

人間中心のAIを築く:Scotiabank の Gautham Pallapa 氏が語る価値定義とエージェント活用
Practical AI
要約
Practical AI の Chris Benson が、Scotiabank でAI・イノベーションを統括する Gautham Pallapa 博士と、銀行でのAI導入や著書『Invaluable』を語った回。Pallapa 氏は、まず組織にとっての価値と成功の定義を固め、そのうえでAIが本当に必要かを問うべきだと述べる。コーディングなど反復的な作業はエージェントに任せ、人は設計や創造性など上位の仕事へ移るという見方を示し、マルチエージェント構成の実験例も紹介した。
この回のポイント
- ●銀行は規制とリスク回避の文化が強いが、COVID を機に顧客体験が変わり、生成AIやエージェント型AIへの投資が進んでいると説明している。
- ●AI適用の出発点は、組織にとっての価値の特定と成功の定量化であり、その後に「AIが本当に必要か」を問うべきだとしている。
- ●現時点では、個人の口座をエージェントに接続して取引させることには抵抗があり、顧客対応やコンプライアンス支援、アドバイザー支援から進めると述べている。
- ●エージェントは非決定的な確率モデルで、人間のような文脈や意図の理解には届かないため、人が関与する前提が必要だと語る。
- ●インターンを「スタートアップ型」チームにして、なぜAIが必要かを問うことで、バイブコーディングではなくエンタープライズ品質のソフトウェアエンジニアリングを教えている。
- ●マルチエージェントは、BA・設計・テストなどの役割を持つチームを模す形が出発点で、スウォームは複雑な分割統治向けと位置づけている。
章立て
ゲスト紹介と経歴
Pallapa 氏が学術界、Pivotal・VMware でのクラウドネイティブ経験を経て Scotiabank に至った経緯と、同行の概要を説明する。
保守的な銀行がAIに向かう理由
リスク回避的な銀行業界が、COVID を契機とした顧客期待の変化を受けて生成AI・エージェントに投資している背景を語る。
AI適用先の選び方と避ける領域
価値の定義、成功の定量化、AIの必要性判断の順に進める考え方と、人の監督なしの自律取引はまだ早いという慎重な見方を述べる。
書籍『Invaluable』の枠組み
600時間超のワークショップ等を基に、価値の特定から測定、優先順位、AI活用、阻害要因、リスクまでを扱う8要素のフレームワークを紹介する。
人間の役割と変化への向き合い方
幸せな人が生産的という信念や、コーディングが低価値化する中で人が上位へ移る見方、エージェントは知的ではなく数理的だという説明を述べる。
インターン施策と開発ライフサイクルの実験
スタートアップ型インターンで「なぜAIが必要か」を問い、ユーザーストーリーから試作までを高速化する取り組みと、開発工程の見直し実験を紹介する。
マルチエージェントとスウォームの考え方
役割別エージェントのデジタルツイン的チーム、オーケストレーター、評価陪審の構成を説明し、まず人間のチームを模すべきだと提案する。
今後の展望と人への還元
FOMO でAIを導入せず価値を問うこと、グローバルな摩擦削減、技術で人の生活の質を高めることを展望として語る。
解説記事
Practical AI の今回の回では、共同ホストの Chris Benson が、Scotiabank でAI・イノベーション担当の Principal Director を務める Gautham Pallapa 博士を迎えた。著書『Invaluable』の内容とあわせ、規制の厳しい銀行でAIやエージェントをどう導入するか、人を中心に据えるとはどういうことかが語られた。
保守的な銀行がAIに向かう背景
Pallapa 氏は学術界や Pivotal・VMware(Tanzu、Kubernetes 周辺)を経て、現在は Scotiabank の二つの事業部でAIとイノベーションを率いている。同行はカナダの大手5行の一つで、リテールだけでなく資産運用や資本市場なども手がけると紹介された。
同氏によれば、銀行はリスク許容度が低く、顧客資産を守る責任があるため、伝統的に保守的だった。ただ COVID でリアルの店舗に行けなくなり、スマートフォンで完結する体験が当たり前になった。その結果、摩擦を減らし価値提供を速めるために、生成AIやエージェント型システムへの投資が進んでいると述べている。カナダの大手行は、先行する米国勢を追う段階だという認識も示した。
「AIありき」ではなく「価値」から始める
AIを使う案件の選び方について、Pallapa 氏の答えは一貫して価値から入るものだった。まず組織にとっての価値と戦略上の優先事項を特定し、次に成功の定義を定量化し、その上でできるだけ早く達成する方法としてAIを検討する。価値は利益だけでなく、顧客満足や利便性も含む。
AIの使い方についても、人間がタスクを指示するボット型から、エージェントが実行する形、人とエージェントのチーム、さらにエージェントが運用を担う形へと広がっていると整理している。一方で現時点では、個人の銀行口座をエージェントに接続して取引させることには抵抗があると明言した。まずは顧客対応、コンプライアンスやリスク、アドバイザー支援のような領域が現実的で、人が介在しない運用までは信頼が足りないという立場である。
『Invaluable』と人の位置づけ
『Invaluable』は、同氏が以前のエグゼクティブアドバイザー兼フィールドCTOの時期に行った、600時間超のワークショップや400件超のインタビュー、500社へのコーチングを土台にしている。価値の特定、成功の定義と測定、優先順位付け、AI活用、阻害要因の解消、リスクの重ね合わせといった要素を8つのセクションにまとめた枠組みで、表紙のタコがその象徴だという。AI活用の章は変化の速さゆえに6回以上書き直したと語っている。
人の扱いについては、「幸せな人は生産的」という信念と、技術の根底には共感があるという考えを挙げた。AIによって、コーディングや文書作成、テスト作成といった反復的な作業の価値は下がり、人は設計や創造性など上位の仕事へ押し上げられるというのが同氏の見方だ。Benson 氏も、長年コーディングしてきた者として、この切り替えは感情的に難しかったと共感を示した。
エージェントは知的ではなく数理的
変化への抵抗にどう向き合うかについて、Pallapa 氏は、エージェントは非決定的な確率モデルであり、知的に見えても数理的に最適な行動を選んでいるに過ぎないと伝えていると述べた。AGI は遠く、実現するかも分からないという。指示を増やしすぎると混乱して幻覚を起こすこと、障害対応で培われた経験に基づく文脈や意図の理解は人間側にあることも強調した。
そのうえで、人を置き換えるのではなく、手作業の摩擦を減らし高価値の仕事へ移すことが目的だと対話で伝えている。個人にとっての利点を示すことが、納得を得る鍵だという。
インターンとマルチエージェントの実験
具体例として、同氏が率いるインターンシップの刷新が語られた。インターンを安定したチームにして、テックリード、デザインリード、プロダクトオーナーを付け、経営陣を投資家に見立てたピッチ形式にしている。AIを使いたいと言う学生には「なぜエージェントが必要か」と問い、流行だからという理由を退ける。「バイブコーディング」という言葉は使わせず、エンタープライズのソフトウェアエンジニアリングとの違いを教えているという。
開発工程そのものも実験対象で、ユーザーストーリーから設計と高速な試作へ進み、顧客と反復する流れを試している。従来の逐次的なライフサイクルをフィードバックループ型に変えられるかも検討中だが、まだ確かなデータはないと断っている。
マルチエージェントについては、BA、設計、テストなどの役割を持つエージェントをデジタルツインのように並べ、オーケストレーターが統括し、共有メモリやガードレール、評価役の「陪審」を置く構成を説明した。スウォームがすべてを解決するわけではなく、まず人間のチームを模すことから始め、複雑な分割統治が必要な場面で次の手法を検討すべきだという。MCP Dev Summit で5つのエージェントパターンを共有したことにも触れた。
まとめ
今回の議論は、AIの導入を技術ではなく価値の問いから始めること、そして人を置き換えるのではなく上位の仕事へ導くことを軸にしていた。日本の企業、特に金融のように規制や慎重さが強い業界にとっては、低リスクな領域から始める姿勢や、「なぜAIが必要か」を問う習慣が参考になるだろう。また、エージェントの能力を過信せず人間の文脈理解を活かすこと、マルチエージェントでは既存の人間チームの型を参考にすることなど、すぐ試せる視点も多い。なお、内容は Pallapa 氏個人の見解と経験に基づくもので、効果を示す定量データは番組では示されていない。
文字起こし(英語・自動生成)
Welcome to the Practical AI Podcast, where we break down the real-world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, X, or Blue Sky to stay up to date with episode drops, behind-the-scenes content, and AI insights. You can learn more at practicalai.fm. Now, on to the show. Welcome to another episode of the Practical AI Podcast. I am your co-host, Chris Benson. Daniel Whitenack is out. I know that some of you who listen to the show regularly know that he was expecting and he has had his first child, a daughter named Dot, who was born a few days ago.
And Daniel is wisely staying out of the show briefly long enough to spend some special time with Dot. So I hope you will join me in wishing the family very well as wife Brittany. And with that, I'd like to dive right into our show this week. We have a fantastic guest, so much experience. I can't wait to ask questions here. I'm welcoming Dr. Gautam Palapa, who is Principal Director for AI and Innovation at Scotiabank. And he is also the author of a book, Invaluable. And looking forward to talking about all those topics. Welcome to the show today. Thank you, Chris. Thank you for having me. And congratulations, Daniel. It's a great time, especially becoming a new father. It's an amazing thing. Your entire world is going to get rocked. So, yeah, good luck for that. Yeah, I think he's going without a bit of sleep. I know he really wanted to join today. We were texting earlier, and I was like, stop talking to me and go back to your baby.
So anyway, with that said, I was wondering if you could give us a little bit just about your background. You have a really interesting mix of kind of skills that come together, I know, and I know we'll get into that in this conversation. Can you just tell us a little bit about kind of where you come from and how you've ended up at Scotiabank and the position that you're at? Yeah, sure. So I actually come from academia. I spent about six years in academia teaching various topics. I did my PhD in machine learning and artificial intelligence. I'm kind of an old person working in traditional AI and ML, as they are called today, or classical AI ML. I worked at a number of companies, have always been in the cloud native and application development realm. And then I worked with products for a long time. I worked at Pivotal and VMware, where I was working on Tanzu, more on the cloud container space,
more all about Kubernetes, trying to make sure that we can transform applications and move them to the cloud. And then I got this opportunity. My current CIO, he reached out to me and asked me if I could come in and help transform portions of the bank. And so here I am. I'm helping drive AI and innovation at the bank for two large business units. We do a lot of experimentation, and I'm trying to bring in rapid prototyping, iterative development, and all the cool things that, you know, everyone's doing nowadays. And I guess as we dive into the conversation, I know a lot will be about the bank. We have a global audience. There will certainly be some people in different places that may not be familiar with it. Could you do a brief introduction to the bank? Of course. So, Special Bank is one of the top five banks in Canada. It has a lot of presence in the entire North America corridor and EMEA. So we focus on not just retail banking, that is what people think of when we look at banks, but also global wealth management, banking and capital markets, private lending, a lot of other things.
So anything related to money and to financial services is kind of what we do at Scotiabank. I have the honor of working with a lot of very smart individuals and teams across the globe. and our entire goal is to help people on their journey and to help them use the financial assets that they have and help open doors and opportunities for them so that they can grow richer and wealthier and also have a much higher quality of life. So that's kind of the purpose of Scotiabank and I help lead AI and innovation for two of the business units there. So I'm kind of curious. I think traditionally when people think of banking, they think of, you know, conservative and maybe not right on top of the latest technologies and stuff. And you have clearly, you know, from what I already know, have dived in and really made some differences.
Can you talk a little bit about how the bank got to the point where they wanted you to come in and do this? And what does that look like in the world of banking to bring AI? What do some of the key things look like before you've introduced AI? And then as you've done that, how does that change? And what does that mean? And how does that relate to the rest of the banking? Yeah, so banking traditionally, and by good reason, is extremely risk-averse. They have to make sure that they're compliant. They have to make sure that they are protecting not only the assets of the people, but also of the bank itself. There's a lot that happens when you have that level of risk. And so in the past, banks have always been very conservative. But that is because, you know, the risk profile and the appetite is low.
We want to make sure that people don't have to ever think about where the money is and they always have access to their wealth in any way, shape or form. I think what happened over the years is that as organizations started moving faster and people and customers started expecting things and services in a different way. And a lot of it actually goes to COVID. We were completely virtual and there was no way that we could go to a brick and mortar building and actually transact or perform any kind of services. So it was inevitable and forced every single organization to try to take a step back and change the way that they're approaching customer experience and the customer journeys. And I think a lot of that is what translated into this modern view of looking at things banking included. So now everyone has a smartphone, they have an application, they want to do pretty much everything on their phone. They don't want to go and talk to someone or, you know, introduce artificial friction to make a transaction occur.
And so it's inevitable that we need to look at our processes and our approaches and make them, you know, ready for the newer generation and for the various people. And this is why I think all the banks and every financial institution is looking at generative AI and artificial intelligence in order to help them achieve that, because you are accelerating the delivery of value and the speed to market with all of these things. That's what the banks started to look at, especially the top five banks here in Canada. The U.S. have already been doing a lot of it in the past, and the top five banks are starting to catch up with them. And so there's a lot of investment into generative AI and agentic systems to try to help them because it's one thing to improve a process. It's another thing to reduce the manual friction and the way in which people interact with us and with the systems.
And I think the organizations that take a deeper view and take a step back and try to look at what business value they're trying to develop and how can AI or any other technology help them are probably the ones that are going to have a huge amount of benefit. I'm curious, as you talk about kind of the risk profile and, you know, the fact that there's regulation, there's a lot of laws around banking. As you come in looking for opportunities, how do you make the assessments, given all the constraints, you know, that the whole business is operating under? How do you pick a ripe target for generative AI to come in and make a substantial difference for operations or for customer interactions and stuff like that? And how do you know what to stay away from? Like, if you could also talk a little bit about I'm not going to go there. You know, I'm sure. So this is a very vast field and it's it's it depends from organization to organization.
And the reason why I say this is because every organization has its own heritage, its evolution, and it's set of opportunities that they can work on in order to improve. If you look at every domain. So let's focus a little bit on banking. There are only certain things, certain services that banking provides. It's not like a huge catalog of things. There are a finite set of things that need to be done. So I think the first question is to find out what exactly is value for the organization, right? Like, let's identify those. What does value mean to us? And how does it connect to our strategic imperatives for our organization? How does it drive? Not just the bottom line, because at the end of the day, it is getting profit. And especially if you're publicly trading, you want to generate more revenue and you want to become more profitable, increase the ROI. But in addition to it, it's also customer satisfaction.
It is about how you can make it much more convenient to people. And so there are so many other strategic imperatives. So once you identify the value, the next thing is to actually understand what is the definition of success. Just because you identify what the value is does not mean that you know what success means. So quantifying what success means is probably the next thing. And once we have those two things, the next question to ask is, so how can we achieve it as quickly as possible? And probably that where then the question is how can AI help And it is not just AI because there are multiple ways right Like in the past we always especially for fraud we been using traditional AI ML algorithms for a very very long time But now with generative AI and agentic AI systems, you have this, it started off with this human, you know, task agent bot approach. And then afterwards, as we have started evolving more, it's become the human task agent.
So now we have the agent actually going ahead and doing things as much as possible. And then we've gone into the human agent teams and finally agent operators. So there are multiple ways by which you can use AI for this. And I think exploring that, we would definitely start looking at what is the low-hanging fruit or the things that are not going to impact us that we can use. And then afterwards is probably going to be what are the things we should not. So, for example, at this time, I probably don't have the appetite to connect my personal financial accounts or bank accounts to an agent and have it perform transactions or do things. because pretty soon I may not even have visibility into what the agents are doing. So maybe today I don't have that appetite. Maybe we'll look at much more simpler things like customer service agents or, you know, chat assistants and trying to help with compliance and risk, any kind of, you know,
looking at the help of the portfolio of the person itself and trying to provide support to the advisors is probably what we're looking at. But I don't think we're going to be radical enough where we're going to say there's not going to be a human in the loop at all or a human at the helm. I just have agents do things. I don't think we're ready, and I don't think we trust the agents that much yet. Imagine a smart home you run by talking to an AI assistant like you're talking to another person. I could see myself creating a website and showing people how to set that up. But making that site look professional and feel intuitive, that's where I'd want help. Framer is the AI website builder that helps creators, teams, and businesses ship production-ready sites faster than ever while getting every detail right. Prompt, inspect, edit, and publish in one place at a whole new pace. For a project like that, I'd look to Framer to help turn the idea into a polished site.
Agents help draft and refine the pages. I decide what goes live. Learn how you can get more out of your site from a Framer specialist or get started building for free today at Framer.com slash PracticalAI for 30% off a Framer Pro annual plan. That's Framer.com slash PracticalAI. Rules and restrictions may apply. So as we come into our next segment, I really wanted to start asking you some questions about Invaluable, the book that you've written. I know there's some key things in there that I know our viewers will be interested in. Could you tell us a little bit about what Invaluable is about, why you wrote it, and some of the key messages that you're trying to impart with your insights and wisdom? Yeah, sure. So, invaluable is something that came out of a lot of interactions with executives and with organizations.
So, in my previous role, I used to be an executive advisor and a field CTO. So, I used to work with executives from Fortune 500 companies and help them in their transformations. So, we had various workshops and exercises that we used to do with the executives and various interviews. where we try to help them understand how they have to transform their organizations. And it's not just from a technology perspective, but it's also from a process and a business perspective. Because the goal of any transformation is to reduce the amount of friction from ideation to business value, right? So that's the goal of what we were doing. And so over the last, I want to say, the last five, six years, what we've been doing is talking to all these executives and running like over 600 plus hours of workshops and maybe 400 plus interviews with various executives and coaching 500 companies. And then we started realizing, my co-author and I, we started realizing that there's no actual
framework to help identify the value and the delivery of value. And so we started writing about it, and that's what invaluable is. It's a 360-degree view of helping executives, transformation agents, change agents, and even technologists to identify what is truly valuable for the organization. Then we help them define and quantify what a definition of success means, and I touched about it a little bit earlier. And once we identify and quantify what success means, the next thing is to actually measure it, because if you don't measure it. You don't know if you actually have to pivot or procedure in the direction that you go. And then after that, measuring success, the next thing is to say, okay, let's establish the priorities because everything is not created equal. So we do the prioritization of it. And after that, we ask the question, how can I leverage AI? Especially in this generative AI and agentic AI world,
the next thing is to ask yourself, how can I leverage AI to achieve the success that I want so that I can deliver the value that I need to deliver? And once I leverage the AI, the next question I ask is, so what is preventing us? What are the impediments that are preventing us from doing these things? And that goes into optimizing the processes and any other tech stacks that are necessary to leverage the value. And then finally, we overlay the risk around it because everything that we're doing is about risk. So it's kind of a happy mix of understanding complexity, understanding productivity or value, and then understanding risk. And then seeing which of these levers and appetites we have within the organization and providing a much more objective and very detailed way for people to start delivering value within their organizations. And that's why we introduced the value transform. It's got eight different sections that you look at.
And that's why we went with an octopus. Like if you look at the cover of Invaluable, it's actually an octopus. So we very fondly call it an octopus. And that was why we wrote Invaluable. It's actually battle-hardened. We have used it with a number of Fortune 500 companies and helped transform them. and so I think it's just a culmination of all the work that my co-author and I have done. So it took a while. I think it took us about, I want to say, two years of writing. And there's one chapter, especially the one around leveraging AI. We had to rewrite it more than six times because the world was changing so fast and finally our editor was like, you need to stop at some point. Otherwise, there's no way you can publish the book. And so, yes, it was a great experience. That's fantastic. I'm busy working on a book in the late stages, and it's been going for quite some time, like, you know, a couple of years.
And you just now said a couple of things that really resonated, like, where do you stop? You know, where do you wrap it up? So I may offline have to pester you as one author, aspiring author to another to figure that out. You know, you were well known for, like, bringing empathy of the people in. Like, how you have this sophisticated, you know, battle-hardened process that leads to bringing technology, AI in this case, into that. How do the people fit in? We're in this moment in, you know, in human history where all this, as we've talked about, is changing so fast. And it's having, you know, huge and unforeseen results, the impacts on people and their families. How do people fit into the process so that everyone wins and that they can kind of retain their humanity as they go through?
Could you talk a little bit about how you've implemented that? Yeah. So I want to start off with a couple of things. The first thing I want to say is I want to share my primary belief. My primary belief is that happy people are productive people. So everything that I do is to make my people happy and then automatically amazing things happen. It's not that hard to get into that mindset because at the end of the day, it's never technology that actually creates value. It's people. Technology can only be an accelerator and it can take you in the direction that you want it to go. But it's people who are really the true value creators. So that is the lens that I start everything with. The next thing that I firmly believe is that technology is actually rooted in empathy. If you think about technology and everything that you're doing, the reason why we use technology and deploy it within our organizations is to reduce some kind of pain. It could be pain for the customer, and that's where you have customer empathy,
And there are lots of techniques and ways by which you can understand, put yourself in the shoes of the customer and try to understand the pain that they're going through. Or it could be internal to us, right? Like as employees, we have employee empathy as well. And that's where you're optimizing processes, reducing the amount of friction points and manual toil that people are going through. So that's the second lens that I have. Now, with artificial intelligence and especially with generative AI and authentic AI, what has happened is it's really changed the way that we value our tasks and the things that we do. So, for example, coding used to be like one of the most valuable things for software development. People used to say, like, if you're a software engineer or you're a developer, then you are set. Like, it will never go away. But all of a sudden, that fundamental principle is shook. Like, you don't actually need people to be coding anymore. And actually, you shouldn't because the systems and the agents and the applications, they can actually write code a lot faster than we as humans can.
Maybe it's not performant. It's not, you know, as elegant as we would as humans, but it gets the job done for now. And maybe it's not enterprise grade either, right? So what technology has done, and especially in this last few years it has done, is it has helped us start looking at and this is where I get super excited it has helped us start looking at everything that we do and then identify things that are high activity for us as humans And then our entire goal is to make these low activities like coding writing documentation, writing tests, all those various things that are systemic and repeatable. Those are now low-value activities, and so they can be handled over by systems and they can be done. And so all of a sudden, what excites me the most is it's taking artificial intelligence for us to really value humanity and human values and creativity and all the things that make us as a human being and pushing us up the stack.
We are evolving thanks to our digital assets. That's an amazing time to be. It's a great insight there. And I love that perspective. And I certainly, as someone who most of my career was software development, you just described an adjustment that I personally had to make was, you know, after decades of coding to recognize there was a better way and that whether I thought I was a very competent coder or not, that still wasn't the path forward. So I appreciate that. That was hard to do, though. It was hard to make that emotional adjustment. As you're bringing this insight and perspective into the organization, how do you get all the humans, the resistant Chris's that you're dealing with, to recognize that and to understand that you are pushing them up the stack and you are kind of supercharging them along the way? How do you get them to make that adjustment?
Because that's not an easy one to make, speaking from experience. Yeah, it definitely isn't, right? And I've grown up the ranks. I started off in development as well. So I totally empathize with you. I understand. Like, it is hard. It's a mindset shift. And people get nervous when the status quo or when they're, what would they perceive as identity is questioned or shifted. And so this is one of those things. It's like all of a sudden you read so much of hype and hyperbole about, oh, you don't need people anymore. Like agents can do all these various things. One of the things that I do regularly, and I talk a lot about this, not only within our organization, but also externally, is I try to tell people that just because we feel that these agents and these AI systems are intelligent doesn't mean that they are. At the end of the day, they're just non-deterministic probabilistic models, right? They're giving the perception or the perceived approach of them being intelligent.
But at the end of the day, all they're doing is they're doing particular things. They're evaluating the pros and cons, a risk-reward function somewhere, and then they are executing the most optimal thing that has to be done. So it's not intelligent. It's just mathematical. So that's the first thing that I do educate. And I tell them that we have not come to the point where you have true artificial generative intelligence, right? Like AGI is like years and years away. And I don't even know if we're going to get there to that point. What we have as humans is the creativity. We understand the context and the intent of things. Working with an agent. So let's say anyone is trying to spin up HNTKI using the hardness. They have all these instructions that they're putting in there. The MD file is like as massive as possible. It still does not understand the inherent context. It probably doesn't understand the entire environmental conditions and the situation and the intent and the goal. Maybe it understands what it's supposed to achieve, but it probably cannot connect all the various points.
And this is where experience from humans comes in. This is where all the scars that we had when we had an incident and we had to be on a bridge. It's like, oh, shit, now I know this is a pattern, and I can recognize that the next time something blows up. Those are things that these agentic chaos do not know. They're looking at it from a very theoretical way. So I relate that to someone who has a lot of book knowledge and trying to now be street smart, and that is not possible. and so while they can process and do things a lot we have the context we have the intent and it's very hard to explain it and the minute you increase the number of instructions or the conditions and all the various things to agents they start they start going crazy they start hallucinating because they don't know which one is really important and as humans we can evaluate conflicting pieces of information in real time and agents cannot at this time so So to people who are feeling really concerned, I really tell them, like, we are not at a point where we can do all this.
We still think that we are giving agency and autonomy to Agentec AI, but we're doing it in small pieces. For me, trust is really important. And I currently, while I do work with Agentec AI day in and day out, they really need to build that trust and reputation for me. And I'm not there yet. and so having these conversations very openly with my teams and with other teams and telling them like we're not here to replace people we're here to make you all reduce the pain that you all are going through the manual friction and convert that into high value activities and improve your quality of life that's truly what we're trying to do and I think when you approach it from a humanistic way a lot of people will understand it and then you turn it around and say okay if you want to evolve here are the various things that you can do to help us on this journey so that way you become much more valuable um for humans the most important thing is to understand what's in it for them right like at the end of the day they do want to um every every person
however uh you know um altruistic or you demonic you are you you do want to have some self-worth and as long as you can show that you're increasing the worth of the individual and making them more valuable and it is in their best interest and it is not something from a corporation perspective. I think humans will understand this and come along into this exciting journey. Every week, our practical AI production calls jump from guests and ads to AI topics and scheduling. And I know this problem all too well. I'm taking notes when I should be listening. That afterwards, I'm piecing together what we decided and who's doing what. That's exactly where I'd put Granola to work. Granola is an AI-powered notebook that makes your meeting notes useful. It captures what happened in your meetings and turns it into clean, structured notes with decision and action items pulled out and made easy to find. For me, that's the appeal.
More attention for the conversation, less time reconstructing it. I'd rather spend that energy making a better show. Once you try it on a first meeting, it's hard to go without. You can try it totally free. Head to granola.ai slash practicalai. That's granola.ai slash practicalai to get your time back. Go to granola.ai slash practicalai and try it on your next call. So tying together the kind of the process that you've described from Invaluable and where you're bringing people on this journey of evolution, of change. And coming back to Scotiabank and some of the hands-on experience that you've had doing that process, could you share maybe a particular task or effort that you and your team have tried to do
where you have kind of brought people into this process? They understand how they fit in. It's not a threat. And they see the productivity that they are capable of with agents and sub-agents. And kind of just give us a glimpse of a real-life situation that you faced on that. Sure. So let me start with people aspect of things and then talk about some of the technologies. Sure. So I also lead the internship program for our two business units here. and we have interns that are coming in. And these are extremely smart people from universities all across Canada, and then we have a couple from the United States as well. Now, all of these people, they ache to just use AI. They still use the word wipe coding, though I've told them to. I've told them they cannot ever use wipe coding as a word or a phrase within our purposes because what I really want is I don't want to have a developer, a coder,
What I want to have is a software engineer, and there's a lot of difference. And I actually walk them through the difference between wide coding and enterprise software development. Very, very different beasts. So one of the things we've done completely is reboot the internship program and make it more like a startup. So bringing the Bay Area to Canada is what we're trying to do. So we create a team. It's a stable team. They have a business problem that they have to solve. it can be both external, so like actual customer-related kind of things, or it can be internal, like a workflow improvement or a process improvement, developer productivity kind of a thing. So let me talk a little bit more about the developer productivity side, and it's very similar to that for the business side as well. So from a developer productivity standpoint, the thing is we want to improve the experience of the developer, right? Like DevX is really, really important for us. But that starts from ideation.
It doesn't start from just the development portion of things. Even today, there are a lot of organizations across the board. How much of our agile they claim they are, like a lot of them do end up being very powerful when it comes to like requirements and design. And now with all these various tools, what we are trying to do is looking at ways by which we can rapidly prototype, get to like minimum viable product as quickly as possible. So we put these interns in this high pressure situation. I have a tech lead and I have a design lead come in and help them. I have a product owner. And so now they're an actual team, a startup. They have to win money or get funding from the venture capital, which is me and a couple of other people, the executives. So there's a competition. They have to pitch and they have to convince us that this is one of the best things that they have to do. And every time they say they want to do all these various things, I ask the question, like, why do you need an agent for this? Why do you need AI? Like, tell me, tell me what is the value agentic AI is providing or generative AI is providing.
And if they like well because everyone else is doing it then I like okay you doing resume development at this point Like if you not actually adding value and you can do something with AI and you can do something with automation go for it So we try to change their mindset and their thinking. And so with the developer experience, what we're doing is we're taking user stories, and these can be very, very non-deterministic ways of creating it. And you need a lot of context and intent. And then using those user stories, now converting it into design and rapid prototyping and iterating with the customer, whoever the customer may be, and getting to a semblance of what it could look like and then giving it to the software engineers. What we're exploring, we're experimenting a lot with that. But that's something that's been very interesting. Like this is a traditional software development lifecycle. And one of the things I'm experimenting with is should the software development lifecycle change with AI? Why are we going through the same steps?
Why is it not looping multiple times with the feedback loops of continuous improvement rather than being like this sequential approach? Like, is there a better way to do it? Can we solve this problem with having like a swarm of agents working together and just day talking and making it happen? Like, these are some things we're exploring right now, especially with developer experience and developer productivity. And, yeah, I don't have the right data to say this is the way to do it, but we definitely are having a lot of fun. When you mentioned the swarming, that's a topic of keen interest to me. And I'm experimenting myself in those areas in terms of, like, having a swarm with a bunch of adversarial roles to kind of keep everything honest, if you will. I'm not sure if that's the right word. But to your point about not VIDE coding and trying to actually have to bring the tools in with the result being high quality enterprise architecture,
as opposed to VIDE coding whatever the agent wanted to go at and having the human experience do that. Do you have, let's put you on the spot a little bit, do you have any insights into that? I think the swarming experiment was a great one. And could you expand a little bit on that? Because I think one of the things we hear a lot from folks out there is, you know, there are a lot of people that are doing it just because everyone else is doing it. And I think benefiting from insights into how to discriminate a good opportunity versus just going with the flow would be really useful to people. Anything that you have to share along that lines? Sure. Sure. So I think this is a good example of where we have shifted the value line, right? So initially we thought development, but truly what we were thinking of is as humans, we want to be the system architects and have a really robust way to design systems
from a systemic view perspective and then hand it over to other people to actually create it. I mean, that's truly what we are starting to do now. So we'll have a swarm of agents. I have, so at MCP Dev Summit recently, we actually shared five agent AI patterns that we have consolidated and we've done a lot of experiments on and we've decided to do. So there are several patterns that we can use now with multi-agent systems. But swarms are not going to solve everything, right? Like there are very specific reasons why you want to use swarms. ideally you want to try to mimic whatever we have. So Congress law, everyone talks about Congress law and you look at the organization and the organizational structure and the software gets built that way. But let's turn it around and say let's look at a collaborative team, a high-performing team. You have people who perform various actions or wear different hats to perform the tasks.
Now convert that into the digital twins. each one of them is there is a so you have like a BA agent a design agent, a testing agent and all these various people so now all of a sudden you have a balanced team that you're using to solve a particular problem and get it together so as a human you create the software design all the specs, everything that is needed like from a holistic perspective and then you give it to the manager of that team and use like a magentic or an orchestrator pattern and say alright make this happen for me. And this is the outcome. This is what I need to do. And this is what I don't want it to happen. So you give positive and negative examples there. That team will go off and do it. It doesn't have to be a swarm. But now you can observe all of them starting to talk to each other. You have shared memory. You have all these various guardrails that you put in. And then you have like a recommendation jury or an evaluation jury that will then evaluate and say, okay, you're getting close to it or you're not getting close to it and helping them guide. It's very
similar to how we work today. And I think starting to mimic that first will then help us get a much better understanding of how we want to move forward. And after that, if we have much more complex problems to divide and conquer, to our panel processing, all those things, then we can start looking at different technologies like swarms, or maybe you have different agile release trends. There's different patterns altogether depending on the problems. And these are not unique problems in this day and age. We have been solving this right, you know, from the beginning of time. Like, if you think about it, these are the things that we've done. Anyone who goes shopping, let's say that you and your significant other need to go to two different stores. You're going to divide and conquer. You're going to have parallel processing. Go do it. Like, we do it every day so subconsciously, but now we have to be very deliberate in our approach when it comes to agentic AI. I think Understanding that and realizing that this is not something new will fundamentally change how you approach these problems, in my mind.
As we start winding up here, I'd really like to get your thoughts, insights, and maybe aspirations into where this is going to go. As you are leading a large enterprise in a major industry in banking to kind of into new territory here, you're truly trailblazing for the enterprise. Could you talk a little bit about where you would like to see things end up? There's so many organizations out there that are trying to navigate the same problems that you're already doing, probably more successfully than a lot of folks out there. And if you could kind of talk about where this is going, where do you expect people to land, both not only from the technology standpoint and the business standpoint, but from the human standpoint as well. I'd love to get your insight. That's a deep one, Chris. Let me try to see if I can make some justice to it.
I think as we start experimenting more and more with authentic AI, It's going to be inevitable for us to try to shoehorn AI into every single thing. We're already seeing it. But I think the most important question to ask is, do we really need AI, especially in this day and age? And organizations that will probably ask that question up front rather than have the FOMO effect and then just go and do, or just because it is shiny right now, we probably will not, those organizations may not succeed really well. It's those who really dig in deep and, you know, introspect and ask themselves, what is the value, where are we going? That's the most important thing. And a lot of times the value is not just monitoring. A lot of it is around the human. So centered around the human and the customer is probably what we already We always need to ground ourselves in because at the end of the day, the reason why organizations exist is to generate or deliver value to someone.
And the next thing is that the world is becoming extremely smaller. We are seeing like within seconds we can transfer money anywhere. We can fly to other places. We can do all kinds of things. Like we need to really recognize that and see how we can reduce the amount of friction we are introducing into these kinds of global activities. Our goal is to deliver value globally as quickly as possible and to as many people or as many systems as possible. So you want to increase your global outreach. the third thing that we want to do is and this is this is this is me as a as a person talking not just from an organization representative of an organization i honestly believe that we as people we probably want to really ask ourselves how are we giving back like how are we improving the quality of life of other people using technology and i think that is going to
be the biggest drivers. It's one of the things that is very fulfilling if you try it. Using technology to improve human quality of life is going to be the ultimate goal, I believe, and it's going to improve the happiness index across the world. And that's something I really, really am aspiring for and driving to. And the fact that I can take technology and help improve people's lives is what drives me day in and day out. That's a beautiful way to finish up right there. I definitely love that. Subscribe to that. And I think often it gets lost as people move forward. So I think that's a great way to leave folks. Galton, thank you so much for joining today. A lot of great insights that you've shared. Looking forward as you continue your journey on this. I'm looking forward to having you back on the show in the future to share more insights as you continue developing this.
Thank you so much. Thank you for having me, Chris. It was a great conversation. Thanks for having me. All right. That's our show for this week. If you haven't checked out our website, head to practicalai.fm and be sure to connect with us on LinkedIn, X, or Blue Sky. You'll see us posting insights related to the latest AI developments, and we would love for you to join the conversation. Thanks to our partner, Prediction Guard, for providing operational support for the show. Check them out at predictionguard.com. Also, thanks to Breakmaster Cylinder for the beats, and to you for listening. That's all for now. But you'll hear from us again next week.
番組の概要欄(原文)
<p>How can enterprises adopt AI and agentic systems while keeping people, business value, and trust at the center?</p><p>Chris is joined by Dr. Gautham Pallapa, Principal Director of AI and Innovation at Scotiabank and author of <em>Invaluable</em>, to explore how organizations can approach AI-driven transformation. They discuss identifying meaningful opportunities for AI in highly regulated industries like banking, rethinking developer productivity, and using multi-agent systems to bring different roles and capabilities together.</p><p>Featuring: </p><ul><li>Dr. Gautham Pallapa – <a href="https://www.linkedin.com/in/gpallapa/">LinkedIn</a></li><li>Chris Benson – <a href="https://chrisbenson.com/">Website</a>, <a href="https://www.linkedin.com/in/chrisbenson">LinkedIn</a>, <a href="https://bsky.app/profile/chrisbenson.bsky.social">Bluesky</a>, <a href="https://github.com/chrisbenson">GitHub</a>, <a href="https://x.com/chrisbenson">X</a></li></ul><p>Links:</p><ul><li><a href="https://www.amazon.com/Invaluable-Define-Execution-Deliver-Outcomes/dp/B0GYHFXRBX/ref=sr_1_1?th=1">Invaluable: Define Value. Align Execution. Deliver Outcomes</a></li></ul><p>Sponsors:</p><ul><li>Framer: The enterprise-grade website builder that lets your team ship faster. Get 30% off at <a href="http://framer.com/practicalai">framer.com/practicalai</a></li><li>Granola: An AI-powered notepad that makes your meeting notes useful. It captures what happens in your meetings and turns it into clean, structured notes with decisions and action items pulled out and made easy to find. You can try it totally free—head to <a href="https://www.granola.ai/?via=practicalai&dub_id=tjZ9lvUGbenMyr6k">granola.ai/PRACTICALAI</a>. </li></ul><p>Resources and Events:</p><ul><li>Register for <a href="https://practicalai.fm/webinars">upcoming webinars here</a>!</li><li>Prior <a href="https://www.youtube.com/playlist?list=PLRpRdFmBfdF6boDgAD3nNaLm8Yx4KN2Vt">Webinars from our partner Prediction Guard</a></li><li><a href="https://midwestaisummit.com/">Midwest AI Summit 2026</a></li></ul><p>Chris’s new book <strong><em>The Art of Swarming</em></strong> will be published soon, written to be the seminal foundational book for the fledgling swarming industry, defining the principles and architecture that enable autonomous agentic UxS swarms with emergent distributed embodied intelligence. You can find out more at his <a href="https://chrisbenson.com/">website</a>: <a href="https://chrisbenson.com/">https://chrisbenson.com</a></p>
出典・この記事について
- 番組
- Practical AI(Daniel Whitenack and Chris Benson)
- 元の回
- Building Human-Centered AI
- 配信日
- 要約作成日
- 公式サイト
- practicalai.show
この要約・解説記事は、元の音声を文字起こしした内容をもとに AI(Claude)が日本語で作成したものです。 固有名詞や数値に誤りを含む場合があるため、正確な内容は元の音声でご確認ください。制作方法について
関連エピソード

Runwayはいかにして AIエージェントでクリエイティブ業務を変革しているか
The Artificial Intelligence Show

ボイスAIエージェントはなぜラボで合格し、実際の通話で失敗するのか:Liam Dunne(Klearcom)
VUX World

すでに機能している企業をどう変革するか――Leidos CIO アレクサンドラ・ギュンター氏に聞く
The AI in Business Podcast

AIネイティブ企業が他社と違う7つのこと
The AI Daily Brief: Artificial Intelligence News and Analysis

連合学習でエージェント型AIを強化する:Flower Labs の Lorenzo Sani 氏に聞く
Practical AI

ICYMI 2026年10月号:AIへの恐れから、AIが可能にする仕事へ
Super Data Science: ML & AI Podcast with Jon Krohn