So here, first of all, I'm Adri here. I am your speaker today to speak about how do we actually really convert our productivity into business outcomes.This is a topic that is really passionate for me. Essentially, as I walk through it currently in in multiple organizations. I've been currently, I'm at BlackRock. Before that, I was at Salesforce and at PayPal and at Amazon where we had used productivity from different lenses. So here is, the I'll I'll go ahead and I'll talk through, and then I'll be opening up for questions at the end. And then, you know, hopefully, it's gonna be open in a discussion. And I thank the opportunity, for women's in tech for me to give this opportunity to share my experience here.

So, essentially, we all know that continuous amount of day in and day out, the push that we get that use AI, use AI, use AI in your daily jobs because you can make it faster. Your productivity can go higher, and essentially, you can accomplish more. I mean, I think that's a trap that we are actually right now in, where essentially we are looking at it from the perspective of, okay, how do I make my daily daily minute tasks faster and faster and faster and better? But at the same time, we need to keep in the the entire, the entire journey to achieving business outcome or essentially achieving a client outcome or achieving a customer revenue growth is not completely tied to an productivity gain. And this is where I will actually go talk talk through in this in this particular, presentation today. So, essentially, most organizations are really stuck in this activity trap because they're all looking at it from AI from a micro lens.

So if you look at it, right, we are driving high adoption essentially through every different business function. So be it could be designers, be it quality, be it developers, be it people who are actually providing business value, like writing BRDs or people who are doing the contractual negotiations, or it could be the customer in operations personnel. Everybody is actually using AI in some way or the other. And I actually help AI helped me also to prepare this particular presentation. So, yes, we are day in and day out using AI. But at the same time, you see that adoption doesn't really reflect into a business outcome, and it doesn't really reflect into an impact. What we are seeing is that while we have, like, really high adoption in the different different pockets of different functions, but what it doesn't give us is consistent usage, and it doesn't give us a direction in which the organizations are steering towards standardizing and making sure that particular little little gains are really affecting the revenue or customer experience in the end.

So how do we make sure that these organizations or all our organizations are how we can contribute making sure that the daily AI that we are using in these pockets are actually getting some sort of an ROI. Right? So, here, like, going back, you know, so what did I, like, think think about? Why does ROI really break down as I talked about scale workflows? Like, for example, we need to rethink the comply complete way a certain amount of, you know, workflow is being actually designed. So I'll give you a simple example. I recently read in a book. Actually, it's, it's an interesting book. It's by Sandeep Paul Choudhary. He I I read his books on platform a lot. He talks about a simple analogy. When earlier, when we used to move, I I like the story there. What he talks about is that we'll go back in history when we look at railroads itself. Before that, what was the main reason?

Like, most people used to move logging across from one state to another state using streams. You would see logs actually going in water streams, and that's how the transportation was being done of logs. Later on, when actually the war, the industry slowly, we got into railroads. And as the railroads kind of evolved and then more more railroads were continuously being built, what happened is that now it became moving some goods or logs from point a to point b became much, much more faster. But that actually helped people rethink not just the movement and the productivity of the logs, but then the biggest thing was that the economies were completely changed. Because then the people wouldn't rely on a local economy anymore to supply whatever they need to supply. At the same time, it actually reflected back into the time zone.

You know, what happens is that now people had to go back and look at how do I summarize time zones Because I need to make sure that point a to point b and I get I I get my goods and so and so time here. So make sure that, okay, I understand now I need to tie my supply and chain to a time constraint. So if you look at it with the smallest micro change that was happening is for chain for the train railroads being built for moving things around, but that kind of made a bigger impact into how time is actually being thought about. So if you look at AI itself, we need to go really step back and and and look at it not with the perspective of, oh, am I actually resolve solving for individual workflows or am I solving for a bigger picture here? So, essentially, stepping out of being the activity obsession as it is is one of the, you know, ways that we can move forward. And then we look at it from the perspective of business outcome of measure. So going further, you know, I mean, this is again reemphasizing that. Right?

That, yes, we have a higher adoption of the initial tool. We are driving productivity. We are measuring it through, you know, how many tokens are being consumed and which particular function consumes more tokens. How but then what is more important is also looking at how are they making an impact and then looking at what is the traditional workflow that is actually being present that we are redefining. And that's the core of transformation, and that's where really, like, 80% of the organizations are really struggling there. And then that directly, you know, impacts into a financial impact. Like, that really reflects into what is a margin and what is the revenue and the throughput that I would I kind of translate my AI work into. And in the end of the day, we're all there to solve problems. We have to save and and improve customer experiences. And that's where the financial impact comes in.

Am I improving financial experiences, or am I getting something whole workflow redesigned such that I'm actually doing more with, you know, what I have available? So as I, you know, re you know, reiterating it, the code challenge lies in moving beyond, like, isolated activities that we do and automating the entire set of tasks and looking at the entire redesign of the systems. So, as, you know, again, like, if you look at it from the perspective of task automation alone, what are we looking at? We are replacing small, small, isolated tasks, like how a business analyst is writing their particular, you know, requirements document or how a contract, you know, reviewer is looking at the contractual documents, and how is it happening that a developer is actually developing within using their particular code gen AI tools to improve their productivity in their particular environment.

What we are losing in the picture is how do we standardize all this. Right? How do we standardize all this is that is is where actually the orchestration value comes in. That is a repetitively how do we make sure that the repetitiveness of the work that we are doing, like, all contracts are being reviewed in a same standardized way. So what is a template we would use? Code is being written to a certain amount of enterprise guidelines and enterprise architecture principles, and we standardize it completely where the code is written and comprehended and reviewed in the same way. That's where you actually the main value comes in when you orchestrate that in a large in a large transformative way.

Then you also look into your workflows, where in your continuous workflow is it completely we get eliminated. Right? So that's where your your pure task automation is not alone. So, I mean, I know there is a lot of, you know, chatter and discussion about where is the human value here. Where is the human I mean, this is particular insight I got directly from Anthropic. They said value is in a combination of humans and AI together. I know we are we are always looking at, oh, is AI going to replace my job? AI is gonna do this. AI is gonna do that. Yes. If it is a job that is of low value, that is a repetitive, that can be standardized, yes, it could. Only only it's going to be only effective if you are orchestrating the entire workflow redesign.

And you are actually are, as humans, providing oversight to exactly that AI agent's behavior. Because what we can act we cannot completely influence what our jobs or what our, you know, individual jobs that AI could be taking over within the value ecosystem, but we can really emphasize how effective that is by providing our oversight and judgment. So, essentially, we are replacing our roles itself from being just executors to over being, like, overseers, like, they're giving oversight. At the same time, we are providing guard rates to these AI agents and then what it should and what it cannot do and what are the decisions boundaries that it it needs to be with them. And that's where we look into, you know, judgment comes in and then orchestration comes in and orchestration is both human and AI completely working together. So, I I kind of you know, this is a a a a hodgepodge of a multiple a multiple, research papers.

I reviewed the HPR, McKinsey, this research, Anthropic, and OpenAI And was essentially the same thing. Like, if you look at it earlier, the economic adoption patterns are completely changing. So in the past, it was like doing one, two, three, four jobs that you would give it queries and do jobs. It just shifting completely into workflow orchestration because that's where the effectiveness is. And you see some of the industry focus of the AI is sitting in finance, coding, and customer operations right now. That has been the primary focus. And AI is get unless you embed it into all of the native flows, like in a sense your workflow itself, I, you know, pretty much the there's a lot of literature out there and research done saying that it's not really effective. Right?

So and if you look at it, like, if you look at the cross industrial business impact, the impact, if you look at the metrics earlier, it's not tied to complete productivity, but it's also tied to how actually I'm effectively making, like, credit decisions. Am I able to provide immediate customer satisfaction at a heightened level because I'm able to respond back to my customers quickly? So we got to look at it beyond the micro and the macro effect that we are producing with whatever business impact we're driving with the workflow orchestration. So at the same level thing, like, you know, the customer ops function is not simply, like, answering you and they have been on those bots which kind of really frustrate me because they just keep you going in the loop. But then what becomes much more effective is that how can they properly triage as well as direct your queries to a live customer agent depending upon the situation. Right? So that's a workflow orchestration. So, essentially, reiterating, you've got to look into business workflows. And most importantly, we need to anchor on governance right now.

We need to look into what are the different AI operating leaders we have and where we can embed into the decision making process in the workflows. Right? So, I mean, I we talked about it the earlier, the benchmarks and success patterns have been that, you know, decision velocity increase. That's another new benchmark. That's how you drive drive outcomes. Right? You look into a new benchmark. Mark. You look into a new metric that you got to drive rather than just driving productivity alone. So for health care, it is like, am I actually making my diagnostics faster? Is is it that, you know, for software, it's going to be not only release velocities, like, quality going up up to the effectors emission time as well as quality essentially, you know, going up and then your incidence going down.

Right? So similarly, like, you know, as you look at it, the the real playbook is redefine your workflows, look at your human and AI orchestration elements, figure out where is your harness that you need to, you know, work upon, what is your complete AI operating harness, and then how do you make sure you're mentioning your outcomes properly.

So this is another interesting thing. Actually, I follow, she is Lea Benzor is on LinkedIn. I follow some of her articles, and she really put it very well. Like, we have to architect the AI first operating model. And then she talks about it, how we we kind of have to move from orchestration to leverage. What does the mean by leverage? Leverage is identifying the right opportunities in which you can drive completely behavior changes of people as well as the systematic workflow changes in which you want to drive a a big business outcome. It could be a client outcome or a large impact. Right? So it's it's very theoretically speaking, but if you put it in a real a real you know, if you drill down it, we got to look at our operating principles of an AI.

For making an effective business outcome, you know, and and driving that, you need to make sure you have the right operating model. And right operating model includes the platform, the people, as well as AI LLM elements behind LLM models behind it is only a smaller piece. So you look at this and we've got to focus on the layer one, which is a governance, having a clear vision on what exactly you want to drive. Do we and then what is your harnesses that you want to use? Because you want to make sure the that AI writes code, AI reviews code, AI deploys code, where do we want to make sure you have the right card rates in it? Do we want to make sure there is a human in the loop and what is the data and context?

Context engineering is completely a newer kind of a whole function, and and security engineering is really evolving very, very rapidly rapidly right now as the models really advance because we gotta look at it from security and compliance lens. Am I actually, in the danger of leaking my PI echo the PI information outside? Am I actually in the danger of making sure my known vulnerabilities are going to get good exploited immediately right now? So having a security guard raise is also important and observability. How can you distinguish between human audit versus agent? We need to come up with complete distinction and different audit, permission and policies, identity completely for agents and humans separately. Then we have to look at how do we define our thoughts.

Earlier, it was domain level, you know, domain level, functional level division where you have software engineers, quality engineers, or you would have your product people. But I think a successful squad right now is moving towards a cross functional squad. A smaller squad which has everyone in it and which has actually driving one impact. So and if you look at the AI first, Raya, it's like the pods. We have to think about the pods. We have to think about the harness, and we have think about really the strategy where you want to optimize so that we are not having collision. Collision is like duplication of work or collision with security or compliance guardrails, you know, internal collision itself. So designing your pod, meaning, as I mentioned, right, we have to be okay to not just having, like, a a silo domain where some that there is, like, a sequential handoff that happens.

Even in an agile environment, we do have some sort of handoffs. But now we are having a pod which has everyone together in it. A developer, DevOps, scrum master, all of it together. And that's a really successful part to deliver an outcome that you want. So here we talk about the strategy layer. You think about where which AI systems that we need to really optimize the strategy, the sense where exactly are the outcomes you want to drive, what exactly are your leverage points that you want to infuse in. And harnesses essentially that governs the whole base. Right? It's like shared shared governance guidelines, shared prompt, workflows, architecture, or, you know, having the audit and security guidelines. Now going back to how we're shifting the metric paradigm, we're shifting it from focusing on outcomes and move the needle. Forget counting prongs. Forget counting tokens. Forget counting, you know, just that marginal revenue or marginal productivity.

We need to look at the larger revenue growth, larger margin expansion, what they do for customer experience, and then cost savings. Right? The evils we have to pull here is redefining the harness, enabling the teams to think beyond their own, you know, work that they're doing and think big about the workflow, then they're prioritizing these. So the newer economics of measurement is instead of looking at the older way of doing it, what is cost for cognitive task? This is, you know, some of these guidance came from anthropic. You get the throughput. You get the human AI orchestration. You get your delegation that you are doing. What is your faster decision that you're making, and what is your cognitive cycle time? Like, in the in the end of the day, are you moving from insights to action quickly? Right?

And the definite the leadership in this whole AI era can change completely from management to designing these harnesses, designing the judgment points, designing how to frame the problems, and looking at how to adopt your metrics. Right? And, essentially, these are my basic references. I know we are at time basic and then, you know, just to make sure productivity alone does not create enterprise value. I know. I didn't go fast. I don't have any questions here. Yeah. Get the reference in slides. There's a McKinsey, Anthropic, OpenAI study, and Thank you for joining, and thank you, Women in Tech, for this opportunity. Reach me out on LinkedIn if anybody wants to continue discussion further, and have a great day.