotting risks, spotting opportunities, becoming easy, but then how do you… how do you then make that a system of execution?
The hard part’s finding what to do after you’ve spotted the risk. So, how do you then take a risk, decide what actions need to to be taken, go and take those actions, assign the person that should be taking those actions.
Maybe they automate that and do that themselves, maybe we have an agent doing that end-to-end, but that system of execution is where the value is really driven in this new world that we’re living in with AI, and that’s where we’re seeing a lot of our customers spend a lot of time with us, which is, let’s find the risk, that’s easy, but then how do we turn that into a system of execution that can be built out into The workflows across your… across your business.
And so, I think also an important thing to state before I hand over to Sophie in a couple minutes is, we have a really strong point of view about what an agent is at Hook, so… It’s not something that will listen to a call and follow up. and write a follow-up email for you.
It’s not something that will listen to a call and maybe spot risks, but it’s something that, end-to-end, you’d pay a human to do in your business if there wasn’t an agent doing it.
And so, think about, like, an end-to-end renewal outcome, from the risk being spotted, or 6 months out from renewal, all the way to that customer signed the renewal paperwork and we’re moving forward. Like, having an end-to-end, outcome-driven agent is how we think and define an agent in Hook.
And that’s all of the way that we are packaging and talking about the agents that we take to market as well. So it’s about… it’s about driving outcomes, and systems of execution get us there.
I think the one thing I’ll leave you with before I hand over to Sophie to talk through a little bit about what we do, how we think about this a bit more tactically, is… Any AI really can tell you that a customer is at risk, or there’s an opportunity in your customer base.
what we do, and what the systems of execution that we’re working with, and partnering with are doing, is we’re… we’re determining the risk, but then we’re taking, like, what is the next best step? What should we be doing that’s worked before?
And then, once we’ve done that, we’ve triggered the outreach, maybe we’ve assigned an owner to do that work, or we’ve had an agent do that work for the CSM, or the AM, how do you then track the response to understand what worked, what metrics were improved, and how do you tie that all the way back into a renewal outcome?
And then, how do you scale that across your business so that ex-CSM over here, that hasn’t had any interaction with this team that’s done this work, understands what good looks like, and keep training the system to understand how to best drive those outcomes.
So that’s what we’re… that’s what we’re spending a lot of time focusing on, how you take those risks you spot, those systems of record, if you like, and how you turn them into execution layers. For businesses that operate within your workflows and help you drive to the outcomes that you want to drive to.
And with that, I’ll pass over to Sophie, and Sophie, I think it makes sense for me to just click through for you.

Sophie Cornish:
If you could, thanks so much. And hi everyone, thanks very much for having me. I guess to introduce the more practical side of Hook, and really how we think about this, really what we do is use AI to mimic what a human would do if they were going about their day-to-day.
And when you think about that at a really high level, what they’re trying to do is understand customer needs, so where might there be risk, where might there be expansion opportunity? They’re using their interpretation to create some sort of plan for that. And then at some point, they’ve actually got to reach out and execute on the plan.
It sounds so simple, but, in reality, when you’re gathering intel, it is in a ton of different places.
And I think for this room, as kind of a leadership view, it is so hard to get a roll-up view of what are my leading indicators, where do I need to be focused, what are the asks I need to be making internally, whether that be of product, support, exec. Creating a plan, very similar, relying on gut feel to interpret data.
No human can read through every single support ticket, look at every single product usage metric, review every single email across their book. And then taking action is maybe a little bit easier, but… Really hard to tie that back to impact, so to work out, okay.
sent this thing over here, how did it actually relate to the intel that I had to begin with? And then how do you start to scale it out to say, okay, this works in this type of customer, this is our new playbook for this type of scenario? And that’s really what we try and solve for.
So we’re using AI each of these steps to gather intel, understand customers, create a plan, and then to actually generate action for either reps to take, or when we’re looking at a fully agentic approach, to run automatically in the background for a more digital touch. Next slide, please, Jay.
And the way that we do that is we take a ton of different data. So we take your structured data, which is things like usage, adoption, how often you’re meeting with them, what their spend is, really any hard structured data that you could see in a table.
Your unstructured data, which is that context layer, so what’s going on in emails, what’s going on in calls, what’s going on in support tickets. As well as external data. So we buy third-party data from the likes of LinkedIn.
to say, who are we speaking to in the account, where do we have coverage, where do we not, and where are some of the gaps that we need to go and plug? As well as what’s going on in the news.
So, has one of your customers been through M&A activity, and actually that presents massive expansion opportunity, or have they been through, big rifts, or they’ve had a really bad earnings call? And actually, that’s going to create risks in accounts. And we use all of that information to do two things.
We use machine learning on the structured data to build out bespoke machine learning models that tell you, okay, here’s my churn risk, here’s my expansion opportunity, to 85% accuracy 6 months out from renewal. And then we use AI on all of the data to spot for risk and to spot for opportunity.
And probably the best way to think about it is a little bit like a rep that’s always on, kind of always listening to everything that’s going on across the book. Next slide, please. And so if you go into core tenets of the rep role, again, we’re gathering intel, we’re creating a plan, we’re taking action.
Really what Hooks AI is doing is, using AI and machine learning to listen across the book, detect risks, detect opportunity. Then we’re going a step further, and we’re generating what we call play iBooks, which figure out, okay, what is the risk here? What is the thing that solved that risk in the past, or what is the expansion opportunity?
What’s the thing that’s closed one that expansion opportunity in the past? And then automatically puts out a 3-5 step campaign, cadence, or advice for the rep to take. And then we integrate back into your typical workflow tools, so things like emails, Slacks, Teams, Salesforce.
However your team are working day-to-day, they have hook as the center of their operations. All of their content drafted for them, and then pushing out versus via the various, kind of, systems… systems and tools. That’s the human-in-the-loop touch.
We can also automate that full journey, end-to-end, so that you don’t actually need reps to manage those workflows. I’ll show you what that looks like in the demo environment, unless, Julia, do we have questions now, or should we jump in and demo?

Julia Nimchinski:
We can address one question that came in, and the question is, what are the signals which determine the risk? Is it product usage declining?

Sophie Cornish:
So the signals are triggered by anything that has been created a risk before, so created a material impact on the renewal likelihood of the customer before.
So that could be any combination of, hey, product usage has declined, and actually your previous champion hasn’t logged in in 3 days, and they’ve been through, I don’t know, M&A activity, and all of their leadership team has left. It could be any combination of those things.

Julia Nimchinski:
Perfect. What can you say?

Sophie Cornish:
Amazing. Right, I’ll jump into the demo. Checking, you’ve now got my demo environment, what looks like Hook. Amazing. Cool. So, I guess to bring this to light a little bit, I’m in a demo environment at the moment.
This is Hook, and I’m in a single account page, and really what I’m going to show you today is how we take an account, we ingest all of the data, we spot for risk, spot for opportunity, and then Take the next step for the reps.
I’ll start by showing, kind of, human-in-the-loop view, so that, you can look at, I guess, the mechanics of how things are working behind the scenes, and then I’ll move into showing you what this looks like in an automated flow, what is the next gen, and how are our most progressive leaders thinking about, the future of customer success?
And so, looking… logging in and looking at ACL, you can see everything that you need to see about the account.
But the main areas of risk or opportunity that we’re scanning for are the key metrics, so usage, adoption, how often you’re meeting with them, what trends are we seeing that might signify risk or opportunity, as well as unstructured data, so what’s going on in meetings, what’s going on in emails, what’s going on in support tickets, what’s going on in the news.
And in ACOR, we can see that Hook’s Echo, RAI, has detected two signals to review. We’re gonna provide an exec summary. So, in this instance, hey, there’s been low workflow completion and rising ticket volume, which is indicating process friction, potential of risk, potential risk of partial churn, or reduced platform reliance.
We’re going to cite the source. I think any time you introduce any degree of AI into an organization, it’s so important that people feel like they have control and they can go in and validate. But we’re going to go a step further, and we’re actually going to look back over everything that’s worked in the past to resolve similar issues.
And generate the next best actions for the teams to take. So in this instance, they need to run some workflow optimization workshops with the key stakeholders, they need to deliver some more training to the team, and they need to deploy, some dashboards.
We’re gonna draft the copy for the team, so that they’re not starting from a standing start, using, your company’s tone of voice, and making sure that it’s appropriate to, making sure that it’s appropriate to the way that you would, or a human, would manage customers.
I’ve shown you that for churn, but we do exactly the same thing for, expansion opportunity, so strong dashboard adoption and increased user logins indicate expansion readiness, opportunity to expand into new modules. Again, I’ve shown you this in a single account view. In practice, where most people live is in this Actions tab.
Which is an aggregation of everything that Hook’s agents have surfaced, for reps to come through and review. All the copy is drafted for the team with the right stakeholders looped in. Say you want to also, I don’t know, CC your manager, you can do that here.
And then when you’re happy with it, you can just click through and say, hey, send and complete, sending complete, send and complete. The journey we see most of our customers go on is they want to start with human in the loop. They want that control, and it also gives opportunity for refinement.
When we get to 90-95% acceptance rate, so no edits on, this copy and content plans. People start to move to fully automate the flows, to remove the human in the loop. Typically, they’ll start with their, kind of, lowest spend customers, and then move that threshold higher and higher, in terms of ARR as they get more confident.
What this looks like in the back end, and this is the automations of Hook, is, hey, if Hook spots something, then activate an agent to, activate an agent to, take action on my behalf.
All of these agents are configured in Hook’s backend, so we have ECHO, which detects risk and opportunity, but… We also have Activator, which is an onboarding agent, which fully automates the end-to-end onboarding process without any human interaction whatsoever.
And the way that we think about prompting these agents is, hey, we’re going to give the agent a goal, we’re going to say, meet this milestone. We’re going to feed it with all the resources it needs to get there. So, which metrics does it need to look at? Which benchmarks does it need to hit?
What support documents or help center articles can it lean on to, get the right context about your customers? And then set the goals that it needs to hit throughout the onboarding process, and what each metric means.
So in this instance, hey, you’re an activation agent, your mission is to guide newly onboarded accounts through their first meaningful product interactions, help them reach value as quickly as possible. Get the metrics to look at how they’re using the platform today, how long they’ve been active.
get the customer conversations, important when you’re coming in from, the pre-sales to post-sales handover, and understanding, okay, what is the context of this customer? What journey have they been on before they’ve even entered implementation? And then assess the activation stages. So, hey, how long have they been live?
When have the actions been completed, and what’s the engagement history? give the AI some guardrails around how it wants to speak to customers. So, what tone of voice do we want to use? What level of intervention do we want to offer for different tiers?
For example, don’t offer exact escalations for customers spending less than I don’t know, $10, for example. And then based on your assessment, take the appropriate action.
And what Hook’s agents are gonna do is act totally on behalf of the… of the customer until they… until they get to that goal, or on behalf of the CSM, until they… until they get to that goal. I think that was everything from a… from a demo perspective. I guess, Joe, anything you’d add there?

Joe Klin:
Nope, I think we’re good.

Julia Nimchinski:
Amazing presentation. Thank you, Sophie and Joe. And let’s address the questions from the audience. One of them is, what retention signals turn out to be less predictive than expected?

Sophie Cornish:
Oh, I can take this one. So I, I actually come from a post-sales background, so I was VP of Customer Success at my previous business.
Did the series, B2D journey with them, and… Found that as my… customer growth grew, I had way more data, but actually way, way less insight into where my risk was, where my opportunity might be, and I spent probably the best part of 6 months, trying to build a manual… manual health score.
I think the thing that… was the… interestingly, the least predictive view on where risk might be coming down the track was MPS. And actually, I found an inverse correlation with MPS and, actual churn risk. And I was… I used to lose sleep at going to bed trying to think about it.
And, where I landed on the hypothesis was, yeah, these customers might be angry, and they might be upset with us, but they’re angry and upset because they want us to do better, and they’re giving us feedback because they value the partnership, as opposed to the silent majority in the middle, who we just weren’t speaking about.
And I think that’s… that’s why I fell in love with Hook, because you have that combination of machine learning, which gives you that structured data.
How are end users, how are admins engaging with your product, on top of that qualitative, kind of subjective feedback, where one person might be incredibly angry about something, but actually another person loves that feature, and it’s difficult to… unless you’ve got the marriage of, I think, the qualitative and the quantitative data, it can become really difficult to identify where risk or opportunity might be.

Julia Nimchinski:
Love it. So, Veh, one more question here. How do you measure whether AI interventions actually change renewal outcomes?

Sophie Cornish:
Yeah, great question, and so important, because, if AI is just doing AI for AI’s sake, and not actually driving an outcome, it, It’s nice to have, rather than a business critical. So we track all of these things, so we can see, and actually, if we’ve got time, I can show you what this looks like in the demo environment.

Julia Nimchinski:
Yep.

Sophie Cornish:
And so what this looks like in the demo environment is we have, a customer table, which shows you all of your customers in one view. And actually we can filter by, I don’t know, where AI has taken an action, and then start to compare engagement scores, or actually any metrics across time.
So, say, over the last quarter, we activated this agent, this agent, and these customers have plays associated with them. We’re gonna compare the difference across these metrics. So, hey, what was the metric that we were looking to move? And actually, what difference has it had?
So, in this instance, engagement score over the last 30 days, these are the changes. Where are the differences?

Julia Nimchinski:
Amazing.

Sophie Cornish:
And I think you’re right. It’s important that if you introduce an AI tool, you want to be able to say, hey, we’ve had this amount of impact. Firstly, so that you can justify the cost and the time investment, but secondly, so that you can actually, and most importantly, move the needle on the metrics that matter.

Julia Nimchinski:
This is great. Another question, what languages work well? French or Swedish?

Sophie Cornish:
Oh, great question. So the hook UI is in English, but we can, we can create output in any language, so, Jo, you’ve probably got better visibility into this than me as head of sales, because it’s not one of my customers, but I think we’ve just launched in Mandarin, is that right?

Joe Klin:
Yeah, like, the… to your point, the hook UI is in English, but the AI can interpret, understand, or the hook platform can interpret, understand, and then suggest in the language that the person is using. So, we have customers in Germany, in APAC, so yeah, it’s quite broad.

Julia Nimchinski:
Another question here, folks were asking you to say more on the comment that the human in Google starts, it feels better because it gives, more control to the end users. Are there certain activities that customers feel that are more important to control, or they’re just nervous and, you know, nervous to let control go?

Joe Klin:
I think it’s a really good question. I’ll start here, so if you can jump in if I miss something, but I think the way that we think about this is for the different types of customers you’re serving, and the activities that you’re doing that with.
So, human in the Loop for the companies that we support that have big enterprise customers that want that high touch, being in front of their customers, that value-led relationships, there we see Human in the Loop a lot more, and where agents are helping is pulling together the information, suggesting what to do, and basically taking away a lot of the admin work.
Where we see the full end-to-end driving toward renewal outcomes is, take a Vimeo, one of our customers. Hundreds of thousands of customers, their entire scale segment, how do they manage that?
And so, using different agents that can end-to-end manage that without a human even needing to be in the loop, spotting the risk, understanding what needs to be done 90 days out from renewal, let’s say. The steps that have to be taken, the metrics that need to be driven to for a customer to be successful, is sort of all done on autopilot.
So that’s what I’d say we sort of think about it differently for different customer bases, but Sophie, would you add anything?

Sophie Cornish:
No, I think that’s… that’s totally right, and I think also, I think… people become… people become more confident as they have more… more exposure to things, right? You’re probably not going to go live on day one with fully automated, unless you want to, and you can, but actually just making sure that you are being robust in… in your strategy.
People tend to move to… move away from human in the Loop pretty quickly, though, when they can see how powerful things are.

Julia Nimchinski:
Super helpful, and one more question here. How do you avoid AI-generated customer actions becoming generic at scale?

Joe Klin:
That’s a good question. Every single customer we have has bespoke models built and bespoke agents built for them, and so we’re not just taking into account I guess, like, the prompt that is a good CS prompt to drive to a renewal outcome, but also your internal FAQs, your internal educational content, things like that.
So we have a really, really deep understanding. It’s all about the data that we take from the customers and use to work with that helps us determine what needs to be done in an account. So, because of the amount of data that we’re able to ingest. really helps us be specific to accounts.
That’s not to say that there might be, you know, similar suggestions for similar accounts, because that does… that does happen, but the more data that we have from our customer base enables us to be really focused around what’s worked, what hasn’t, and that nuance that can often be the difference between a renewal and not, that we can… we suggest up to the… Up to the CSM or the customer.

Julia Nimchinski:
Yep.

Sophie Cornish:

Sorry, I was gonna say, I agree with that, and I think… Joe, just to level onto that, the reason AI is so powerful in this context is… it can adapt, and it can ingest, and it does have historical context, and I think the world that I certainly came from looked something like this, which was generic playbooks, had no customization, no flexibility, no context, kind of simple if-this-then-that decision trees.
Which don’t account for changing customer behavior. They are exponentially complicated, and really hard to personalize, and so, actually, I’d say it’s one of the areas that AI has added Exponential value is we can get away from these horrible, kind of, 30,000-step decision trees, where you really have no contextual data to feed into them.

Julia Nimchinski:
Amazing. Joe and Sophie, can we discuss Hook’s uniqueness? Because there are so many solutions now, and it might feel overwhelming, especially for our community. I’m just curious your, you know, favorite customer story, or how, like, how do you see Hook? Why Hook? I know.

Joe Klin:
Do you want to take it safe, or shall I?

Sophie Cornish:
There you go.

Joe Klin:
I think the uniqueness is the system of execution and the ML combined with AI. So I think if you look at, tools on the market, there’s a new YC Combinator company that might pop up, that’s built an agent that’s got a name that is a CS agent. We do that.
But then if you also look at the flip side of the, sort of, systems of record, systems of intelligence, all of the data that you want to be available at the fingertips for your customer success teams. we have that part.
But then the bit that I think is… really ties it all together is we have the machine-learned piece, which is looking at your product usage, your users’ usage, the historical, the historical outcomes that those different metrics and changing metrics have led to in terms of renewal, or opportunity, or churn, or whatever it may be.
And when you combine all of that, as well as external signals, as well as your educational content, and the amount of data that we have to ingest, that’s a really, really powerful, powerful system that there’s not many out there that we see, at least, that can talk to that.
And then our customer stories, we, like, it depends, but at scale, like, we’re now automating Vimeo’s book of business for their scale segment.
So they’ve got hundreds of thousands of customers, thousands of those are running on hook on autopilot, and so, they don’t need teams that they used to need it to be of a certain size, and they’re starting to change the way in which they’re operating and building their orgs out in CS, which are… agent managers that are CS-focused versus teams of individual CSMs that are going to focus on individual accounts.
So there’s loads of stories we can share, and please reach out on LinkedIn or check us out on our website, and happy to talk some more.

Julia Nimchinski:
Sophie, how about yourself?

Sophie Cornish:
Yeah, it’s, I’d obviously agree with Joe, but I think for me… for me, it’s the combination of machine learning and AI. I think… I think most people, or most, most of the companies that I see in the market, skip the machine learning part, because it’s really hard.
It’s really difficult to ingest that volume of data, and to build robust machine learning models. But the thing that makes that special to me is it gives you revenue predictability, it gives you guardrails around your agents to say, hey, go and sense check that you’re not hallucinating against this structured data.
And also as a… I think as a revenue leader, you… live and die by your ability to predict revenue, and to have revenue repeatability. Machine learning gives you the foresight, it gives you structure around your AI, and the AI is a layer on top, which gives you color, it gives you context, and enables you to execute.
And so I think that kind of partnership between the machine learning, which has typically been considered the ugly stepsister of AI, And, and… I guess what we know as AI today, so LLM functionality. is… is incredibly powerful, and to my knowledge, I think. We are the only people in the market doing it at the moment.

Julia Nimchinski:
Awesome, and what’s the next best step for our community? Where should they go? How should they test drive this?

Sophie Cornish:
Jake?

Joe Klin:
hook.co, reach out to me and Sophie on LinkedIn, yeah, happy to continue the conversation.

Julia Nimchinski:
Amazing. Thank you so much.

Joe Klin:
Thank you. Bye-bye.

Sophie Cornish:
Thanks so much for having us. Bye!