How AI Transformed a Sales Crisis in 7 Days

Speaker A: Welcome everybody. This is Ryan Staley and I am back with Mike Haylen for part two. Mike is the GM M of AI Studio over at Asana. Was a lifer, I guess I could say that, in enterprise sales. Enterprise sales leadership and now is leading up AI Studio for Asana. Mike dropped some amazing insights on how they used AI transformation internally. So if you missed episode one, go back and check it out. But now what we're going to get into more specifically is not the internal component of what Asana is doing, but more the external component.

Speaker B: Right.

Speaker A: What AI Studio is and how it's serving other people. Mike, welcome. Happy to have you back on the show.

Speaker B: Thanks Brian. Great to be here.

Speaker A: Yeah, I'm uh, pumped. I want to get into this. I'm curious. I love checking out an awesome AI tool, AI agent, workflow, whatnot. Super intrigued with what you're doing over there and love the thought leadership coming out of Asana. So walk us through exactly what AI Studio is and like how it could be applied for go to market.

Speaker B: Yeah, probably best if I uh, take you through a couple slides.

Speaker A: Yeah man. Show us some visuals. People love visuals, especially my YouTubers.

Speaker B: We can do this together. So AI Studio is a no code workflow builder. If you're experienced with Asana, you'll appreciate how intuitive the application uh, is that extends to this product as well. If you've used our uh, rules engine before too to create automations with if then statements that can help perform actions that you would otherwise have to intervene and do on your own, you'll be very familiar with AI Studios. We've now just embedded AI to perform lots of these functions right within that uh, workflow builder. Any workflow really has these four main processes. It's intake, planning, execution and reporting. And, and so you could imagine for an event that you're running Ryan, or the work that you go do and doing discovery with your clients and having to intake inputs from surveys that you're launching, you can really. The manual process you would have had to go through to go back and forth and make sure you collected all the information. There was a QA check that all the fields got filled out, that you had follow up and ask specific questions about things uh, that didn't get filled out or weren't, didn't have the context you required. You can automate this all now and remove yourself from it until the point in which you would need to intervene. Which now instead of all those steps might be at the point of, you know, planning out the work off the back of all that information that you collect, uh, have prompts embedded within this workflow that could produce the output that you're ultimately sharing back with customers about what you've learned without all the copying and pasting between different systems and intervention that you would have had to take prior to communicate or email back and forth to collect this information. And so if any workflow that you built, how this extends to go to market, you know, applies really in the same way. We have a large company running their events, uh, large scale event out of uh, this uh, workflow engine now where they had to deal with hordes of vendors and you know, people and speakers and, and now they've been able to automate a lot of this process up front to collect the information they need to communicate back, uh, out to customers, um, whatever uh, they need from them, and then to put their team in a more strategic position to actually focus on the execution of the event as a result. And so really these are steps along the way that we were just speaking to that any one person would have to perform. And now you can see really where AI can start to take shape. Renaming requests, automatically, doing a quality check of the information that's actually in there, communicating back with the requester submitter with custom questions based on what hasn't been filled out, checking for duplicates and as you get into planning phases, maybe reading documents that you've embedded in there to only pull out what uh, you need from it to uh, input into the account plan that you're using, uh, to recommend you know, uh, due dates based on good examples that you've shared within the prompt and the workflow that you've built and really start to take this all the way through to really more effective reporting on it because you've eliminated a lot of the duplicates without all the manual effort that that required. You've renamed inputs to, for a more consistent structure that you require so your data is a lot richer and cleaner and you're analyzing documentation that doesn't require, you know, people to have to sift through all of that, but instead really just uh, double uh check the work of you know, uh, the AI that you've embedded with it to say hey actually ah, I approve this. This looks good. This is, this did yield, you know, what I expected it to, and that will take you all the way through to um, you know, a smart workflow. And so you know, I think Kasana's collaborative work management platform lends itself really nicely to this because you can think of it as a directions post Google Maps and pre, with our uh, work graph that you see here, it gives the AI the context and structure it needs to go to be pointed to a specific place where it can go pull that information as opposed to, I think these agents that get deployed into a large database and get asked to perform a function, uh, where there might be different versions of that document or um, old versions of that document, or even the lack of context or structure to point to a particular part of that large database. That's where hallucinations occur and it becomes much more difficult to perform these. And the workflow builder really can take you through this uh, in a much more seamless way. And so these are starting to get into screenshots of exactly how you'd go step by step where you can initiate a rule. When this form comes in, move it to this section. As that form gets read qa, check it. If it doesn't have an information, enough information, put it into this section. If it does, show it as complete and continue to take it all the way through to workflow. And then the prompts that you're writing are where, how you're going to instruct AI to perform functions within that. So you might have a document it needs to read. This is what a great blog post looks like. So that it has good examples and good context of what it needs to produce and all of that can be automated within this step by step instructions that you've helped give it to navigate as opposed to. You know, I think a lot of where the discussion goes today is just take an agent, let it run wild, it'll do all the work for you. We are far from uh, you know, that reality and I think uh, context, human in the loop and a little bit more structure we found lens it really sell really nicely to taking advantage of AI in the world that where we sit.

Speaker A: Excellent man. So what, what models do you support as well? I see the model builder in there.

Speaker B: Yeah. Uh, so we are very intentional about the models we select. We partner of course with OpenAI and Anthropic. Our CEOs on the board of Anthropic. We have two high reasoning models, one from each and then uh, two really low cost models that help you perform, let's say high volume, lower reasoning tasks, uh, that are less expensive and then a couple that sit right in between that. I think as we go forward we'll continue to look at supporting more and more models, but those are the two that we support today.

Speaker A: Okay, so this is good man. I love this. I know we're short on Time, because I've kept you a little bit longer with uh, some of the stuff that you're doing and focusing on what are you seeing as like you're. Let me ask you this, is there any like running while um, you're asleep? Like this is a good analysis where you wake up in the morning and a ton of different task analysis optimizations have happened and it's done by the time you wake up or it's happening in the background. Any great use cases like that that you could point to and what you're seeing from that as well?

Speaker B: Yeah, when I stepped into this role we had, let's just say many, many dozens, Uh, I won't give an exact number but of opportunities that uh, and I had two people on a team and I report directly to our global CRO. And in late February he was asking me about individual opportunities and my answer was, I've got a waterfall here to report my forecast. But you know, having details about each and every one of these is difficult at this stage. And he said you days. And in those seven days we used AI Studio, we built a workflow that integrated in with Salesforce and uh, amalgamated all this data from call transcripts, from email, back and forth, from notes and account plans between Asana and Salesforce and put it into a simple interface and dashboard within Asana where I could. I wrote a prompt to say, here's what a good deal looks like that's likely to close. Here's what a not so good deal looks like and give me your interpretation of whether this is going to close this quarter or not. And then give me a justification for why. And uh, now I've got uh, in seven days with the help of a, of a builder, a creator here who's you know, really adept at integrations and building workflows because there still is a little bit of a learning curve. Today we were able to build this really powerful workflow engine that created an interface for me that uh, gave me an in out call on every deal, it gave me the justification for each of those. And all of a sudden within seven days I sounded really, really intelligent. I think the uh, alternative to that would have been having to go out to each of these individual teams, bothering them about where things sit going and searching that information myself in the many, many hours of the time that that would have required prior. And I was able to more confidently call a forecast as a result in only a week's time and then do so with really great detail about, you know, why these deals were trending in the right direction or not, that would have been impossible prior. Well.

Speaker A: And how many deals was it that you had to do this with?

Speaker B: Many, many dozens. I'll say. It's, uh, you know, we've had a huge demand for this. More than you could any one person have to spend time learning about in a week's time, put it that way.

Speaker A: Excellent, man. Well, we got, uh, I got one last question for you before we wrap. Where, like, what do you see as the future of where this is going specifically as it relates to agents? Because there's I think real agents that are out and then there's branding agents that, where people talk about agents that aren't really agents. So where do you see the future of AI over the next 12 months and specifically as it relates to go to market or just business as a whole?

Speaker B: Yeah, I mean, you know, there's some bias here, but I think if you go out and talk with as many CIOs as I've been fortunate to, this is rooted in reality too. And that is that the cost of value for agents today is really just not there. The reason for that is what we talked about a little earlier, which is agents really require still direction, good instruction, context, and that's what really great prompts help to do. Uh, if you've ever tried to just pose a simple question without that In a, a ChatGPT or Claude, you see what you get back versus if you give it context and you know, a task to perform, goals and then examples of what good or bad looks like. And I think that's what, you know, uh, our platform really lends itself really nicely to. It has the structure of the work graph so we can point you to specific tasks or projects as opposed to the entire system. It has the context that it can give you from within that workflow. So the task description or documents that are attached or commentary, that's happened back and forth and where agents can't perform that function, you could put humans into the loop. I think this is, you know, I was just at a panel last week talking about this at a marketing conference and you know, there was talk about these SDR bots and you know, enterprise search. And the unfortunate reality is today, and this is why CIOs are saying the cost of value isn't there is when you deploy an agent out into the wild without those step by step instructions, lots of hallucinations occur. And you then, as you and I talked about in part one of the episode, this is where people start to get deterred about the value that AI is going to deliver them, and then they are even less inclined to go through the transformation that we want them to go through. So, uh, I think while there's a lot of talk about agents, I think what customers are looking for really today are practical ways to get value, uh, out of AI today. And they require. And that's why we've had so much excitement and demand around our workflow builder, because those things that are still integral to get value out of AI are really, are there. And you don't have to spend a lot of time or energy to go find it. It can walk you through and guide. Guide you much like it would the agent along the way.

Speaker A: Great feedback, Mike. And, um, I totally agree with you that that's. You got to give a direction, context, prompting. Those are the three core elements for a good outcome. So, unfortunately, we're up on time again. Where can people find you? Where can they find more about AI? Uh, studio at Asana?

Speaker B: Ah, yeah, asana.com. we've got all the resources, you know, right there for you to learn everything you'd want to know about Asana and AI Studio specifically. And then you can find me on, um, LinkedIn. There's two of us. I am a junior, so, Mike Hal h a y l o n at GM of AI Studio at Asana. It's been great. Ryan, thank you for having me here. I appreciate it, man.

Speaker A: Thanks for being on, man, and appreciate you joining us as well. And we will see you all on the next episode.

Speaker B: All right.