Presentation by Arun Saraswat, Methodologist, Scaled Agile, Inc.October 13, 2026
Arun Saraswat, AI Methodologist at Scaled Agile, joined our SAFe and AI-Native Australia meetup to introduce the AI Value Architect: the role that connects AI capability to business results for a team, a solution or a whole ART.
Arun works through the role's five responsibilities: Coaching AI Adoption, Unlocking Value from AI Workflows and Tools, Bridging Business and Technology, Facilitating AI Solution Development, and Optimizing Outcomes. He follows one example the whole way through: a customer service colleague drafting a first reply to a customer, and shows where SPCs, RTEs, Scrum Masters, Product Owners and Product Managers already bring most of what each responsibility needs. The gap he names is AI fluency.
We first saw Arun give this talk at the SAFe Summit and asked him to bring it to our community. Watch the recording below, and read the full transcript underneath.
About the Speaker
Arun Saraswat, AI Methodologist, Scaled Agile, Inc.
Arun Saraswat is an AI Methodologist at Scaled Agile. Drawing on his experience as an Enterprise Architect, Fractional CTO and Founder, he has coached global leaders in aerospace, education, and telecom on turning innovation into business value. Arun co-creates the AI-Native Learning Path, providing the clear playbook organisations need to upskill their workforce with AI, overcome "proof-of-concept purgatory" and embed AI into their core DNA.
Em Campbell-Pretty: Good morning, folks. Welcome to the SAFe and AI-Native Australia meetup. We have extended our reach beyond SAFe to include AI-Native. We're following in the footsteps of the greater SAFe in that sense.
Today we are super excited to introduce Arun Saraswat. Did I say it correctly? Arun, I should have practised beforehand, shouldn't I? But I got it right, so it doesn't really matter. Arun is an AI methodologist at Scaled Agile. He has been there for, I'm going to guess, 18 months-ish. Am I close? Give or take?
I don't know; the first time I recall meeting Arun, and maybe I met him before, I don't know, was in Frankfurt last year, as Scaled Agile was kicking off the AI-Native program. A handful of folks around the world were invited to either Denver or Frankfurt. We got invited to Frankfurt, hung out with some Europeans and Arun and Rashid and Odile for a week, and started to see early versions of what today is AI-Native Value Architect.
Adrienne and I saw Arun give this talk at the SAFe Summit three, four weeks ago; we thought it was quite interesting, and we wanted to bring it to the community. So I emailed Arun, "Can you come and talk to our meetup?" And he said, "Absolutely, tell me when." So this was the first available time slot.
Folks, if you are new to our meetups, we love cameras on if you can. Questions in the chat as we go, and we will try to get them answered. And on that note, Arun, I'm going to pass to you.
From Change Agent to Value Architect
Arun Saraswat: All right, Em, thank you so much. Em and Adrienne, thank you so much for the opportunity.
This is one of the topics that I'm really passionate about. If you've been around the AI-Native scene, you know that we started with a class and training, and a role characterised as an AI-Native change agent, and that's evolved into what we're going to talk about today, which is an AI-Native Value Architect.
If you've been in the AI scene since ChatGPT, give or take, around the end of 2022, organisations have tried to figure out where they can leverage this incredible, revolutionary technology to get more value, produce better results for their customers, and introduce better ways of working. The Value Architect is our answer to that question. So today we're going to talk about a gap that exists between having powerful AI and getting results that an organisation can actually explain.
We're going to spend about 40 minutes or so on the role itself. We're going to cover five responsibilities. This is that wheel that you see up there: a five-responsibility wheel that characterises what the Value Architect is responsible for. And we should have about 15 to 20 minutes at the end for Q&A.
If there is a pressing question or issue, please put it into the chat, and Em and Adrienne, please feel free to just shout it out if there's a question that really needs answering. Otherwise, we'll try to get some room at the end for some Q&A. The discussion is what's going to make this topic really lively.
While we go through this, I want you to think about a potential workflow or case in an organisation that you work for, or a client that you may have, that has a question about where they can really use AI, or particularly generative AI, in the way that they deliver value to their customers today. That's really going to help contextualise the conversation as we move forward.
I'm not going to spend a lot of time on myself, but my name is Arun Saraswat. I am an AI methodologist. Not sure what that means; we do a lot of work at Scaled Agile. Obviously there's been a big demand for AI, and with the latest iteration of AI-Native SAFe, we're incorporating a lot of what we've learned over the past 18 months or so into the framework itself, into the revered, decade-old framework that has helped thousands of organisations deliver better, with better ways of working. AI can absolutely accelerate that, enhance it, and we're looking for a really great future, introducing AI best practices with the SAFe framework and all the things that we've learned over the past decade or decade and a half, with agility and agile ways of working.
The question I keep coming back to, as I think through the role of a Value Architect, is: what should we be able to do differently on Monday? What can you take away from what we're going to talk about today that you can leverage tomorrow, next week? Is it going to really change the way that we do things?
This community, I assume most of the folks here are from the Scaled Agile or SAFe community. You are SAFe practitioners, you've helped organisations adopt the SAFe framework, or at least agility. You already are quite adept at organisational change, and we're going to connect that experience to the particular demands of AI. So that's what today is all about.
Where AI Value Gets Stuck
Before we look at the role itself, let's get some interaction in the chat, guys. When you look at where AI value gets stuck in the organisations you're working with, or clients, do you see that it's stuck at A, where people have access to tools but struggle to use AI well? Put a B in chat if you see experiments work, but nobody changes the workflow that delivers value for organisations. Or do you see a C, where something launches, they have successful solutions and initiatives, but the results are really not clear? It's hard to demonstrate what actually changed for the organisation.
Give me an A, B or C, in terms of what you're seeing. And a combination, if you're seeing... okay, cool, all of them, yeah, so a number of Cs, very nice. Absolutely. Okay, good. I think this is really characteristic of what we're seeing in the marketplace today.
Being at Scaled Agile, I've had the opportunity to talk to hundreds of organisations, hundreds of teams, dozens of ARTs, and had a chance to hear about the challenges that they're having. Typically, they fall into one of these three categories. So again, keep these examples in mind that you've already seen in some of the organisations as we walk through this role, and how the role addresses some of those areas.
A couple of these slides, I'm going to throw some statistics at you. Don't get overwhelmed; it's just a couple of numbers. In a recent survey by McKinsey, they asked over 1,700 people across almost 100 countries about AI, and 80% of the respondents reported that AI improved their individual productivity. But only 6% actually met McKinsey's definition of a high performer, which means that the organisation can quantify significant value: a 5% increase in EBIT, which is earnings before [interest] and tax, an actual bottom line improvement that's attributable to AI. These are different measures.
It's not to say that 94% got nothing, but what's really happening is that there's a difficult translation between what people are feeling when they're using AI and what's actually quantifiable by organisations. Workers are saying, "Hey, look, I got my afternoon back. There's processes and work that took me a week to do; I can get it done in a few hours." That's great. But the finance folks are like, "Well, what are we doing with that better afternoon now?" There's a mismatch there, a kind of mistranslation.
Again, the numbers establish a question for us to work on, not a verdict that AI has failed. And this is what we're seeing across the board. Another recent survey, and I particularly chose these surveys because these are from the past couple of months. There are a lot of other surveys done before, but this is the latest and greatest. Grant Thornton surveyed 950 senior executives across 10 industries, with 39% of CIOs and CTOs describing the workforce as fully ready to adopt AI, whereas only 7% of chief operating officers say the same.
In the chat, why do you think there is a divide between folks on the technology side of the house versus people who are responsible for the operations, or see into the operations? I'm interested in what you've either seen or think about this. Why do CIOs and CTOs think, "Yeah, you know what, we're ready," or almost half of them are saying, "We're good for adoption," from the tech side, the infrastructure side, but operations folks don't see it?
All right, I'm seeing change impact. "CEOs understand the real challenges of adoption," yep. Perspective, yeah, they're seeing the same situation from different angles, absolutely. Nice. Okay, I like that one: AI is a nice, shiny tool for tech, and ops is, "What does it actually do for me? What is the definition of success?" Absolutely.
Again, these statistics and latest studies show a divide where tools, infrastructure and capabilities are being made available to the workforce. The workforce is seeing impact, anecdotally, personally, but it's not necessarily easily measurable, and from an operations perspective, we're seeing a gap. Just because something is made available doesn't mean that it is being leveraged or measured. And that's where we're seeing the gap and also the subsequent opportunity.
The Whole Workflow Is the Unit of Improvement
So this is a familiar way that a gap can arise. If we look at a work stream or workflow, let's say we're looking at the overall workflow, and we're thinking about a draft, say an email or communication. We can make that drafting step much faster. But then it sits in the same approval queue. The work stream hasn't necessarily changed. We're still waiting for the same person to copy the same information into the same system. We've improved a single task while leaving the entire customer experience almost unchanged.
So where does the request actually wait? Who makes the decision? If an exception happens, who's repairing that? We have better prompts that can improve our drafting, but the organisation still has to answer those other questions. Connecting the work across the work stream is actually the opportunity for this audience, and you are already quite capable of that.
I come from more of a technical background, although I was a former SAFe partner before joining Scaled Agile. I think the SAFe practitioners- people who've helped organisations adopt agility- already have a really fantastic skill set that is primed to help these organisations figure this stuff out, and really understand where AI can provide real value. As we get into the Value Architect position and its responsibilities, I want you to think through where you see yourself along this entire pipeline of delivering value, rather than just AI solving a single task.
The AI-Native Value Architect Role
And so we really want to measure the whole workflow as the unit of improvement. Now, Scaled Agile defines the AI Value Architect as a role that helps teams and ARTs connect AI capability to business results. If you haven't already, there's a really great article on our framework site, the Scaled Agile Framework, describing the Value Architect and the role they play within AI-Native SAFe. I would recommend everyone read that.
But really, the idea is you want to connect AI capability to business results. That's the core function of the Value Architect. We can support a single team, an entire solution, or the ART more broadly. The person brings a systems view across adoption, workflow, delivery and outcomes. If you were at the last SAFe Summit, we moved from outputs to outcomes, so that's a big emphasis in SAFe now, and AI can certainly help do that as well.
So we're not asking one person to become the business owner, or the developer, or the risk specialist, or even the coach simultaneously. We're keeping folks accountable in their particular areas. But somebody needs to make the whole path visible and ensure that the gaps between those accountabilities get addressed. And that's where the AI-Native Value Architect sits.
They're characterised in terms of what they do by five responsibilities. Namely, Coaching AI Adoption. That's going to help develop capability and confidence within the Value Architect themselves and also the people that they work with, the workforce.
[The second is] Unlocking Value from AI Workflows and Tools. We want to be able to improve how work moves, and make better use of the capabilities that these organisations have already invested in. This, to me, is what we call in the industry low-hanging fruit. This is one of the easiest ways that you can demonstrably help an organisation actually get value from AI.
The third responsibility, and this should be familiar territory for everybody here, is Bridging Business and Technology, and this helps people reach a workable decision. This is a very classical problem with cyber-physical systems and solutions, and it takes a new dimension when we bring AI into the conversation. We're going to cover this in a bit here. And then the fourth responsibility is Facilitating AI Solution Development. This connects the steps needed for dependable use. And finally, Optimizing Outcomes. This is asking whether the results actually justify the effort in the first place, and then what should happen next.
So they interact throughout an initiative. Evidence can send us back to the workflow. A delivery problem can expose an adoption need, and vice versa. These five are quite complementary responsibilities, and they inform one another as you work with organisations to help realise AI value.
One person or a small group can carry the role, while specialists keep their accountabilities. So again, the idea is not that one person is a Value Architect for the whole organisation. You can exist within a team, within an ART, within a solution, at different levels, depending on where you can influence the organisation to get value. So there is quite a large spectrum here. And again, where are you already on this wheel? Think through that. I'm sure, as folks who are used to bringing organisations better ways of working, some of these responsibilities might sound quite familiar.
What I want to do and demonstrate for the rest of this presentation is that a lot of folks in the SAFe ecosystem are already mostly there. There are a few gaps when you want to move towards not just surviving, but thriving in this AI world that we find ourselves in, where everything is getting disrupted by AI, and there's a lot of apprehension and a lot of uncertainty. Because you're already used to bringing organisations towards better ways of working, you're actually quite well positioned, probably better than most, to adapt to this new world. And it's actually quite exciting.
All right, guys, so we're going to ground the rest of this presentation in these five areas of responsibility. So let's go into the first one.
Responsibility 1: Coaching AI Adoption
Coaching AI Adoption. I want you to imagine adopting Scrum by emailing everyone the Scrum Guide. How's that going to go for everyone? Congratulations, transformation complete: everybody has a Scrum Guide, everyone can read on their own, and that should be good to go, right? Obviously I'm being facetious. Nobody here is going to expect that to work. The guide matters, obviously, but capability grows through coached practice.
The same applies to AI. A licence and a webinar are useful inputs, but they don't tell us whether someone can actually evaluate a result or recognise a dangerous answer. With AI, the coach observes a real task, helps the learner choose an approved approach, and examines the result with them.
"With them" is actually quite a loaded term. You want folks embedded in the organisation, definitely to provide at least training, definitely to show people the path, but also to walk that path with them. People need to know what information they can use, how quality gets checked, and whether they can admit that a trial didn't work. That is building a safety mechanism, and you're all familiar with psychological safety.
This is brand new territory. Are we going to have folks doing things that don't work the first time? Absolutely. Are we going to have folks who thought AI gave them the right answer, but it didn't give them the right answer, ultimately? Is that going to happen? Absolutely. So you need to develop an environment, or help the organisation understand what those guardrails and guidelines are. Organisational conditions matter as much as individual confidence.
So, what I'm going to do is walk through one example across the rest of these five responsibilities and demonstrate that. Our example is a customer service colleague drafting a first reply. Imagine we are working for an organisation and we want to implement an AI solution where we get a customer service request and one of our customer service agents drafts a first reply. We're trying to see when the customer actually gets a good, meaningful first response to their problem.
We begin with a sanitised request in an approved environment. Our colleague tries an approved AI approach, then compares the draft with a clear quality standard. They check that it's accurate, useful, respectful, and supported by the available information. We discuss what they accepted, what they changed, and why. Then we repeat on a different case, so we test judgment as well as familiarity.
And we ask our powerful questions. We all know the skill set of asking powerful questions. As SAFe practitioners, or agile coaches, or team coaches, we don't necessarily have authority, and we can't just rely on authority to tell people what to do. We're not managing; we're trying to coach. So we might ask, "What would make you escalate this?", as an example, rather than just replying on your own. That's a much more useful coaching question than simply celebrating a fluent answer. "What quality checks would you use for this reply?"
Here, people also need time to practise these tools, access the approved ones, and practise those safe routes. And if an error shows up or a mistake happens, they need the psychological safety to learn from those issues. Again, this is very familiar territory to a lot of SAFe practitioners.
As we walk through the responsibilities, I want to show you where different folks in the SAFe ecosystem fit into each of these areas. For SPCs, you can spread learning patterns through communities of practice. Scrum Masters already know observation and feedback. That's already part of the job. RTEs can elevate a repeated access or learning barrier, so every team knows how to solve it. Product Owners and Product Managers make "good" concrete through customer context and quality expectations. They help define what good is, they help define what done is.
The new capability you need to develop is gaining enough AI fluency to coach responsibly. That is the major gap between where most SAFe coaches and practitioners sit, and where they need to get to in order to coach AI adoption effectively. You need to build your fluency up.
Shameless plug: if you want my recommendation for building your fluency, take the AI-Native Foundations class, possibly with a great partner. Possibly with Pretty Agile. If you have an opportunity to take AI-Native Foundations, it's something I would suggest. Keep that learning up, because the gap between seeing where you can coach AI adoption and your already well-practised and well-honed coaching skills is your AI fluency. As you build your foundations and become part of a community that's really doing this, you're just going to get better at it.
And I'm telling you, there are not a lot of folks, especially if you think of people who come from a technology, development or engineering background. Let's just say they're not well versed in coaching. They're not well versed in helping elevate the rest of the organisation with AI practices. They may be good at developing. They may be good at using Claude Code to create AI solutions. But an organisation is not going to look to them and say, "Hey, can you help our sales folks figure out how to use AI to increase our win rate by 10%?" It's a different skill set. You're already there with the coaching practice. You need to fill that gap with AI fluency, and you're good to go.
Responsibility 2: Unlocking Value from AI Workflows and Tools
Let's move into the second responsibility, Unlocking Value from AI Workflows and Tools. A practical early move is to get more value from tools that the organisation already has. This is absolute music to most organisations' ears. These organisations have already invested in tons of tooling, infrastructure and capabilities. A lot of the existing tools already have amazing AI capabilities and features in them. We're not asking them to purchase anything new. What we're saying is, as Value Architects, we can come in and help determine where you can get the most out of the money you've already spent. So: find an unused capability in an approved platform.
Quick poll here. Put in the comments: what is the most common platform or tooling that you're encountering in the field, either with an organisation you work for or with clients you're working with? I have an answer to this, but I want to hear from you. What are most people using where they're actually working?
Interesting. It's actually more diverse than I expected. I'd say 85% of everyone I've spoken to is [using] Microsoft Copilot. But I'm seeing some of you putting in ChatGPT and Claude, so it's getting a little diverse. Honestly, though, Microsoft Copilot is a tremendous tool with a ton of capabilities. And out of the hundreds of teams that I've spoken to over the last 18 months, 99.8%- and I'll give it the 0.2% benefit of the doubt- 99.8% of the teams are severely underutilising the capabilities they have in Copilot. So again, massive opportunity.
What we're actually finding, and there were a couple of studies done, is that AI assistance increased issue resolution for customer support agents by 15% on average. McKinsey also finds that AI high performers often redesigned workflows. So there are two major areas here. Number one, help the organisations get the most out of the tools they're already spending money on. And then, once we have better use of the tools, help the teams figure out how we might redesign our work to take the most advantage of it.
Let's follow the example we're giving here. Someone gets that customer support request, copies some information, searches for context, drafts a reply, and then waits for a review. In a redesign of this, we might say that the request and approved context stay together. AI proposes a draft. A person checks the answer and sends it. And then exceptions go for review.
It's a different way of looking at the workflow. Once you've opened up some of the tools and their capabilities, once you've given AI, in this case, access to the knowledge base, to the customer support ticket history, and so on, it opens up an opportunity to redesign the workflow. We don't necessarily need to send every response to a review, because we're pretty confident now that the AI draft is fairly solid. Every single response starts with an AI draft rather than a person draft. And again, this opens up different opportunities.
This particular responsibility of the Value Architect gives organisations immediate value, because they already have these tools in place. If you're working with organisations, besides the actual AI platform, like ChatGPT or OpenAI or Copilot, and they're using tools like Salesforce, Workday or SAP, all of these platforms have a lot of AI capabilities and features already baked in, and I can promise you they're severely underutilised. Part of that is because the features are getting built so quickly and things are moving at such a rapid pace. Very few people in the organisation can keep up. They just don't know what is available for them to use.
So part of that Value Architect responsibility is having a way to keep abreast of what's happening, and then mining out those features that are going to be useful for the organisation. That is a massive value add to any organisation right now.
From a SAFe perspective, value stream mapping becomes a tremendous AI opportunity tool. Team-level observation reveals unofficial spreadsheets and workarounds. RTEs see a repeated need across ARTs for AI tools and capabilities. Product leaders bring customer problems and the economic trade-off. If you discover an unofficial tool, ask what gap caused the workaround, and partner with security and technology on an approved path. If you've been in an SPC role or RTE role, you're already used to dealing with all of these areas of the organisation. You're regularly dealing with DevOps, infrastructure and security folks. It's just part of what you do when you're trying to help deploy a solution. This very naturally extends to getting the most out of existing AI tools and reconfiguring workflows.
All right, is there any question that I can answer now, or should I continue?
Em Campbell-Pretty: I think we've been doing a reasonable job of being your question answerers, but people can dispute that.
Arun Saraswat: Awesome. Cool, so just interrupt. We have three more responsibilities.
Responsibility 3: Bridging Business and Technology
Bridging Business and Technology. I think I'm speaking to an audience that really understands this problem. It's very classical, and architects in particular used to straddle this interesting bridge between technology and business. But I would argue that as SAFe practitioners, you're constantly dealing with both of these areas. We want to get people to agree, and even to escalate disagreements.
Typically, you'll have a number of parallel conversations happening at the same time. Business is coming in with its concerns. Finance has concerns about cost. Security's asking about data. Technology's asking about delivery. These conversations happen in parallel. We want to be able to bridge them and decide which problem we're actually trying to solve. What is the outcome we're trying to achieve? What are the boundaries? Oftentimes the right people aren't even invited to a seat at the table.
Let me ask you a question. Have you ever experienced a situation where you're discussing, let's say, a customer success problem, or customer feedback, and there's nobody from the customer support team actually sitting at the table? You've got a bunch of heads of finance and leadership there, but no actual representation from the people who deal with the customers. That happens all the time. Value Architects are there to bridge this divide.
Like I said, this isn't a new problem, but what happens when we're talking about AI? I characterise AI as an amplifier. Because of its nature, because it moves so fast and enables so much capability, if you have a problem like bridging technology and business in the organisation, AI elevates it, highlights it, amplifies it. These cracks in the change management or the organisational way of working get elevated and amplified when it comes to AI.
Let's think through our customer service example. The intended outcome is less delay while maintaining answer quality, protecting that quality. The boundaries are going to include what approved information we can use, human review, and the escalation route. A named business owner owns the service outcome, supported by technology and the specialists. This is concrete enough to reveal a disagreement. If one person's definition of success is that an automated answer gets sent to the customer in under five seconds, that may not be everybody's definition of success. What if it's a poor answer, an answer that didn't really address the concern, or it wasn't reviewed? Where do those boundaries sit? Where does the disagreement happen? It's almost as important to highlight the areas of disagreement as it is to get alignment, so we know what everyone's dealing with.
So the Value Architect is there to connect the conversation. Business owners decide the business trade-offs, and the specialists retain their professional authority. What's interesting is that if you take the AI-Native training, we include a lot of artefacts and assets to facilitate these discussions. You already have a toolbox, you've been doing this work, but it takes a bit of a different shape and form when AI enters the discussion. Bridging business and technology is a classical problem you've been dealing with already, but it takes a new and interesting form with AI in the conversation.
RTEs can keep the outcome intact while multiple teams negotiate dependencies. Product leaders turn "we need AI" into an actual proposition that can be tested. SPCs can get the right people into the conversation, and Scrum Masters help surface assumptions without making disagreement unsafe.
There's so much in here, guys. My background is software architecture and enterprise technology. I speak to a lot of technologists, engineers and software folks, and this is not familiar territory to a lot of them. They're good at AI, but they're not necessarily good at bridging technology and business. Whereas this audience is. All right, that is responsibility number three.
Responsibility 4: Facilitating AI Solution Development
Number four is Facilitating AI Solution Development. Remember I said earlier that getting the most out of existing tools and looking at workflows is the low-hanging fruit. A lot of the time, organisations want to deliver AI solutions or initiatives that require things to be built. If you've been on ARTs and helping coach Agile Release Trains, Solution Trains, large solutions, coaching teams, the delivery lifecycle is your bread and butter. You know this stuff. We've just adapted it a bit to AI.
Here I want to introduce our AI-Native Solution Lifecycle. This is provided in the training. It gives you a route through sense, discover, design and deliver. These are four phases. You sense the signals and needs. You help discover an opportunity. You help design a viable approach. This doesn't mean that you're the architect of the solution. It doesn't mean that you're the technical developer. You're going to bring the right folks into the room to help facilitate that. And then you're going to help deliver it with evidence that helps you decide what to do next.
Q&A
Attendee: So can I jump in with a question?
Arun Saraswat: Yeah, absolutely.
Attendee: Okay, so this is all really good stuff. So, two things that I think you might find funny. I've had a consulting practice [for many years], and the theme of the practice is bridging business and technology. I had been in the technology space for 20 years and realised that was always going to be the easy part. The difficult part would be: how do you leverage that technology for good?
I have two questions here. I'm curious whether SAFe identifies this Value Architect outside the context of AI, because it's needed whether you have AI in the picture or not. We also know that companies grow up as silos, functional silos, and that you don't have anything that brings those silos together to unify them. This value role is a logical thing to do that.
So my question is, do you feel like this AI-Native Solution Lifecycle is going to just scale the silos, or do you feel like it is really going to do something to unify and connect those silos and bring an enterprise lens to these solutions?
Arun Saraswat: Yeah, great questions. Thank you. So interestingly, I've actually never thought of the Value Architect role outside of AI. Well, we have postulated, and it started as a role that wasn't necessarily dependent on SAFe. So, whether an organisation is leveraging SAFe or not, the AI Value Architect makes sense to help do those five responsibilities, whether they're using SAFe or a different methodology. So that has been postulated.
We used to call this role a change agent, and I think generally that's a general term that could be applied outside of the realm of AI. Value Architect as a term isn't new. I don't know if it's a specified title, per se, but it has been postulated, similar to what you're saying, that it is a role that brings a bridge between technology and business. I don't know if it's been formalised as that outside of the AI space, although I certainly think there's room for that to…
Attendee: Yeah, because I am now [working on value management], and that's the point of alignment to bring all this together. So there's very, very similar thinking here.
Arun Saraswat: Yeah, absolutely. We certainly built on established practices. The thing that makes our offering and thought process unique is the uniqueness of the challenges that AI brings to the space. So, to answer the silo question as well, when we walk through what an AI-Native solution workshop looks like in a team or organisation, we need to bring the right folks with the right seat at the table.
These things down here, the Alignment Brief, the Value Blueprint and the Implementation Tracker, are artefacts and assets that we provide that are tailored to the AI challenges. So specifically, let's say for risk management and compliance, we discuss the EU AI Act. We discuss some NIST standards that are particular to AI guidance. So these take a particular shape, because the nature of the solution is AI-Native, or embedded with AI. That's where the difference is, and that's where we put the uniqueness of the program around those areas.
But certainly, it's definitely meant to break those silos down, and that's where the Value Architect positions themselves as the connector between all of these different concerns. We boil it down to six areas: a value proposal, data strategy, AI solution, prod ops, risk, and then finally a viability assessment. These six areas, if done properly, would really help resolve so many of the issues that organisations are seeing. If you looked at some of the studies that were done earlier in 2026 and late 2025, anywhere from 80 to 95% of organisations were failing to realise the return on investment in their AI initiatives.
A lot of times it's because this alignment didn't happen before actually making a go-no-go decision. We distilled a lot of those factors into a very doable one-day workshop, or an async workshop, that could work with the right stakeholders in the room. And the whole idea is to break those silos down and get the right people in the room discussing the situation.
Attendee: Let me just ask one more question, then I'll be quiet. Where do you make that decision about what the right place is to bring AI into part of the solution? Is that the right investment for AI? Is it worth it? So, in the solution space, where does that conversation take place?
Arun Saraswat: It's happening almost everywhere. We'll begin it from here, where the Value Architect is keeping an ear out and building that sixth sense. It's grounded in reality. Teams might have problems, individuals or leadership might be complaining about something. You're listening and sensing those signals. It doesn't start with where can we put AI, it's what are the problems we're facing?
Then, because the Value Architect has that level of fluency and understanding, they can start saying, hey, you know what? Let's look at what shape these signals are taking. Are they an operations issue? Is it a competitor issue? And based on what I know, and the folks that I can bring in, from understanding AI fluency, where could AI possibly help? And that starts the conversation.
So it definitely starts with a practical question: where are the problems, or what are the challenges we're facing? Not the other way, which is, because of board pressures and market conditions, a lot of organisations find themselves saying, hey, you need to use AI. Well, that's nice, but it should start with the problem. Not find the nail to use this hammer. We've got nails sticking up, so where do we use it?
Yeah, great questions. Thank you.
A First Release with Clear Boundaries
All right, guys, so, continuing with facilitating these AI solutions. Taking that example again, we might have a first draft where there's a useful reply from approved content, or a colleague checks and sends it. Autonomous refunds and case closures stay outside its scope. So we want to scope what's in and out of this particular solution. What are the boundaries for the first release? What are we going to use to actually launch and then learn? Again, PDCA, small batches of work. This is all stuff that's really familiar. But it gets thrown all over the place when it comes to AI solutions because of whatever pressures are there.
If you, as a Value Architect, can bring this kind of discipline and approach to AI solutions, it actually makes a massive difference in the way that organisations operate these solutions. Ultimately, the customer still gets a service. What happens when AI isn't available? What happens if we can't reach our inference? What are the edge cases? What's the failover? Again, all the things that we've already been talking about when we talk about delivering in a SAFe framework applies to AI solutions.
I'll just leave this here quickly if you want to take a screenshot. Scrum Masters here can help the teams make evaluation part of the Definition of Done. RTEs can expose operating and integration dependencies. Product Owners and Product Managers help define a useful release with explicit limits, and SPCs can support the organisational practice that makes all of this repeatable.
Responsibility 5: Optimizing Outcomes
All right, I know we have a couple of minutes left. I'm going to quickly go through the last one, and I would say this is one of the most important ones: Optimizing Outcomes.
As a responsibility, ultimately, are we looking for more customers served, less delay, better quality? When capacity frees up, what do we actually do with that? What does AI enable us beyond just the outcome of the solution itself? Oftentimes we'll hear an outcome of, okay, we want to be live by Q3. That's not an outcome, that's a delivery date. So we want to move from, number one, understanding our current baseline based on metrics. Currently, we have two days for a first response to get to our customers, and we want to move that down to two hours. That is a real outcome. That's not saying we want the system live by a certain date. What are the guardrails? Well, we want it in two hours without sacrificing quality answers, without sacrificing our customer satisfaction, and remaining within operational costs.
All of this is going to help us make a scale, revise, pause, or stop decision. We all know our old adage of... pivot or persevere without mercy or guilt. You're taking the same practices and applying them to AI solutions, and it's just going to be amplified so much more.
Once again, as SPCs, Scrum Masters and RTEs, you can bring that discipline in. Honestly, this particular responsibility, Optimizing Outcomes, is one of the biggest challenges, and it takes us all the way back to the beginning of the conversation. What did we want to do with AI in the first place? That whole bridging of technology and business. What are the actual outcomes? How do they get defined? How are they measured? And ultimately, it's going to help us improve. Again, all the same roles are well adapted to this.
AI Skills in the Australian Hiring Market
The last things that I want to say, guys, beyond the five responsibilities. I just did a quick search on the Australian market for roles that require AI fluency and AI skills, just in their job descriptions. It's increased by a factor of two over the period of 2024 to 2025. There were about 20,000 roles in 2024 and 41,000 in 2025 that had that embedded in their job descriptions, and this is across multiple roles. We see it as increasing. You may not necessarily find the role AI-Native or AI Value Architect in a job listing yet, similar to how when RTE started, people were doing that job, people were performing those responsibilities, but it wasn't a formalised role or title. Same thing with Value Architect, but you are seeing certain listings for AI consultant, AI process transformation, and forward deployed engineer [FDE].
Give me a number one in chat if you've heard of the role forward deployed engineer and if that's a popular, increasing role that you guys have heard of.
That's been popularised by a few organisations, OpenAI for one, for sure, and Palantir is another one. There's actually quite a bit of overlap between the Value Architect and an FDE. An FDE lends itself more towards the building solution side of the house, where they're expected to build a solution: number one, helping define it, and then actually building it out. We don't focus so much on the engineering side, but if you were a Value Architect, my proposition is that you actually have a broader set of responsibilities than an FDE, because you cover a lot more on the business side of things. And if you get your AI engineering chops up, you can absolutely perform incredibly well as an FDE.
FDEs are getting paid well right now. So to me, it's a great stepping stone to an FDE. There's just a bit of engineering discipline that you have to develop, and you're extremely well positioned for that role. In fact, like I said, for Value Architect, we cover a lot of things, I would say getting the most out of existing tools and infrastructures, that FDEs aren't really positioned for. And we put a lot of emphasis on that.
Where to Start
So guys, your experience, like I said, really belongs in this role. You already come with a vast strength and background in coaching, systems thinking, product and delivery judgement, and really the idea is to add some AI fluency and get started.
Again, thank you, everybody, for the opportunity. I have a suggested path here. You don't need 90 days to do any of this. AI helps you cut down a lot of this stuff, but first thing, take some training if you can. Otherwise, find an opportunity to practise some of the things we talked about. Look at the tools that you have available, how you're getting the maximum out of those tools. Look at the organisations you work with. Can you help coach them in their practices? Try and experiment and assess where you are, and pick one workflow, one conversation, one date. Start today, basically.
All right, guys, we don't have 20 minutes for Q&A, but I think we've got some good questions in there. I'm sure Em and Adrienne answered some great questions. Otherwise, thank you all for attending, and I hope you enjoyed the presentation.
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behaviour or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.