Leveraging AI4All_ Pathways to Inclusion

healthcare, finance, education, urban planning, but I’m going to only focus for a few for this particular evening. First, access is a multi -layered problem. Good technology by itself does not bring in or include people. By adding AI, you’re automatically not going to include more. The last mile gap is still a problem. You need to be able to focus on connectivity, in skilling, in the interfaces that people use. You must take into account the needs and wants of multiple communities. One of the other key observations that was important for this was understanding the power of the purple economy. The market of assistive tech products for people of persons with disabilities and people with special needs. These are often perceived to be on the margins of our reality, but they are not.

As one of the largest populations of people with disabilities in India, India alone has the potential of 150%. We have $150 billion just in this space. These are people who can purchase. These are people who can access these products. We need to be building for them. It’s not a charitable cause. It’s a simple business proposition. Second, a lot of AI products are stuck in the AI in the pilot stage. You often have a great idea, but you’re not able to execute them. These are for a lot of reasons, but fundamentally, they’re usually around the surrounding system. Like I mentioned, last -mile diffusion, funding, or limited support to be able to scale them up. Third, and this is something you have seen across the summit, language is foundational for enabling inclusion.

Whether it is a banking system which is using a voice AI for credit facilities or an educational AI tutor which you made for a rural village in India, all of them require to be understood in that local context where it’s operating. And this is something you would have seen across the summit in various. Exhibition halls over the past few days. And the last one is institutional capacity. this is a break or it can make a variable as well what you’re going to see is a lot of governments need to build technical expertise in the space of AI we need departments to understand this further this is already happening and once you see this you will see this reflected in procurement standards in technical specifications that these departments are making and this will lead to increasing adoption as a result of these findings then what do we have to suggest at the report there are three interconnected pillars like I mentioned in the beginning design, access and investment anything around AI and inclusion needs to take this into account first, looking at design you need to ensure that you’re embedding inclusion from the start a lot of AI systems are shaped very early and at that stage our recommendation is to have participatory design involve the people as you’re building it out if you’re making something for ASHA workers and you don’t involve them that happens that product is bound to fail in the last minute access.

This is where you have to make sure AI is usable in real world conditions. I know we’re in the AI Impact Summit, but something which you need to know is at least 33 % of the world, that’s 2 .6 billion people, still don’t have access to the internet. So when you’re thinking about building AI tools, you need to take into account those real world contexts, low bandwidth environments, not everyone has high speed internet or a full fledged smartphone. The third is investment. We need to align procurement, capital and incentives. Governments here can play a crucial role by acting as anchor buyers for these kind of products. By embedding standards which reward accessibility and open standards, you will be able to shape market incentives.

Creating these incentives we believe is very important to be able to scale inclusion through AI deployments. The last part of our report, and this is something which is my favorite, are these use cases. And And our report documents a bunch of them. Over the past few days, you would have seen a lot more than we could even account for. I’m just going to focus on two of them, two, three of them, which I really like. One is Shishumapin from Badbani AI. This is a very small tool which allows ASHA workers, frontline community healthcare workers, to take a photo or a video of a newborn baby and get accurate measurements. And this is very important and this is a very simple tool.

It can be used with low internet and can be used offline as well. Second, and you will hear from Augustia soon, I really like the reban glasses. I even tried them out at the Meta stall here. The Be My Eyes feature of that is something which a lot of people with visual impairment are using across the world. Something which helps them navigate the world around them. This is something which Meta has built by involving these people in their design process, involving them as they took decisions. And lastly, this is a shout out to the YesSense. To access app, you may have seen them in installs here. This is a very interesting tool where… you go around, take photos of buildings and physical spaces and understand whether they can be accessed by people with disabilities or not, creating a database which then allows for future greater policymaking in the future.

The crucial thing to note in all of these use cases is that all of these products and tools follow the principles which I talked to you about. They look at design, they have been supported by different government departments and finally they are looking at low resource context environments to be deployed. I am sure at the end of these five days we know that AI is going to expand access and opportunity. The question or doubt really isn’t that. It’s whether ecosystems will now choose to build systems that are required to make this expansion durable, equitable and sustainable. Our report will be out online soon. Thanks. Thanks so much.

Thank you so much Nirmal for those insightful findings. May I request? Now everyone at the panel to please come for a photograph. this is the launch of the report as well so we’ll just take a quick photograph up so if you could come ahead with the report up front Nirmal please the project team who worked on it Yes Thank you very much We’re now going to move to a very interesting part of the event which is hearing from people who actually build these products. To take us through that we have Rutija Paul who’s a partner at Ikigai Law at the panel Rutija over to you.

Thanks Rahil and thank you Nirmal for that wonderful presentation and to the audience for staying back for so long on a Friday evening. So thank you so much. Panelists, incredibly grateful for your time. I know it’s been a very hectic week for all of you. So thank you for taking out the time. And I think Nirmal, he set up a really good context about the three things that we thought were important from our findings. Design, access, and investment. And how do we sort of, you know, use them interchangeably and together to ensure that inclusion is not just a concept but really becomes, you know, really common in the conversations and all of our products. So I’ll start with actually Aragya.

Help us understand how has your product, tell us first about your product and how did you go about designing it, but also how has it enabled access to justice in a country as big as India and all of the issues that it has in the justice system.

Yeah, sure. Firstly, thank you so much for having me here. I’ll probably start by painting a picture of a district court. A lot of you, I’m sure, have been to a district court. by virtue of your profession, but there’s towers and towers of paper everywhere. I’m not a lawyer. The first time I went there, that was the most surprising thing for me. I saw more people writing with typewriters and not computers. And then there were people spending a lot more time looking for the right files than actually going through them and understanding what’s written in them, right? And so when you look at all of these things, it becomes quite clear that justice in these settings is really not a question of law.

It’s become a question of logistics. And that’s where Adalat AI comes in. We build AI and technology to make courts more efficient at a daily and weekly level. And the hope is that when you do this at scale, you can affect the case pendency problem in a rather positive manner. Now, coming to your question of how does AI actually enable access, I think what we are seeing is that there are two tracks. One is the more direct track, and then there is the indirect track as well. When it comes to the direct track, which is how does it enable communities to access justice better, I think there is a huge information darkness problem in the country.

It’s very hard to access judicial information about your cases. If you are in one, what’s going on with it? When is your next date of hearing? And there’s always multiple layers of middlemen that you need to sort of go through to access justice. I think the one use case of AI which we feel is quite safe now is to access information easily. And to that extent, at Adalat AI, we’ve built a WhatsApp chatbot which any citizen can access. They can talk to it in any language that they want. You can just give your name and your PIN code and it’s going to tell you if you have a case. And if you do have a case, what’s going on with it?

When is your next date of hearing? What happened in the previous order? This is not suggestive by any manner. In fact, we discourage any sort of legal advice using AI models at this point. I don’t think. That’s the right use. This is more around. given the information that already exists in the systems behind rather broken, you know, sort of websites can be kind of sort of bridge the last mile access. The more indirect sort of opportunity is by making the institutions of justice be more efficient, which is what we do with our core judicial product. We try to make courts more productive, you know, so writing everything down by hand. And in a courtroom is a big pain point.

Ninety percent of India’s courts don’t have stenographers. So we built a legal transcription tool, which is multilingual. You could understand the legal jargon that lawyers love to use, like rest your decata and whatnot. I’m not exactly a lawyer. It understands Indian accents and dialects. And what we are seeing is that courts that do use technology like this are able to improve judicial productivity two to three X. So if someone was recording two witness depositions per day, now they’re able to record four. to six. Now, when you do this at scale, you can get a lot more done at a daily, weekly level and then hopefully that helps the case pendency problem. We’re also sort of tackling a lot of other different judicial tasks like going through thousands and thousands of pages.

Can we help them navigate it? Can we digitize the entire workflow so that you don’t have to go through a lot of bundles of paper? What we are steering away from at this point is anything that involves legal intelligence. For example, something as simple as summarization too. We don’t think it’s safe enough right now because the summary for a citizen looks very different from the summary that you need for a judge versus a summary for a lawyer. And so that’s something that I would advise everyone to tread with caution on.

Alright, that’s interesting. Thanks. I’m going to come back to you on the aspect of what has been safest to access information. But, Olivier, I wanted to come to you. next. One, very curious to know about Rwanda’s AI scaling hub. And second, Kinyarwanda, if I’m pronouncing it rightly, it’s your go -to language, right? But it’s also a very low resource language. So when you look at using an AI tool based on that language, how has it been? Has it been incredibly difficult? What have been your learnings? And just everything about the hub, please.

Thank you. I hope everyone can hear me. And thank you, first, for having me here. And I’m happy to share. So, as she said, I come from the Rwanda AI scaling hub. And you wonder, she asked me a question when we were out there. She said, why the scaling hub and not just the AI hub? But the whole idea is we, as Rwanda, took the approach of thinking of working on solutions that can be scattered. so that we do not end up just having pilots and we stay in pilot mode, if I can say. So in that case, the AI Scaling Hub has one main mission that has two key pillars. And the mission is really to drive the AI implementation while ensuring that those implementations are aligned with the national priorities for socioeconomic development.

We focus on mainly AI solutions. And then we have two pillars. One is to encourage or accelerate the adoption by basically looking, scanning the world, and find those use cases. Those solutions that have succeeded elsewhere. and see which one inspires that should be brought to Rwanda, adapted to the context of the country, and then implemented to be scared and do the impact in the society. That’s one pillar. The other pillar is now build the ecosystem all around it to make sure that, one, those implementations can be scared and sustained. Two, they open up the door of possibility to actually be able to, I would say, create much more than this. That basically the ecosystem of innovators and all the other institutions and key stakeholders that really needs to make sure that this movement does not stop.

So that’s because we look at AI as, you know, Rwanda as a country have taken the direction of making sure that the country becomes… African hub for AI research and innovation. So that requires now to really go into this thing, and we are the scaling hub because we are also powered to really move as fast as possible in order to show the impact. So that’s in summary what we do, and we have three key sectors that we focus on, but we are not limited on this. Since we talk about the ecosystem, we really drive this whole thing as much as possible in a very agile way. We are the startup -ish type of institution. If I can say it like this, we find a way to make things happen.

So that’s why. And now talking to King Aruanda when it comes to AI solutions, there is something that in India many people may find or take for granted. But which is not somewhere everywhere. when the AI revolution started India had mature DPI which means that the focus has been more to actually implementing the AI already on existing and mature and trusted DPI that are in place it’s not a scenario in many places the Rwandan approach is actually building the plane as we fly it there is a lot of advancement into DPI I would say if I look at it from a technical standpoint everything is at least at 80 % but not necessarily at 100 % it’s more of plugging into things as we go, the DPI stack is being completed but the AI also needs to take off and go into this so there comes basically with that approach that’s why looking at it holistically is key and when it comes to Kinyarwanda definitely Rwanda is a small country compared to India in terms of size and in terms of people but it’s also a country with a high density population when you look at the way it is and the entire population speak one language which is Kinyarwanda as one of the languages that other we speak, basically which means that actually a solution for it to be adopted, it needs to be speaking Kinyarwanda and AI did not originate in Rwanda so AI does not speak Kinyarwanda originally so as we build our plane, there is the time of also now building the models, building the data set for the language be it the text be it the voice in order to get to perfection so we are doing this as we go and there is improvement every day.

I think that a couple of years from now, we have, I would say, a full stack data set of Kinyarwanda language that can now operate all this. But even right now, we are doing things. That’s the approach.

That’s very fascinating. I think building a plane as you fly is going to stay with me. Thank you for that. I’m going to come to Archana next. I’m just going to pivot a little to a B2B conversation. You help businesses across the spectrum, be it healthcare, BFSI, education, scale up and transform digitally. What does access and inclusion mean in these rooms? How is it that you really convince your clients that inclusion and even access needs to be really embedded in the first thought of your transformation journey?

Thanks for that question, Rituja, and thanks for having me here on this panel. I’m going to take three examples. Recent ones. The first example, we were working with a humanitarian agency which deals with refugee crisis. So they had approached us to develop an AI solution for the field workers who operate on the field when a refugee crisis is happening to look at real -time where should the aid go. Because when refugee crisis happens, assume a blast happens, something happens, there’s a lot of aid that flows in. But is it reaching the right places? For that, you need to process real -time information. For that, you need to look at what is happening there on the ground, which you could be getting bits and pieces from the representatives who are there.

You need to be able to access information that’s flowing around the media. So there’s a lot of data crunching intelligence that needs to be baked in. And typically before AI, a lot of this was relied on telephone calls. That’s manually done. with AI this is something which helps but in this kind of situation most of the time your internet doesn’t work most of the time the connectivities are down because in this situation the connections go away and your AI still has to work you cannot say that I don’t know where to give the aid because my cloud connection went down or my net didn’t work or the connection was down by the government at that point in time so when you design an AI system like this you need to be able to figure out what needs to work offline what should work online where to bring in how to architect it and that becomes crucial so that’s first example where AI needs to be accessible inclusive by design I’ll take a second example so second example a global bank one of the largest bank in the world approached us and their request was, hey, I have a lot of financial literacy videos on my website.

Typically, those are in English and from an accessibility standpoint, there are some captions in English which come in but those don’t necessarily serve hearing impaired because for them, their first sign language, first language is sign language, not English. What can AI do here, right? So the question was, can we use for a little bit technical terms like the vision LLMs and some of the processes that are out there, technology, to create videos which probably were not accessible initially to a large set of population and make it accessible. So again here, something existed but you are using AI to put and add a wrapper on top of it. of it. So you are not accessible by design in this case, but you are trying to use AI to make it accessible.

Whereas in the first case it was accessible by design. And let me take a third case, which I was getting into quite a bit of heated conversation with the CTO of that insurance company, where they did a small POC with AI, where it was a conversational thing. Somebody calls at the insurance help desk and the AI kind of response on what queries the person has called in. And of course, like in all demos and POCs do, it worked beautifully. And the second question was, hey, let’s scale it up. And immediately the person with whom we were working, the CTO said, you know what, let’s do it in English for phase one. And let’s look at other languages later.

Now, my argument was that if you do it this way, most of the folks who are calling you are the ones who speak Hindi because you are operating in that region. If you don’t do that, you are alienating 70 % of the people and your customers. And why are you then putting this bot for? Why are you even attempting it, right? And their answer to that was, you know what, I have to show ROI from AI. And I have to show that quickly. And hence, hence, please go and still do the English one first. Let’s look at Hindi in Phase 2. Right? And you can imagine what kind of heated conversations I was trying to explain them. That’s not the right approach.

You need to be thinking of Hindi right from the start. Because if you do this, it will work beautifully in demo because it was all English. It was a sample data set with which you were working. It may still work in your Phase 1 a little bit. but in phase 2 it’s going to fail miserably and it will bite you even bad when it comes and fails at that point in time but it was a hard conversation we finally convinced them but to get to that there was a lot of education that’s needed so what I’m saying is if you look at these 3 examples where in certain cases due to the virtue of the business that humanitarian agency was you had to be accessible by design in the second case because it made good business sense the company said make it accessible whatever financial solutions we have whereas in the third case it was a very difficult conversation on accessibility because somebody wanted to prove a point to their management that AI gives the ROI which is there now if I look at various cases where most of the corporates are today of the businesses which actually are dealing with this economy and responsible for bringing AI out there, most of them are still hovering in the bucket three, which is the last one, where it is still not inclusive by design, still they feel, I have a POC, I can scale it up without being as inclusive with the data, with languages, with other things, and I can do that in later phases.

So this was the story till the entire last year. This year, and thanks to the summit and more and more forums like these, businesses are appreciating the fact that if they don’t do inclusive by design, they are leaving money on the table, and it’s just plain, smart, good business. So I think now the conversations in the boardrooms and the rooms and in corporates are shifting, where the question is not necessarily, get me the ROI and prove and show that AI works. but make AI sustainable and working for me for a long term, which means I have to be inclusive. So that’s what I would say.

That’s wonderful. I think, I mean, kudos to the summit. It certainly made the conversation inclusive, really common and very boardroom, entered into the boardroom finally. So I think that’s a good takeaway. I think moving from the third bucket to now, Agastya, I wanted to come to you to just help us understand the way we, what we’ve seen in the research findings of our report has been that in many ways, AI is a force multiplier. It is going to enable at a much faster, at a much larger scale, right? So tell us a little more about the, at the back end of the design team in Meta, how when you look at designing a particular device, what are the instructions you give your team that this is what you need to follow A, B, C, D, so that the device you’re creating is definitely inclusive.

It respects the idea of the people that it’s going to be useful for.

which are the divots on sidewalks that allow wheeled devices to transition from a sidewalk to a street to cross the street, they are ubiquitous in the United States due to regulatory pressure to protect the rights of people with disabilities that use wheelchairs. But anyone who’s encountered them while using a pram or stroller or a trolley, shopping cart, or luggage has benefited. They just make cities better. And so taking an extra step and thinking holistically rather than just being pressured by regulation, which of course is still an important component, is critical to making the end result good. I don’t think anyone’s perfect, but I’m doing my best to instill this mindset within Meta.

All right. I mean, yes, I don’t think anyone’s perfect, and we’re all trying our best. It’s a good takeaway from the summit and everything that we’ve learned from here. I wanted to pivot to the conversation around investments and just, you know, how do you make inclusion and creating sustainable pathways? For inclusive AI, really, you know, in the context of India. or even globally for that matter. And I first wanted to actually come to Olivier again. Could you help us, just give us some idea of how did you go about making the procurement policy, which I understand is very innovation -friendly, for the national AI strategy? What were the considerations that went behind it, and how have you seen it pan out on ground so far?

That’s a good question. So, Remy, paint a picture a little bit. So you see the whole journey to get to there. So procurement is normally most seen in public sector. And, you know, we are in a country where accountability is something expected from everyone. And when it comes to public funds, it’s even to another level. which leads that the classic procurement, if I can say, it takes a lot of time because in order to really avoid any way of any conflict of interest in the process but when it comes to the ICT space most innovation products look at the journey, he’s talking about about graphic user interface and you know the touch screens, look at the social media, he’s from meta you know, Facebook before before it become meta but just if you look at the journey you will see that normally into this space there is a change, there is a new thing every three years 2023 we were talking more about DPI, DPG and people were even having hard time to differentiate the two And now, three years later, we are talking more about AI as if it’s a new thing, but it’s basically the large language models that are new because of the revolution of social media that gets a lot of data sets and creates something that we can interact with.

If you go into the all -time procurement, you can try to buy 10 phones, and it takes you three years, which means basically by the time you follow this process, things have changed. You may have the right process, but not the right product because things have changed. That’s how the idea of now having the public procurement for innovation concept. Which was put in there, let’s say, in some space to some categories. Let’s consider a way where instead of really going through the classical time, how about… we bring together key players, potential institutions that can deliver to XYZ solution that we see is needed. And then give them a chance. That is a bit like, you know, they compete to see the best, who can deliver to this, and then they are empowered to do this.

So we go more into the agile mode of having these, you know, small step development along the way that can adapt to the change instead of waiting for that long process and end up getting a product that is no longer relevant to the market or to what we need to respond. Or maybe it’s relevant, but it’s way too old. So imagine trying to get, now we are at iPhone, what, 17? You know, how many times have we seen these basically these evolutions? So think about the process that started five years ago. It works for building roads. but not necessarily for technology projects. So that’s a bit of the picture of how we end up to this.

That’s interesting. Even for us in India, it’s been that oftentimes the law and the policies is playing catch up with the tech. So you really need to find a creative way of finding solutions that you can smartly look at regulating the emerging tech. Aragi, I know you have a lot of thoughts on this one, especially around the procurement rules and how do courts adopt your product. Please do come in. We’d love to hear more about how do you think the existing procurement rules have shaped the way you’ve been able to access the courts and deploy your product there? And what do you think needs to change so that it’s faster and more usable for the courts?

Yeah, I think I’ll take a more solution -oriented. We could talk a lot about the problems of policy playing catch up with tech, but I’ll take a rather solution -oriented approach to how we… We’ve worked with the courts at Adalati. So when we started Adalat AI, which is about two years back, AI was very new. Courts are still, you know, working to adopt generic software technology. And so AI is extremely new, right? I think a couple of things worked really well for us. Number one was to build painkillers before vitamins when it comes to solutions. So we actually went for a very big pain in courts, which is judges are having to write everything down by hand.

And so when we say that, hey, there is this new technology, but it solves a really big pain point of yours. This is not a vitamin. This is something that you are all struggling with. There are a lot more open to adopting technology. But in terms of the creative solution around procurement, I think I want to emphasize that nonprofits as a pathway to creativity, creating impact are highly underrated. specifically in the space of justice and law. You know, there are all these non -profits that work with education and with healthcare to support doctors and teachers, but not enough non -profits doing this to support our court staff and justices in the country. And so, Adalat AI is exactly that.

Now, what do I mean by non -profit as a vehicle? Being non -profit helped us align incentive with the courts better. It automatically took away a lot of the stress around, oh, what are they going to do with my data? Are they going to profile the judges? It took away a lot of stress around, okay, are they going to charge me? Where am I going to, how am I going to evaluate the new technology? Now, so this helped us get into courts initially. And, you know, within two years, we are now in nine Indian states. We are in one out of every five courts in the country. And as of a historic mandate by Kerala, it actually became, mandatory to use Adalat AI in every courtroom in the state to record witness depositions.

It’s absolutely not allowed to do this by hand. And I do think that, you know, sort of this impact vessel vehicle really help us do that. In terms of sort of the other side, which is that at the end of the day, eventually courts and all institutions need RFPs. They need to sanction budgets and they need to sort of make sure that they pick the right player for it. Being a non -profit, some of the ways in which we are seeing we are able to influence this process is that now that they’ve been able to work with us, they have a lot more experience of what it means to scale these products. Their tech teams have a lot more experience of working with us in knowing what do they actually need out of these products.

And so they have a lot more in terms of ideas of how to draft the RFPs. And so I think that’s the other big benefits. If it’s that coming from being a non -profit, you know, all these non -profits in the ecosystem, they’re able to help these institutions. sort of design better RFPs when they actually do go and procure solutions.

Right. That’s interesting. I love the Kerala example. I wish to see that happening across all states sooner in the country. But I wanted to now move to Agastya. I know that Meta Ray -Ban Glasses represent a significant investment for Meta in terms of the AI -powered hardware that you’ve created, right? From the inside, help us understand how do investment priorities shape the design journey for that product?

That’s a good question. And I think in reality, sometimes the plan or the intent doesn’t necessarily match with where things land. For example, the Ray -Ban Stories, which were the first iteration of smart glasses we shipped, they were great. They had some really cool features. When we built them, we initially thought that the use cases would be around taking pictures and audio would just be used for making phone calls. While myself and a couple other engineers were doing hackathon projects, combining multimodal AI to help blind and low vision people, this was before the AI hype had caught the zeitgeist of the industry. And then the next iteration, the big focus we put was we found that people were using them for music much more than we expected.

And so we thought the biggest use case, the biggest investment would be on making the speakers better for Ray -Ban Meta version 1. And we did that, and the music and audio quality was much better. But you’ll notice something missing from the product plan that I mentioned for both of those products, AI, which is now not only front and center, but it’s literally how we market these glasses. They’re AI glasses. I say this not to drive cynicism, but nobody has a crystal ball. And so I think the key thing is learning to be nimble and understand the direction things are going and being able to jump on trends versus being too fixated on what the original plan was and maybe giving a sunk cost fallacy.

I love the painkiller. Vitamin analogy. And maybe adding to that. the really important thing to do is to avoid the temptation of eating the candy before either of those two. That’s my take on it.

That’s interesting. Thanks so much. Arshna, I wanted to come to you. Same question, and I think you touched upon it in your earlier remarks that the executive wanted to show ROI. So really the question is when you have these routine discussions in boardrooms with your enterprise clients, did you start with positioning inclusion as a CSR initiative or just a good -to -have thing in your strategy, or has that shift significantly changed? I mean, of course, barring the summit and the two months of change in thinking, but in the past, how did the pitch start for you, and how was the reaction really like from the leaders?

If you, and this is my personal view based on what I’ve seen, and my experience, If you position inclusion as a CSR initiative, you are also going to get budgets which match the CSR initiatives, which don’t necessarily translate to good products or make good economic sense. So that never works. Don’t do it. That’s first. The second is that when you are positioning these kind of conversations, remember that in corporate world or any business for that matter, it’s always a trade off. How much you are willing to spend versus the returns that you are getting. Now, if you want to be it more and more inclusive, especially in an AI context, you can do that. If you have more and more diverse data sets feeding into that.

Do those exist today? at a cost which is palatable to all enterprises? The answer is no. So first thing what enterprises look at is great, I want to be inclusive, nobody wants to say no. But if they don’t have those data sets or if the cost of getting those data sets or cost of cleaning those data sets to make it more inclusive is going to be much higher. So typically in AI we say $1 spent on AI, you have to spend $3 on data. So if that’s the kind of economics you are dealing with, there is definitely going to be a point where the company says inclusion is going to come later because economically it stops becoming as viable to them.

Now if you look at the inflection point that AI is there today, it’s hyped up. it’s yet to show tangible outcomes across all the sectors. Yes, it’s shown great promises and results in some, but has it universally shown those promises? No, we are yet to see those. So when you are dealing with clients which are in those areas where they are yet to see those, you will see inclusion taking a backseat, not because of the intent, but because of the cost in certain cases. Whereas everybody realizes that inclusion is plain good business, but those trade -offs is what they look at. Increasingly, with the data being made more accessible, governments taking initiative. In fact, at India, we have AI Kosh, which the government of India has put in where you get diverse data sets of India and you can feed in those data.

And you can use those data sets to make your AI systems more inclusive. more tuned to custom local traditions, you will see the cost of this implementation going to come down. So economies of scale kicks in. Moment that happens, you will automatically see corporates and companies adopting this because now while there always was intent, now that intent is also becoming financially viable for them. So I would say it’s a combination of these kind of different facets which play together when certain decisions get made.

That’s helpful to know that CSR is not the go -to route to see, but a bunch of things that determine the decision -making. I think in the interest of time, I’m going to move to the last segment of our panel discussion and my favorite, which is design. So I think I’m going to first come to Agastya again. Tell us about how can AI devices really drive accessibility first innovation? And I remember reading this at the Metastore, as well, earlier in the… weeks. So just help us understand the company thoughts behind it and how have you gone about executing it across different devices, including the glasses?

Sure, thank you. Accessible design is good design. Universal design is good design. I think opening with that mindset that if you build things in an inclusive way, you make the product better for everyone, people with and without disabilities. I think that’s the critical factor. I think the second thing tied into that is the notion of nothing about us without us. On this panel, we discussed that a model is only as good as its data set. The same is true for a development team, for an organization. So I think it’s critical to hire people from all sorts of different backgrounds, not be stuck in your own bubble because you’re building products for people with all sorts of backgrounds.

It’s not just good karma. It’s not just charity. It’s good business. So I think those are kind of the two philosophies I’d push on, is that hammer home that innovation actually is seeded by accessibility. There’s so many innovations that started from accessibility efforts. The flatbed scanner, text -to -speech synthesis, OCR, these started as efforts to read books for blind people. They didn’t start as just industry -wide things. And yet here we are. So I think working with your leadership teams to call those examples out, show concrete examples of how things get better, and ensure that you are building with everyone.

That’s incredible. Thanks, thanks. I know in the interest of time, I’m just going to quickly come to all three of our panelists to help me understand, of course, obviously with your own case study, but, Olivia, one case study from your country that you think the design aspect of it where, you know, from the very initial you’ve looked at, and inclusion has just been visible, and that’s helped in many ways. So just give us one example. and the same thing for you, maybe perhaps from the jury that you looked at on AI for her, that would be helpful. So, Olivier, then again, then Archana.

All right. A quick one. We don’t have so many AI -powered solutions that are there, but just an example, we are working on an AI -powered advisory solution for agriculture. And right from the beginning, we need to think about the end user before we even think about the technology, because what AI is doing to us actually makes the tech easy. You know, a chat bot and a robot, it’s like even a code bot can make the code. But the end user now, in this case, we are talking about a smallholder farmer who does not use a software. He doesn’t use a smartphone, but uses a future phone, who may be in a place where the connectivity is shaky.

but who only speak Kinyarwanda. So going from that angle now basically there is that inclusivity right in the stage so that if we can deliver to this then the technology can work. That’s one example I can set in there and a couple months from now I should tell more success story because we are beginning into now those solutions to scare.

That’s good. I look forward to a couple of more months and then some more case studies from your country. Raghav, do you want to go next?

Yeah, I think I’ll talk about two things. Number one is design and design of product and the second is and I want to contrast this is design of the intervention itself the entire solution and with respect to the problem that you’re trying to solve at Adalat AI with respect to design of the product there’s one thing that we’ve done from the start that has helped us we force our engineers, designers, everyone to go to court sit with judges, show them the designs, get an in -person approval from them before any piece of code is written, before they come back and touch their laptops, right? And that’s one thing that has helped us tremendously in being able to make sure that design is extremely inclusive.

The second is when it comes to design of intervention itself, it’s not enough to build technology. You know, we build transcription solutions, but if the judge doesn’t understand that they need to turn on the mic at the podium when they’re kind of dictating, then the mic just becomes a very expensive paperweight, right? It’s of no good use. And so we do extensive trainings. In fact, we have something called the Adalat AI Academy. As part of that, we go to courts, we teach them how to use the technology, and we had a very interesting insight. We were trying to teach them AI. But what we learned was a lot of judges don’t know how to update their Chrome browser.

And so that helped us then understand what exactly is needed to drive that intervention forward and make sure that impact is actually realized on ground. And I think now a lot of Adalat AI Academy has become a part of the official curriculum of becoming a judge in a lot of states. And so that’s kind of helped a lot in terms of design.

That’s great. I think moving into the curriculum always helps that you’re planting the seeds early on for the training. Archana, the last word.

I’ll be real quick. So as part of the jury for AI by Her came across several startups, which were, of course, led by women and conceptualized and supported with AI. One of the startups which kind of stuck with me is a startup in fashion tech. And the interesting piece was that that startup. Helps the designers to show and envision how the finished product could look like. and what it does is not just show it so that you can reduce the time it would take to develop certain samples and then discard them so it’s not just sustainable fashion and sustainable designing, but it also shows in different shapes and sizes. So that makes it even more better and inclusive.

So some of these kind of things is what I found in the solutions which were there in AI by Her, which kind of makes you think that, yes, these are truly being sustainable and inclusive by design.

That’s wonderful. All right, do we have time for questions? No? All right, cool. So we’re going to… Sorry about that, audience. Thank you so much. But probably we can catch all of the panelists once you’re done with the last segment. Thank you so much. Thank you. Thank you so much for a very insightful panel. I think everyone who stayed back has at least the last hour has been more informative as well and we were left with something. from all of you. So thank you so much. Please do catch the panelist. Thank you everyone for staying here. I know it’s been a long week. This is the last session at the AI Impact Summit, so just thank you all for being here.

And a big shout out to Metta who’s partnered with us for this project, so thank you for your continued support and we look forward to engaging further work. Thank you all. We do have some mementos from the India AI Summit for all the participants. So Rutuja, if you would please give them Yes, there is. Thank you.