Hi Channel Insiders. Welcome back to another episode of Channel Insider Partner POV. I'm joined today by two individuals from Kalent to dive into all things Anthropic AI and what it means for the growing channel ecosystem supporting Anthropic's push into enterprise and mid-market customers. Jason Cutler is the SVP of Anthropic Consulting and Engineering, also known as ACE, the new business unit that Kalent has spun up to further help its customers as they adopt AI workloads and push AI further into their business operations.
And Ryan Gross, now head of Anthropic Engineering in that new business unit at Kalent. You might have heard Ryan's name before if you've listened to the podcast. He joined us last year to talk about AWS migration and modernization efforts that Kalent's also known for. We dive into all things Anthropic AI with me, the two of them, and my co-host for this episode, Grant Harvey, co-host over at The Neuron. It's a great episode not just for those looking for more information on Anthropic, but also how partners are addressing crucial needs across governance, security, adoption, and scaling programs past pilots.
Hope you have a good listen, and thanks as always for trusting us. All right, hello. I am joined today by Jason Cutler, SVP of Anthropic Consulting and Engineering at Kalent, and Grant Harvey, familiar face on the podcast from over at our sister show, The Neuron. Thank you both for joining me. >> Thanks for being here. >> Good to be here. >> Cool. So, I guess maybe Jason will kind of dive right into the work that that Kalent's doing with Anthropic, but before I ask that, maybe we'll back up and ask you to introduce yourself and introduce Kalent to the audience. >> Yeah, happy to be here.
Jason Cutler, SVP of Anthropic here at Kalent. I joined Kalent back in 2021 when we were about 50 employees and made a real partnership at that point to go all in with AWS. And so, from 2021 to today, we've grown from 50 to over 1,000 employees. 90% of which are consultants and engineers tied to the cloud and AI. But, our journey is pretty unique in that we started with AWS, and in 2022, after the creation of ChatGPT 4, which lent itself to everyone getting very hyped about GenAI, Amazon released a program called Prompt 100.
And Caelen had to go deliver 40 rag chatbots in 90 days. And I wouldn't wish that upon my worst enemy, but we learned a ton over those 90 days that it actually lent itself to go be the AWS GenAI partner of the year for the last 2 years. >> Wow. >> So, as part of that, we started building great AI products on top of Bedrock with Claude models. And so, we we became an Anthropic partner 2 and 1/2 years ago as we were building out agents and products leveraging these great frontier models.
Now, when Anthropic and Claude then developed out Claude Co-work, it really enabled us as a partner to start diving in deep with Anthropic as well. And so, we attended the partner summit earlier this year. We met with the Anthropic leadership, and we described to them how important it was for us to develop out a practice and and manage that going forward. And so, we announced our practice about a month ago after the partner summit, and we are now investing very heavily into our Anthropic relationship and our AWS relationship at the same time.
So, it's great to have those two companies coming together and surfacing the great AI models to all of our customers, but we are also really, really ingrained at the Anthropic level as they are creating new products and services to the market. And now, we have people dedicated to that piece of the business, too. >> Lot there to dive in on. I know Grant probably has a about a million Anthropic questions, but before we go on the the model side and the tech side, since we are on Channel Insider partner POV, I'm curious, Jason, I know you've said you've worked with Anthropic for about 2 and 1/2 years now, all in.
I think for many partners, Anthropic kind of started being a a channel player earlier this year when they started building out the partner network and and what that looked like from a formalized approach. Could you walk us through maybe what it's been like to get to know Anthropic as a company, how you've developed that relationship with them, and and what that partnership has kind of evolved into to today's practice? >> I mean, traditionally, Anthropic's been a research lab.
And so, creating frontier models or bringing those to customers was what they were great at. And again, our partnership with them really extended to that level. It was going and building great AI products and making sure that we had the best model to fit what the customer outcome was. And so, we were partnering great with their field account teams, with their applied AI specialists, all the different folks that captures that type of work. I think what we've seen in like the last 6 months, though, is there's been this growth of go-to-market and partnerships on the Anthropic side as their products have grown.
And you know, we've seen across LinkedIn and other social media just them explain, you know, how they were kind of caught off guard by how fast they were growing, where we then actually got a dedicated partner person earlier this year. And so, we've seen the increase in head count has actually associated to an actual increase in go-to-market activity with Anthropic because we can now lean in heavier on what are the things that they care about, what are things that our customers care about, and and we can create great solutions to kind of capture those two things. >> Great.
I know you cover Anthropic pretty heavily over on The Neuron. From what you just heard from Jason, do you want to dive in here maybe on on that growth that Anthropic has experienced or whether that surprises you to hear it all or not? >> Uh no, it doesn't surprise me at all just because basically you know, the meme in the industry is something happened over Christmas break, December 2025. Like all of a sudden people realized how powerful the models really were and they actually had time off to play with them and everyone had a project in 2025.
They all came back Claude code pilled and they were extremely bullish on on Anthropic and then Anthropic realized very quickly that a lot of people are using uh Claude code for non-coding related work and so they needed to launch something that to uh meet the users where they were and so they launched uh co-work shortly after that, I think in January and ever since co-work came out and they actually had a product to meet the demand for people who realized how good Claude code really was um you know, for just doing general purpose office work on your computer.
The demand just skyrocketed and it's been the story of the first half of the year um without question in in the media. So, no no surprises there, but I guess I'm wondering, you know, as they were riding this wave, Jason, they've launched a lot of new products. I just mentioned co-work, but they also offered manage agents and a lot of this other stuff. How has it been managing the growth side including the growth of the products that they're offering now and how are you managing that in terms of how you're working with you know, your customers in order to guide people in the right direction of like what product they should use? >> Yeah, we're used to this type of working together with a partner in terms of new things coming out to market.
I mean, AWS has over 240 services, but obviously what Anthropic is doing is at a pace that no one's ever seen. The fact is is that each week we're seeing three major product releases happening that we then have to capture and somehow be able to explain back to customers. So, it was actually one of the reasons why we've created a dedicated practice was because we know customers right now are facing those same concerns that you just mentioned Grant that how do I keep up with something like this?
And that's why you're going to need partners that are keeping up with it, are dedicated to it, and as part of our enablement and training and the things that we're doing for operating models and governance, we can keep iterating on your environment over time because we are keeping up to speed with what's happening within Anthropic and in the models directly. And so, it's really what been one of the main drivers for us in terms of developing out this practice in just customer conversations in the last 3 months, it's one of the main drivers for customers in terms of why they're looking for a partner because to keep up to speed and do your day job is almost impossible. >> It's kind of funny.
My my day job is trying to keep up keep up to speed and it's impossible. So, I can't imagine someone also trying to work while doing, you know, what I'm trying to do. >> Yeah, and I mean Jason to your point on the why customers and why enterprises are are going to keep looking for partners in this space, right? To your kind of expertise, AI isn't necessarily anything new for Kaylenta, but maybe going to market specifically with some of the Anthropic tech is something you've built out in this practice over time.
So, could you walk us through kind of what ACE looks like as a suite of offerings and a consulting practice and how you built that to maybe address what you were seeing in terms of pain points from customers? >> I mean, we've been the J I partner of the year with AWS the last 2 years globally. So, we know AI and we know J I very well. We've been using Anthropic and cloud models for 2 and 1/2 years. So, we know the model situation pretty well. I think what we're seeing now is work changing with AI and how that is it being adapted both across code, co-work, and API.
And so, what we've done is we've kind of internalized that into three different practices. So, the first practice that we created was around agentic SDLC and PDLC. So, the software development and product development life cycle and making sure that we can speak very clearly and articulate the value of Claude code to that subset of user. The next is knowledge worker transformation. And this is where co-work and agents are coming in in terms of how you can leverage those AI solutions to really see and fulfill against how knowledge workers, SG&A functions, business functions can use AI and leverage AI to get work done more efficiently.
And then the last piece is around AI products and agents. And the idea there is how are you creating either external or internal products and agents that can drive innovation faster. Across those three areas of SDLC, PDLC, knowledge worker transformation, AI products and agents, we try and then understand what is your training and enablement look like. How are you helping people inside the organization leverage these tools? How do you know if people who are leveraging these tools are actually doing so thoughtfully, efficiently, or in a way that's actually driving further value to the business?
Beyond that, we then look at the operating model and governance that is required. So, how do you put cost allocations in practice? How do you choose the different models that that are available to you? How do you put things inside of a regulated industry in terms of having the right security guardrails in place? And so, we're helping organizations understand this, and as you can imagine, doing that for a company that's, you know, 80% software development would be completely different than a customer that's 80% knowledge workers.
And so, we're diving in deep trying to understand, again, through our forward deployed motion, what the business requires and what they need and how to do it safely. And then our mindset is by really honing in on enablement and training and the operating model and governance is actually how you get transformation. That's how you can see AI transformation come out of an organization is once you've kind of tackled those two areas concretely, then we feel like we're on a good pace forward.
And for us, it really stems to working with customers. So, we love the idea of forward deployed engineering and product and AI leads and product managers because for us, that's how we've been doing work for a very long time with AWS. We've always talked about doing with customers, not do for. And forward deployed is the epitome of doing with a customer, that you're ingrained with the customer, showcasing, prototyping to them quickly of how this technology can be working in a very safe manner.
And so, we've seen a lot of interest in a short amount of time. And for us, it's also just a massive, massive, massive help to be partnered with Anthropic because they are trying to bring all this technology to those different type of users across SDLC, PDLC, and knowledge workers. >> Grant, I see the wheels turning. >> >> Yeah, well, I had a couple of questions. And one of the things that you said was interesting. So, this was one of the questions that we had.
For companies with mixed users, for example, right? So, you've got business teams in Claude, you've got engineers in Claude Code. We have leaders asking for agents. How are you recommending that they all come together to avoid like three separate AI strategies? Is Is that sort of where you're anchoring the forward deployed engineering that you do, where you kind of come in. And actually, if you could just explain how that actually works. Um, you know, are you building it for them and then teaching them how to do it?
Are you giving them frameworks that they can use? Love to hear more about that. >> yeah. So, you know, I think it's important, you know, you start with training and enablement and what that looks like. And the idea there is are you teaching someone like where the buttons are inside the application or are you training them on how they're going to change their work habits with AI? And so, while we talk about training and enablement, really what we're trying to do is focus on enabling users to make sure they can use AI in a safe manner that can help them accelerate their work.
And so, one of the things that we always kind of strive for these conversations is, "Great. Let's talk about your personas today inside your organization. What are the goals of that persona? What are the challenges of that persona? And what are the different tools that they have today to really get their work done?" And the idea there is you want to be able to then integrate those tools into AI to really solve for the challenges of the role to accomplish the goals.
And so, we're looking at that through MCP, we're looking at through that through custom connectors, right? But we're driving that innovation through a champions network of personas inside the organization to rethink the way that work gets done. And that has been a massive driver, I think, within our organizations and also organizations as a whole who might have just started adopting AI without the why or the how, and they're kind of stuck. Or they've seen, and this happens a lot, right?
You have some users take off, but that's 10% of your total user base and 90% are still stuck here because they're unsure how to prompt. And so, our idea is to understand, great, how can we codify some of the work within the persona to where it's not the person that has to prompt the AI, but the AI keeps the human in the loop so that they are also being included inside of the work itself, and they get more familiar with it. And so, that's kind of how we've been approaching it.
And again, how how you see that across an organization is making sure that you're you you really have this centralized governance and operating model in place to where there is good use. Like, you understand the security guardrails around it. You understand the observability. All the right things that you need from a centralized governance model are kind of included in that area. >> Where are you seeing the strongest results right now? Are you seeing, you know, following that same thread, are you seeing the strongest results bringing Claude into existing SaaS apps, or bringing existing SaaS apps like company tools and data into Claude as plugins or MCPs?
Like, is is there one pattern that beats the the rest, or yeah, what do you think? >> So, I would I'll give you the bad answer. It depends. >> >> Right? Ev- Everyone hates that answer. But, you know, we see it within our customer base who are creating products. They are looking to find ways to leverage AI within their own products to their own customers to enhance that experience. And so, that would fall more to that AI products and agents kind of category that we create in terms of like how we create Claude, or we manage Claude inside of their existing environment itself.
On the other side, you know, there's been this proliferation of SaaS tools across ecosystems internally over the last 10, 15 years. And at this point, one of the major drivers that we see of I would say a lot of enhancement has been taking those solutions, integrating that into AI, and then leveraging the human in the loop to really drive towards the outcome that you need. And so, you know, there's been a massive growth in what we see with Databricks and Snowflake in terms of congregating data, but now it's like, okay, well, what do you do once you have the data?
And so, where does that go? And so, that's where Claude is coming in. We've had a lot of interesting conversations and deployments already in terms of what people are looking to do with co-work and in finance departments, in accounting, in sales, and marketing. And so, while they have these tools, they haven't yet put it around an an AI wrapper to actually get the work done. >> I'm curious, Jason. I know Cailen kind of offers services depending on the type of industry that that your customer's in, and you kind of verticalize around ultimately what your customer needs from a use case perspective.
So, when you think about and the ways you've seen customers be successful with AI, how specific do you think business leaders need to get about what exactly they want out of their AI systems and what that ROI, so to speak, that everyone's always talking about looks like? And have you seen business leaders kind of evolve their ways of thinking from maybe 2 years ago when you really started diving into AI? >> It's a great question. I think one of the things that we always try and talk about are the outcomes, right?
So, what is what is the ideal outcome coming out of this? And for a long time, that's been, you know, cost reduction or new revenue coming in. And for us, what we try and really understand, again, goes back to the individuals inside the organization, and what are the goals of those roles? I think a lot of times, you know, everyone's put together a JD. But does the JD actually describe what that role does day in and day out or just the themes of the role itself?
And the reason why Cloud Code has been the fastest growing products ever was because developers leverage code effectively across an organization and the AI can quickly learn how to supplement that over time and continue to keep the person in the loop. In knowledge worker information today, nothing is codified. And so one of the things that we've talked about is like, how do you codify the roles so that AI understands what has to get done in a meaningful capacity.
And so that is actually one of the main drivers and conversation tracks that we have right now, which is similar to what Cloud Code did with codifying the developer and engineering experience. We have to codify the knowledge worker experience and really understand where human in the loop becomes a critical function for folks over time to do more work and really the the highly thought of work that you really need people to be a part of. >> Yeah, because it's really hard it's it's easy for AI to improve things that are verifiable, right?
Like code is very much verifiable. So it's either the code runs or it doesn't. You know, it works with the code base or it doesn't. But knowledge work it can be a lot more squishy like I guess. And so having documented pro- even just having documented processes for what does each person at this company actually do from day-to-day and then where can you know, AI assist that? I'm sure that's a very tricky problem. I do wonder so in your public facing marketing you mentioned a 66% productivity increase and a 2x faster prototype to production rate.
I guess I was wondering anecdotally, do you have any specific customer success stories you can share that show what use cases specifically are driving that ROI? >> In a lot of cases, you know, we've been developing out these AI products for a very long time. On the code side where we've seen an enhancement, right? Has been around how engineering teams are moving from [snorts] leveraging Claude code as an assistant to a co-pilot to a multi-agent orchestrator.
So, we were working with a healthcare SaaS company where they were able to get Claude code across the organization entirely and they were seeing some they're seeing more code getting generated than ever before. And now when they were going under the hood of like understanding the code itself, it was code it was good code but it wasn't the kind of code that they would have accepted through QA and and other areas. And so, the idea thought process there was well, then how do we get humans involved to make sure that they're in that process of the loop and that they're spending their time wisely of making sure that the code can actually go into production.
And so, we then have taken them on a journey across those different areas to now their engineers are set up as these agent orchestrators that are coming back to them with cues of things to look at where they might have existing problems or where there's actually dependencies on the code in other areas of the application itself that that they have to look at. And so, that is a big piece and something that we've really spent a lot of time on with our customers and kind of going through that journey.
Talked to you another company today even around their QA process and how they should be thinking rethinking the way that Claude code can help them accelerate their QA across the organization and and they're one of the you know, 10 most visited websites in the world. So, for us it's been you know, a journey of understanding kind of again the adoption curve inside these organizations and then where their engineers are spending their time but really where can they be spending it better. >> That makes me wonder in in trying to take that same process of improving the the engineers workflows from assistant to co-pilot to managing like orchestrator, essentially.
If you're trying to apply that same growth pattern to knowledge workers, do you kind of have to teach knowledge workers to think more like engineers? >> Well, maybe. I would never recommend that in a million years. >> >> You're right. Like, yes, you probably would, but it would never work. One of One of the common pieces of information that we get of of why people are even looking for a partner is because they're finding that people even have a hard time going to Claude to set up the prompt inside of the organization, right?
And so, there's this fear of like, "Hey, how do I do this? What can we do?" And so, for us, enablement is that key function to help that subset of users. And it really goes back to understanding kind of the goal, kind of the role itself, what are some of the challenges they face, and then, you know, how could you make their work easier over time based on the different tools that they use. And so, we aren't necessarily asking a knowledge worker to go in and, you know, code up vibe code the map that connects everything together.
But for us, it's about creating that process where we've automated through AI 90% of what the function is, but in the end, it's it actually still needs that 10% review where that person can actually change the outcome of what AI is looking at it with their own perspective tied to it. And so, it becomes more of like this pull-push, which is we want to push our our folks, our our our customers to use AI to help accelerate their adoption and making sure that we're driving those very high-value use cases to where people are not being intimidated, that they're using it freely, and that they're seeing the benefit.
And I think once you do that, you start uncovering even more use cases inside the organization that can continue to drive further enhancement inside the organization itself. >> Hi everyone, Ryan Gross. I lead the engineering side for our Anthropic Consulting and Engineering practice. So, that means helping to shape both how we deliver software delivery life cycle enablement for our clients, as well as how we drive the applied AI side. I heard Jason talking about quite a bit here of how do we actually drive roles to use an agentic approach while not having them need to build software to do that.
And then, um in terms of background, been doing consulting for about 15 years working in data and AI for about a decade now. And over the last 18 months, I've been doing a lot more work on how do you use generative AI and then agentic approaches in order to actually shape the software delivery life cycle. >> I know this this conversation is very heavily Anthropic focused, obviously, because of the consulting and engineering arm you've launched. When you think about the kind of technical capabilities within Anthropic and its models, whether that's specific forms of Claude or Claude Co-worker, or what have you, what stands out to you on the the kind of engineering side of the house, so to speak, in terms of what you've been able to build with Anthropic and then unlock on the customer environment? >> In terms of what we're building with Anthropic, obviously obviously there's multiple levels to that.
So, what we're building from a technical perspective started, like I said, maybe about 18 months ago, where we started down the path of using Claude for modernization work. So, we actually built our own harness and wrapper around Claude LLM models in order to help do things like database transformation, which Victoria, I know we've talked about in past session I did on this uh podcast. And through that process, we kind of learned this agentic loop and what it looks like to really verify the results and how you can delegate more and more work to an LLM.
Over that same time period, Anthropic was really iterating on Claude Code and the kind of new foundations of agentic engineering. And over time, we found that they pretty quickly overtook our capabilities in terms of what our pre, you know, custom-built harnesses did. And so, we've now been working for the better part of a year building on top of Claude Code and how do we configure it to solve those types of problems? And as we've gone through that and done enablement with our teams, we have started to realize that those tools are actually great for enabling other teams as well.
And so, that's some of the genesis of the partnership between Kalent and Anthropic expanding from the more engineering-focused partnership that we've had for a couple of years to the new ACE practice and the ability to do enablement as the primary service where we're working on engagements together with Anthropic. In [snorts] terms of what we've done, like IP-wise internal to the practice, we have harvested back some of the things that we've learned using this for Kalent to be some of that IP that we can go in and work with clients to ensure that their best practices are getting out to all of their team as well. >> Ryan, what do you find is the biggest unlock when you're first working with a new team that's just adopting Claude Code?
We'll we'll focus on engineers for the first part of this. When they're first adopting Claude Code and they're just getting up and running, what's the biggest unlock that you've found where, you know, they couldn't figure it out and then all of a sudden it clicks into place for them? >> There's almost always two waves to it. It starts out with the biggest unlock being getting the right hierarchy so that you have Claude MD and the various other information about your underlying code.
And then, that gets you to a certain point where people, you know, don't have to prompt everything into every message and feel like they're just using it as like smart auto complete and, you know, they're they're burning a lot of tokens. And over time, that gets to a few people who realize that the skill and plug-in ecosystem is the real unlock. And so, those power users end up running out ahead of the rest of the team. Typically, they're doing it, they're working, they're tinkering on their own, and they'll either start to work to distribute that back to the broader team on their own, or you'll get to a point where the team kind of plateaus, where like some people are doing great, some people aren't.
That's our like one of our most common entry points with customers is like "We rolled out cloud code. It's been 2 months. These four people are just all of a sudden so much more productive. Everybody else is a little bit more productive. How do we actually go from those power users back to a fully enabled team?" And that's actually some of where we've focused. So, like we have some tooling that'll let us come in, essentially harvest back what the power users are doing, use that, like literally run through their session histories to truly understand, not just like have them describe it to us, and then build a plug-in in an automated fashion.
Obviously, then we human in the loop and inject our own expertise into it and use that to drive some of the training that we do for the broader team, so that it's not just the concepts in abstract terms, but it's like this is someone on your team that can help you explain this long term. So, you have a champion already. And also, it's in your repo using the MCP connections to your tools and dealing with the limitations that you are going to have to work with.
And that really leads to the broad unlock of many more people kind of getting up to that power user level. >> Yeah, that's awesome. That makes a ton of sense. And then how would you, you know, if you're doing this already, how do you then apply that same kind of system thinking to the code the the you know knowledge work side of the equation? >> Yeah, and this is it's still emerging there like the tooling is much more mature on the cloud code side especially the ecosystem around it.
I think that ultimately underneath they're the same underlying SDK the same underlying models. So like the capabilities that are being built on are about the same. It's all of the other things that have been built up around it already that are more mature on the cloud code side. So what we're seeing there is that same process of harvesting back a plugin based on what someone who is more advanced works in the context of co-work. If you're just talking the chat tool then there's a lot more limitations in there and you need to be a little bit more careful.
You're you're often times just pulling in some of the Anthropic developed or other custom developed skills that will enable it to understand. The the legal domain is one that's actually come up several times recently in that space. And then the other big thing I will say here is the document authoring capabilities that tend to be the core of a lot of your co-work knowledge worker based workflows are not as clear in terms of what does it mean to quality control that document?
What does it mean to even you know validate that the formatting is right cuz you know if code the formatting you have linters and so forth that will automatically do all that stuff for you. There's no such equivalent for Excel spreadsheets for instance. >> Yeah, for for a word document maybe it'd be Grammarly or >> Yeah, for a word doc sure but then you know a lot of times the more important stuff runs through Excel versus Word. >> You're right. Yeah, I guess you'd have to do some sort of like formula check like check.
They should make a linter for spreadsheets. >> Yeah, I mean it that's kind of where it's like it trends towards that over time and if you think about like harvesting out a a plugin from what somebody in you know FP&A has done. That's essentially what you end up with is you end up with this like set of you know scripts that get derived and a bunch of prompts that are being a linter for Excel spreadsheets. >> I have one more question on this. So, like following the AI space on X, you know, you see a ton of cool demos being built with Claude, but what is and you know, Jason Ryan either of you feel free to answer.
What is your strongest recommendation for workflow and or a use case that an enterprise can actually trust in production and apply like right now. Like what's like the first thing you would recommend they do? >> Yeah. >> I think one of the things that we as an organization having done, you know, a ton of agentic use cases dating back to the 40 rag chatbots, one of the things that we understand is AI is a moving target. Right? And you have to create something that is beyond just a point of time.
I think Ryan mentioned it's like we're building for the future and not necessarily the past. And so, one of the things that we drive towards is we've created a framework to think like that. We call it 321. And the idea is that we spend 3 days to do discovery and use case analysis for our customer with the idea that we could prototype something back to them in 2 weeks and bring that into production within a month. And the idea by that point is you are moving fast enough to where the AI cannot kind of skip ahead of you.
Well, Anthropic maybe can because they're creating three new models a week, but besides them, you at least are creating this box of time to create what the use case is, how it gets delivered, people are getting familiar with it, and then it's in production. And the idea is that we are framing 321 where you could have multiple 321s happening inside an organization at the same time thinking of it as like a module, but we want clear intended outcomes from each of them that real people can feel in a meaningful capacity.
And that's that's the main driver that we are trying to align to, And honestly, it's one of the things I think that customers have really accepted to adopt because they've seen that within their own AI use cases that have been either developed or they've been working on. And I mean, we saw that internally here at Kale it for a long time, where if you followed the same agile framework of delivery, you actually might never get to a a point of done because new new feature enhancements continue to come in, new people want to be involved, and so the backlog just continues to grow over time to where there is no done.
3-2-1 meant to give you that that point of time to say, "Hey, we're going to spend 3 days to tell you when done is done, and you can iterate from that point forward, but after a month, this thing is in production and people are using it." >> I think the key thing there, and like if you're thinking about what to select into that type of framework, is something where you're looking to drive a process transformation of some kind. The reason you don't want to continue to thing 100 times after that is like a lot of the work comes in we've prototyped and we've started to work through what the new version of doing this task might look like.
So, pick an example here of, you know, in finance and you've got a long laborious process for doing month-end close. A lot of that's manual steps along the way. Instead, we're going to think about, "Okay, well, here's the set of data sources. This is really just a data processing challenge. We have these old Excel spreadsheets we could bind back to get a good idea of what good looks like or what done looks like." Doing that, that means that the process is going to change and we're ideally going to try to do it in a day or two.
We'll still start early at the beginning. We can prototype out a "If you had all this information, would that allow you to sign off?" And then we iterate several times through that prototype phase to get to like the full set of details that an actual financial analyst is going to need to see in order to feel confident to sign off on something like this cuz it's obvious like it's a pretty critical asset that's being produced. >> Yeah, you do not want any hallucinations. >> Exactly.
And so, but that allows you us to get to a pretty clear definition of what would drive that change and what would actually get like people to use it as opposed to it exists and wow, wasn't that cool, but now I go back through my whole spreadsheet anyway just to double-check what Claude said. And then in that month we can make it real. >> It's like a trust problem as much as it's a technical problem, right? Like you have to help people build a trust that >> Yeah. >> Especially on the knowledge worker side, I think it was a trust problem on the developer side for a while.
Enough people have seen enough news and you know, oftentimes your first week and a half of trying out Claude code for most developers, they get over the hump quickly these days of like, yep, these things check out and do most of what I would expect them to do. We're going through that same cycle. It does feel like a lot of it is like six months behind of like knowledge worker side is six months behind, although the the tooling is accelerating so fast that, you know, it may catch up and not actually be six months between as we go forward. >> Jason, I want to bring you in here on this point that Grant highlighted with the the trust capacity because I think when I talk to service providers and partners who are also kind of building these AI focused practices for customers, what they hear a lot is you know, the executives all in, the leadership teams all in, great, we're going to do all this AI work, and then you get to, you know, Brenda who's been in the office for 50 years and and has everything down lock and doesn't want to learn something new, right?
Or you have the people who are like, I I didn't ask for this, I have to relearn my job now, how do I what's going on? Is part of the kind of enablement focus you've built into ACE helping the entire company kind of get over the hump, so to speak, of going, okay, this is why my leadership is investing in this process? I mean, how often do you see that disconnect from customers? >> We never see that. Uh cuz no one ever Uh yeah. So, No, I mean, I think look, I'm sure when people started using personal computers, the same exact talk track was happening.
I can't redo the way I work today because um it's going to slow me down. I don't think all any of us could imagine a life without a computer at this point. It would be impossible, right? And so, we're going to be going through that same exact piece here with AI over time, and it's it's really upon us to make sure that that person, Brenda, you know, she feels comfortable enough to leverage AI, but that really that AI is there to help accelerate her with her job and make her job easier, and that it's driving more efficiency to let her do more important things like that in the end becomes the driver and gets actually people more open to using more AI because they start seeing all the great advancements it's made in their day-to-day life, and now they want to learn what else they can do.
So, for a lot of organizations that we talk to, you know, we don't necessarily want to go after the most important thing inside the organization because that most important thing might also be probably the most complex thing. And that doesn't necessarily always equate to the best outcome. Where we might want to find time is, "Hey, what's happening inside of your law firm today where you have lawyers spending 20 or 30% of their time on mundane tasks where they might be able to spend more time billing and working with clients if we're able to take that 20 or 30% of time away from them and automate that through an AI function." While that work is not necessarily high value, it's actually allowing the lawyer to go spend more of their time on high-value pieces.
And so, that is our job as a partner to help understand inside of an organization to make sure that we are setting the that organization up for success, but also we're setting Brenda up for success to do the things that are going to be highly valuable to her. >> To our earlier point about how fast all of this is moving, do you find that you have to spend a lot of time with the Brendas of the world and updating their their priors on this topic? Because like I feel like from my observations, so many people still have AI's capabilities in the back of their head from like where it was 6 months ago and they haven't updated to where we are now. >> The key thing here is we spend time with champions oftentimes that are in the organization alongside these people.
And so, you're not necessarily starting with those people who are coming from a place of not wanting to change. You're starting with those that maybe have already seen it or maybe they got tapped on their shoulder by their boss and said, "You're going to go figure this out." But, the ability to then work through with those people what it really looks like to do this within this organization and then go alongside them to those that maybe are not necessarily going to be the first ones to adapt are generally going to go better.
It's still not a like immediate win, but that capability to build plugins and skills that encapsulate some of what you're going to do does change the equation from past versions of software where you like ERP rollout and you had to go train everyone on like you're going to need to go click all these 17 buttons in this order and these obscure screens. In this case, it's so much more self-describing and you can get it to a place where they just go in and ask Cloud like, "Hey, I need to go do the thing that I always do.
Please go take the next couple steps for me." and they're going to get an output then ideally that you've tailored to give them the information they need to feel comfortable with it. And over time the amount of time they spend on reviewing everything there and then going back to their old way, like as long as that continues to check out the first I don't know, six or seven times, maybe 15 for something more important than somebody that's more uh risk-averse. >> Personal. >> Yeah.
Then the time declines that they spend reviewing outputs. They're still there for the past. The risk you start to run is the opposite. Some people will too eagerly adopt, never read any of the checks, assume that it's perfect, and then get slapped on the wrist or cause some negative outcome, at which point then the question becomes like is that a personal responsibility of that person and you're going to say like you you were bad for doing this or do you want to build the process so that there's less of a way for someone to just blindly trust.
And so they I I think we spend as much time fighting the overtrust issue versus helping get people across the risk-averse kind of chasm. >> So, to both of you as we wrap this conversation up, what excites you the most about kind of the next 6 months to a year of working with Anthropic, of building out continuing to build out uh the ACE program for customers? I mean, how excited are you to to keep growing alongside probably one of the the fastest-growing companies in the market right now?
Jason, maybe we'll start with you. >> I've been working with Google, Microsoft, AWS the last 13 years in a partner capacity, and what we've done in the last 4 months, I haven't really felt in in 13 years. Uh >> Wow. >> Because of the pace at which the product is moving, the pace at which customers are adopting and really the pace at which we have to grow ourselves. And so, you combine those three things together and it really does feel like you're on a rocket ship because we are literally going to uncharted territories in terms of where this is actually going to head.
And that part to me is incredibly exciting. I think we're seeing it from the people that we're interviewing to come to be part of an ace of the ace practice. They are incredibly interested in partnering on this journey because they know that this is the beginning of the next 10 years of Anthropic services partners life. And the fact that we are a preferred Anthropic partner right now, which there are less than 30 in the world, and the fact that we are ingrained with the Anthropic on the field level, that we are ingrained with Anthropic on the product level, that we have our partner ship with AWS, and AWS and Anthropic's partnership is massive.
You know, we're at the beginning of I think is going to be a very exciting time. And you know, I tell this to to Ryan and the rest of the team is like, this is going to be the busiest and the most fun and probably the hardest part of your life because there is no playbook that you can say, "We've done cloud implementations for the last 10 years." They just don't exist. And so, we are building these out, we are finding these use cases that are being common inside of organizations.
And it's just it's a really exciting time to be part of kind of this practice and be part of this partnership with Anthropic. >> Yeah. And And then maybe just to play off of that, ability to use this technology ourselves has been one of the key really exciting things for me so far of like being able to use Claude in a whole bunch of embedded context to make it so that as we're running a consulting business, we're doing it in a way as if you started it.
It really feels like an internal startup that we spun up here that allows us to keep up with the pace that Anthropic is growing because I think that's going to really be one of the differentiators between those that really are the true trusted partners with Anthropic and those that are really good but maybe can't scale as fast as Anthropic needs them to. And then I think that the second thing here, I'm going to give the super nerdy answer as well.
Uh you may have seen in the news that Karpathy joined Anthropic recently and he joined as an individual contributor on the pre-training team. So I've been talking about this for uh maybe the last 6 months or so. Like part of the next step function change here is the models right now are really good at the things that have been able to be verified easily. What he's been working on recently with uh you know, the last couple of releases that he had that it made the news before joining Anthropic is this ability to kind of build those closed loops of learning a new domain.
And so all of this knowledge work and part of the reason I was saying before that like that timeline might compress between code and software engineering and other functions, all of those domains that are out there, there's just thousands and thousands of different ones, they're getting better and better now at making it so that the model can reason its way through a net new domain that it doesn't actually have too much data in the broad internet scale that they've been using to train these models.
So I could really see in you know, I don't know what the horizon is on this and this is not any inside knowledge, it's just me speculating, like the next 6 months, a step change in terms of the model capability. I mean you're already hearing about Mixtral and the step change in terms of code and so forth there that's soon to be released. And so that applied to all of these other domains means that you're going to be just driving this massive amount of change based on something that's not like playing within the limitations of what we're dealing with today where it's like I've got to spend a lot of time custom tailoring all the knowledge and context.
It's that exploratory discovery baked into the model that is, I think, going to be, you know, it really that transformative next step there. >> So, what I'm hearing is we have this conversation 2 months from now, and it'll probably feel very different even then. >> If we had it 2 months ago, it would have felt very different than today, and in a like that 2-month timeline versus the year that we're used to in this world is a lot, I think, what Jason was talking about.
Like, the pace is just so unprecedented. >> Well, I appreciate you both spending some time today with us to dig into it. So, again, Ryan, Jason, thank you for the time. Grant, thanks for piggybacking and co-hosting with me. >> Yeah, it was fun. Thanks for having me. >> Yeah, absolutely. Thank you all. >>
This transcript was generated automatically from the
video's captions and may contain errors.
