Detect Intent, Then Tailor Every Screen to the Person
Episode Summary
Al Shanmugam runs product for Dish Network and Sling TV at EchoStar, which means he owns what a viewer sees, how they search, the notifications they receive and the advertising they are served. He describes AI in that domain as three things: automation, personalization and intent detection. The old approach was segment based, grouping customers together and sending the whole group one offer. The new one learns about a person rather than a cohort, and works out what someone wants without being told, so the same page can look different to two people. Agents close the loop by acting on that signal and reaching back out on whichever channel the customer actually uses, inside guardrails and brand validation that stop a model speaking for the company unchecked. He measures all of it on human involvement removed, churn and conversion. His forecast is that integration layers and interim databases disappear while systems of record stay.
Key takeaways
He owns four surfaces at EchoStar and names them himself: what the customer sees in recommendations, how the customer searches and interacts with the platform, the notifications going out by email, text and push, and the advertising, including whether it is contextual and whether it is useful
His framing of AI in a product organization is three things, in this order: automation, personalization and intent detection. He notes at the end of the same answer that the three sit at different maturity levels inside his organization
Automation is different now because the old version ran on rules and this one is contextual, he says. Software and systems used to do that job, and what he describes in their place improves continuously from the mistakes being made
His own definition of AI, offered as a working description rather than a term of art, is a self understanding or goal-driven system
Segment-based personalization is the thing being replaced. The old pattern was to decide that a group of customers likes something, put them all in one segment and send that segment the offer
What replaces it is one-to-one, and he calls it hyper-personalization: AI learning about a person rather than a cohort, because the interest that matters is often nuanced
Intent detection is the third leg and the one that makes personalization act rather than describe. If he knows what you are looking for without your saying it, he can customize the product for you and tailor every screen
The endpoint he names is that a single dot com page renders differently for two different people, and he says that is where this is going rather than where it already is
He treats agents as the next step rather than a current state, and defines the step precisely: text and images were generation, and what he wants now is actions
The example he works through is abandonment. A customer searches, looks at a product, compares variations and then leaves, and the agent reads that behavior, predicts the next step, judges whether a better offer is what is missing, and reaches back out
His argument for agents is org structure rather than capability. Companies run those same steps today across several different teams, and one agent, or a set of connected agents, can run the whole workflow
Human in the loop is a design requirement in his account, not a safety add-on: you have to think about it while designing, because the guardrails are what stop the AI acting on its own
Brand validation is the specific guardrail he names for anything customer-facing. The model has to be trained on the brand through RAG, and it must not communicate with a customer without that validation
He makes brand concrete rather than abstract: the EchoStar brand is family based and sports centric, so the communication style, the colors and the creative all have to match, and that differs from other brands
The drudgery he says AI takes back is testing. Releases used to absorb days and sometimes months before anything shipped, and AI-based testing is what he names as the change
His Northstar KPIs are two: how much human involvement has been removed, and how much customer conversion is happening
The clearest saving he quantifies is deflection. A query solved in the chat bot rather than routed to a call center with a human saves the cost of that call, and at high call volume that is money on every one
The second saving is structural: skipping systems means fewer interfaces, less infrastructure cost, fewer people maintaining it, and fewer of the small partners whose business is piping data between systems
On the retention side he frames churn as a support problem. Customers used to contact support and then leave, so a falling number of departures after that contact is a cost saving he can point at
What he wants instead of several models is one that handles text, image, video, audio and user intent detection together, which he calls multimodal convergence
Personalization is multimodal in his answer, not just copy: personalized creatives, personalized messages, personalized in-app notifications, and the preferred channel personalized too, because an email person and a text person are different customers
His test for whether the churn of new releases matters is ownership of data rather than choice of model. If the model is working on your own proprietary information and you have a good RAG in house, you are good
He uses LLMs rather than small models, and puts the small-model case on the device. Companies like Apple and others might find them useful, he says, because the deployment target has limited memory capacity, while the bigger enterprise companies run most of their use cases server side with a lot more compute available
His rule for when to stop building is arithmetic. If maintaining and training the models costs a lot and the result is a three or four percent improvement, go with a partner instead
The model decision is made by an enterprise-wide committee, and his reason is sprawl: engineers cannot each pick a tool when there are too many options, and policy on what data may be used for training has to come from one place
Legal sits on that committee, which he flags as new. Old technology committees did not include legal, and now every data set exposed to a model and every API called is a legal question
The committee has no single reporting line, in his experience at EchoStar and previously. It is assembled from engineering, product, marketing, customer support and customer experience, and he says it depends on how the organization is structured, because one function alone ends up solving one set of problems
Agents get their own KPIs underneath the business ones, split the way the work is split: agent performance and reduction on the agent side, guardrails and error rate on the human side
His forecast for enterprise AI starts with goals rather than technology: companies will carry AI-driven OKRs, with a set percentage of productivity expected to come from AI, and will treat AI as an investment
The systems that survive are the ones that hold the record. CRMs stay because the data has to live somewhere and you need a single source of truth
The systems that go are the integration layers, and he puts them first. Anything whose job is moving data between two other systems gets replaced by an agent doing that job
Monolithic systems go the same way, and so do the replicas. Instead of keeping your own copy you read live from the source of truth, which for inventory means reading from the supplier
The casualty he names that is a team rather than a technology is the database function: interim databases, database teams and database systems, unless the data is proprietary to you
SaaS does not die wholesale in his account, it splits. SaaS that holds a source of truth stays; SaaS that only sends email or runs social posting is in trouble unless it adds something
His first instruction to anyone starting is data, not models: build a single source of truth for the customer, a CDP or something like it, because personalization needs to know the person first
His second is to experiment with pre-trained LLMs and not train your own, and to build the RAG in house with your own vector database and your own embeddings, because AI reads through embeddings and will not query your Oracle database directly
His third is to have KPIs at all, and he has seen people run experiments without them. The point is not the cool factor, it is whether the business moved
His closing takeaway is the goal-driven framing again, with a number attached: humans set a goal, devise strategies and iterate over six months, and he says AI can run that loop and reach the goal in ten or fifteen days
About Al Shanmugam
Al Shanmugam is Head of Product at EchoStar, where he leads product for the satellite communications and streaming division, working across Dish Network and Sling TV. His remit covers recommendations, search, customer notifications and ad tech, and his focus is personalization, automation and AI-driven streaming that lifts engagement and monetizes content delivery. He was doing AI at scale at Amazon and Meta before generative AI arrived, which is the background he brings to intent detection and one-to-one personalization at consumer scale. He studied engineering at VIT University.
In this episode
| 00:41 | Welcome |
| 01:05 | Introducing Al Shanmugam of EchoStar |
| 01:22 | The five companies inside EchoStar, and which two he works on |
| 02:36 | What Sling carries: sports, then cricket and soccer, then news |
| 04:00 | Christina: are you the one helping me find things, or selling me ads |
| 04:20 | His remit: recommendations, search, notifications and ad tech |
| 04:52 | Christina: the four obvious AI use cases, search, ads, recommendations, notifications |
| 05:17 | AI at scale at Amazon and Meta, before the generative wave |
| 05:33 | The three areas: automation, personalization, intent detection |
| 05:33 | His definition of AI: a self understanding or goal-driven system |
| 06:09 | Segment-based personalization, and what replaces it |
| 06:28 | A person rather than a cohort |
| 06:46 | Intent detection: knowing without being told |
| 06:46 | Tailoring every screen, so one page renders two ways |
| 07:01 | Maturity levels differ by area |
| 07:26 | Agents as the next step, and the shift from answers to actions |
| 08:31 | The agent reads the behavior and reaches back out |
| 09:00 | One agent, or connected agents, instead of several teams |
| 09:28 | Own agents, for privacy reasons |
| 09:40 | Human in the loop as a design requirement |
| 09:58 | RAG, and brand validation before anything reaches a customer |
| 10:14 | What brand means concretely: family based, sports centric |
| 11:04 | Testing as the drudgery AI takes back |
| 11:28 | Christina: how are you judging the ROI of that workflow |
| 11:51 | Northstar KPIs against system-level technical ones |
| 12:06 | Deflection: the call that never reaches the call center |
| 12:23 | Why the old chat bots could not do it |
| 12:42 | Skipping systems, and the costs that go with them |
| 13:32 | Churn as a support problem |
| 14:22 | Christina: would the images be personalized too |
| 14:42 | Multimodal convergence, in one model rather than several |
| 14:56 | Personalized creatives, messages and notifications |
| 15:15 | Personalizing the channel itself |
| 16:34 | A separate model for recommendations, another that understands ads |
| 16:51 | New models every week, and the catching-up problem |
| 17:13 | Own your data and your RAG, and the release churn stops mattering |
| 18:02 | Small models are for on-device, enterprises have server-side compute |
| 18:35 | A high maintenance cost against a three or four percent lift |
| 18:53 | Spending more to get five percent, and the case for a partner |
| 19:15 | The clean room, where partner and first-party data meet |
| 19:44 | Llama for self-managed, Claude when Anthropic carries the patching |
| 20:06 | Christina: how does a new model get decided internally |
| 20:39 | The enterprise-wide committee, and the sprawl it prevents |
| 21:04 | Policy on what data may be used for training, from one place |
| 21:31 | Legal on the committee, which older technology committees lacked |
| 21:53 | No single reporting line, at EchoStar or before it |
| 22:15 | Assembled from engineering, product, marketing and customer experience |
| 22:35 | Christina: is there someone who leads it |
| 22:37 | Up into the CEO, and in some companies the CIO |
| 23:08 | Christina: do the agents have KPIs of their own |
| 23:26 | Agent KPIs underneath the business ones |
| 23:55 | AI-driven OKRs, and AI as an investment |
| 25:04 | Christina: do you see any of those systems going away |
| 25:18 | What stays: CRMs, and the single source of truth |
| 25:31 | What goes first: the integrators |
| 26:21 | Interim databases, database teams, database systems |
| 26:58 | Which SaaS is in trouble, and which is not |
| 27:27 | Christina: what is your guidance for someone early |
| 27:47 | Start with a source of truth for the customer, a CDP |
| 28:45 | Build the RAG in house, with your own vector database |
| 29:21 | And have KPIs, because he has seen experiments without them |
| 29:54 | Christina: what resources do you recommend |
| 30:33 | Codex, and reading a code base without being a coder |
| 31:22 | Hugging Face Transformers, LangChain and LlamaIndex |
| 32:11 | Christina: what should listeners take away |
| 32:37 | Set the goal, let AI iterate the strategies |
| 33:02 | His closing line: set the goal, use AI to reach it |
| 33:07 | Sign-off, and the invitation to the November summit |
In Al’s words
“a self understanding or a goal-driven system. That’s how I would describe AI”
— Al Shanmugam (05:33)
“About you as a person, not as a cohort or not as a segment”
— Al Shanmugam (06:28)
“So if I know, what you’re looking for. Without you explicitly saying, I can actually, customize my product for you”
— Al Shanmugam (06:46)
“a lot of companies do all of these steps using like different teams, right? So now AI can do all of this as a single agent”
— Al Shanmugam (09:00)
“guardrails are very important. So you have to implement all these guardrails so you don’t really let AI do on its own”
— Al Shanmugam (09:40)
“it has to be aligned with your brand. So it should not, communicate to customers without, brand validation”
— Al Shanmugam (09:58)
“we are talking about personalized creatives, personalized messages, personalized notifications”
— Al Shanmugam (14:56)
“As long as your model is actually working with your own proprietary information, you’re good”
— Al Shanmugam (17:13)
“the systems that are gonna go away are integrators”
— Al Shanmugam (25:18)
“think about, a goal setting and then using AI to reach the goal”
— Al Shanmugam (33:02)
Resources
Websites and projects
Al Shanmugam
Al Shanmugam on LinkedIn: Where he says at 30:06 he can be reached, and he is explicit that he uses it more than X. He also offers a chat or a call to anyone who wants to talk about LLM-based personalization
EchoStar: Where he is Head of Product. He works on Dish Network and Sling TV, two ofthe five companies he lists at 01:28
Sling TV: The streaming service he names as one of the two he covers. Sports first, then cricket and soccer, then news, sold in bundles rather than as one package
Dish Network: The satellite service, with programming he describes as very similar to Sling and with local channels in more markets
What he recommends
OpenAI Codex: The tool he says at 30:33 he loves, and the reason he gives is that you can understand a code base without being a coder or a computer science expert. He mentions meeting OpenAI three or four weeks before this conversation to talk about it
Hugging Face Transformers: He describes it at 31:04 as GitHub for AI projects and AI apps, and at 31:22 says it is where to look at what is currently going on
LangChain: One of two frameworks he names at 31:22, and at 31:40 he gives the reason: you do not need a lot of insight into these technologies to run them
LlamaIndex: The second of the two frameworks he names at 31:22, for the same reason: you do not need a lot of insight into these technologies to run them
Chroma: The vector database he names as an example at 28:45 when telling anyone starting out to build their own RAG, and his own words on the choice are that anything is fine
Named on air
Llama: Meta’s open-weight family, which he names at 19:29, and at 19:44 he gives the reason for reaching for it: you do not want to rely on a company, so you do your own training and your own maintenance
Claude: The managed option he names at 19:44, where Anthropic carries the patching and the model updates so his team can stay on business innovation rather than model maintenance
Amazon and Meta: Where he says at 05:17 he used AI at scale before the generative wave. His current employer is EchoStar
Apple: His example at 17:42 of where small models earn their place, and at 18:02 he gives the reason: the deployment target is a device with limited memory capacity
AWS: Named at 18:02 as the kind of cloud compute that makes server-side inference practical for a large enterprise
Oracle: Used at 28:45 as the example of the database AI will not query directly, which is his argument for embeddings and a vector store
Boost Mobile, Hughes: Two of the five EchoStar companies he lists at 01:28, alongside Dish Network, Sling TV and EchoStar itself. He does not work on these
CNN, CNBC, Bloomberg, Fox: The news channels he names at 02:54 when describing who comes to Sling for what
Ideas and terms discussed
Intent detection: The third of the three areas he names, alongside automation and personalization. He describes it as knowing what someone is looking for without their saying it, so the product can be customized for them
Hyper-personalization: His word at 06:09 for one-to-one rather than segment-based, and he defines it by contrast: AI learning about you as a person, not as a cohort
Segment-based personalization: The practice being replaced. Decide a group likes something, put them in one segment, send the segment the offer
Brand validation: The guardrail he names for outbound customer communication. The model is trained on the brand through RAG, and nothing reaches a customer without passing that check
Northstar KPI: His term for the top-level measure, against the system-level technical KPIs underneath it. His two are human involvement removed and customer conversion
Clean room: The privacy-controlled environment where partner data and first-party data are transacted under compliant infrastructure, with the AI trained on top of it
CDP: Customer data platform, his name for the single source of truth about a customer that he says to build before anything else
Multimodal convergence: One model that understands text, image, video, audio and user intent together, which he wants because keeping up with several separate models is the problem
Source of truth: The property that decides which systems survive his forecast. Hold the record and you stay, move data between systems that do and you are the integrator being replaced
Related AI Realized episodes and events
Smaller Models, Bigger Wins: Verify Before You Answer: Jason Williamson on the case for small models, which is the argument Al Shanmugam takes the other side of
Agentic AI and Revenue Work: What Actually Pays Off: Christopher Penn on agents pointed at revenue, which is what the abandonment workflow here is built for
Only Content That Clears Every Agent Gets Monetized: Adil Ajmal on personalization and advertising at fan-platform scale, the same problem one layer out
Frequently Asked Questions
-
Intent-driven personalization is working out what a customer wants without their saying it, then changing what they see accordingly. Al Shanmugam, Head of Product at EchoStar, describes it as the third of three AI areas in a product organization, after automation and personalization, and defines it by what it enables: if he knows what you are looking for without your having said so, the product can be customized for you and every screen tailored, to the point where one page renders differently for two different people.
Transcript 05:33 to 07:01
-
Segment-based personalization groups customers who look alike and sends the whole group one offer, while one-to-one personalization learns about an individual. Al Shanmugam describes the old pattern as deciding that a set of customers likes something, putting them all into a single segment and sending that segment the offer. What replaces it, which he calls hyper-personalization, has AI learning about a person rather than a cohort, on the grounds that the interests that matter are often too nuanced to survive being grouped.
Transcript 06:09 to 06:46
-
Agents can act on customer behavior rather than only respond to it. Al Shanmugam frames the shift as moving from generation to action: text and images were what the previous wave produced, and what he wants now is for AI to do things. His example is abandonment, where a customer searches, compares variations and leaves, and the agent reads that behavior, predicts the next step, judges whether a better offer is what is missing, and reaches back out by push notification, text or email.
Transcript 07:26 to 09:00
-
Customer-facing AI needs a human in the loop and brand validation before anything is sent. Al Shanmugam treats the human in the loop as a design requirement rather than an added safeguard, because the guardrails are what stop the system acting on its own. On top of that, anything reaching a customer has to be trained on the brand through retrieval augmented generation and validated against it, so that the communication style, the colors and the creative match. At EchoStar that brand is family based and sports centric.
-
Al Shanmugam measures the ROI of an AI agent on two top-level numbers with a set of technical ones underneath. His Northstar KPIs are how much human involvement has been removed and how much customer conversion is happening. The clearest saving he quantifies is deflection: a query solved in the chat bot rather than routed to a human in a call center saves the cost of that call, which matters at high volume. Underneath that sit the infrastructure and headcount costs that fall when systems are consolidated, and he counts the small partners whose business is piping data between those systems in the same saving.
Transcript 11:51 to 14:22
-
Al Shanmugam uses large language models and puts small ones where the deployment target is constrained. His reasoning is that small models earn their place on a device with limited memory, of the kind Apple ships, while enterprises running inference server side have the compute available through providers like AWS to do it differently. He does use several models rather than one, a separate model for recommendations and another that understands advertising, and says one multimodal model handling text, image, video, audio and intent would be better than several.
Transcript 14:42 to 18:02
-
The choice of AI models should sit with an enterprise-wide committee that has legal on it, in Al Shanmugam’s account. Al Shanmugam says the reason is sprawl: individual engineers cannot each choose a tool when there are too many options, and policy on which data may be used for training has to come from a single place. He notes that older technology committees did not include legal and now must, because every data set exposed to a model and every API called is a legal question. The committee is assembled across engineering, product, marketing and customer experience, and has no single reporting line, though he says those leaders all report up into the CEO, and in some companies the CIO.
Transcript 20:39 to 22:37
-
Integration layers go first, according to Al Shanmugam, followed by monolithic systems and the databases that mirror them. Anything whose job is moving data between two other systems gets replaced by an agent doing that job, and replicas give way to reading live from the source, which for inventory means reading from the supplier. Interim databases, database teams and database systems go with them unless the data is proprietary. What stays is whatever holds the record: CRMs, sources of truth, and SaaS that holds one.
Transcript 25:04 to 27:27
-
[00:00:00] Christina Ellwood: AI realized the podcast about everything that is new now and next for enterprise executives deploying ai, hosted by myself, Christina Ellwood, and my collaborator, David Yakovich.
[00:00:17] David Yakobovitch: Our guests will share. What's driving AI adoption use cases and business models for the data and AI economy?
[00:00:41] Christina Ellwood: Welcome to ai. Realized the podcast for enterprise executives leading AI deployments. From tackling security data and operational challenges to navigating organizational transformation, AI deployment offers a unique opportunity to redesign our [00:01:00] organizations from the inside out. I'm Christina Elwood, your host for today's episode.
[00:01:05] Christina Ellwood: We're talking today with Al Shama gum head of product at EchoStar Al. Welcome to AI Realized.
[00:01:14] Al Shanmugam: Thanks, Christian. Thanks for having me.
[00:01:16] Christina Ellwood: For those who are not familiar with EchoStar, tell us a bit about the company and your role.
[00:01:22] Al Shanmugam: Cool. So yeah, EchoStar is a satellite company. And, it has five different, companies associated to it.
[00:01:28] Al Shanmugam: So we have, dish Network, sling tv. Boost Mobile hues and EcoStar by itself. So all five companies are different domains, but Dish Network and Sling TV is where, I work, I work on all of the products on that area. And also the interesting part about Dish TV and, sling is, both are, connected to tV and streaming division. So it's mostly video and audio services and all of the consumer products that comes with it. So I specialize in this area of AI and, streaming personalization. So yeah I feel like a natural fit in that [00:02:00] area. And, yeah, so this is a pretty cool place where, we work on a lot of streaming challenges live tv.
[00:02:05] Al Shanmugam: And how do personalize it for the consumers and, the live TV space, as it's changing quite a lot. So it is very t and we compete with a lot of big giants, in the market. So it is always pretty cool to do it, with the less number of people. How do you compete with, the big giants with the higher technology level?
[00:02:22] Al Shanmugam: It is always, good for me.
[00:02:24] Christina Ellwood: Yeah. So if I were to go to my streaming on my TV and select sling tv, what live programming would I see? Is it sports programming, news programming? What's the live programming type? That I would see?
[00:02:36] Al Shanmugam: I. We have it all. In this case some consumers come to us, for sports specifically, right?
[00:02:41] Al Shanmugam: So basketball, football, college football baseball, all of those games, we specialize, recently, we have this, specialized division, cricket and soccer as well. So a lot of customers come to us for that. And, there are certain segments of customers, who allow us for our news.
[00:02:54] Al Shanmugam: Like CNN, Fox market news like CNBC Bloomberg, things like that. So [00:03:00] it's yeah. So there are different audiences, for the live tv. And also we operate by bundles. So again, it is very, meticulously designed bundles for, specific customer segments because you don't have to really pay for, what you don't, watch.
[00:03:10] Al Shanmugam: So we don't really sell it like, cable where, you know, you have to buy into everything. Here very bundles purchase into the cheapest, option in market. So customer.
[00:03:23] Christina Ellwood: So Sling is the streaming and then is Dish the satellite tv? And is it the same programming or different program?
[00:03:29] Al Shanmugam: Very similar programming. So there is a lot of overlap, but as you said, it is satellite tv. So we have our own satellite and we stream, we communicate using that. So Dish you will get like a lot of different, local programming as well because you are having antenna and things like that.
[00:03:44] Al Shanmugam: So you are, you can consume your local channels if you are in. If you're in Chicago, if you lobby your local TV so you can get that, in sling, we offer that in certain markets, but not in all markets. But this is think of, this is like a much bigger, channel selection compared to [00:04:00] sling.
[00:04:00] Christina Ellwood: Gotcha. Okay. Thank you for that background. I think that's probably really helpful for our listeners. And as the head of product marketing, are you the person responsible to help me as a consumer to find what I'm looking for, or are you responsible for the advertising I'm seeing or for both?
[00:04:20] Al Shanmugam: I take care of, to be, precise both right? But to be very specific, I take care of, what customer, sees recommendations and also how customer interacts with the platform. Like search the audio video four. And also how customers, are getting these notifications, right?
[00:04:39] Al Shanmugam: Like a marketing standpoint, getting email about the new content, getting a text message, push notifications all of that. And also, I play in the ad tech area as well, so where, how, what are the advertisements you're seeing? Is that contextual? Is it useful for the customers?
[00:04:52] Al Shanmugam: So yeah, I look into all of that. Yeah. And they all seem like natural AI use cases like, AI search, AI ad [00:05:00] placements, AI recommendations, AI notifications. Is that, what you're using AI for, or are, am I missing the most important use cases from your perspective?
[00:05:11] Al Shanmugam: Yes. In this case, ai, AI is a very, useful tool, for all of us in the business, right?
[00:05:17] Al Shanmugam: So every area of business, needs AI in different ways and, we can use it in different ways. So again, in my personal, scenario, I also come from, strong background of, AI based companies like Amazon and Meta. So where we use AI at scale, like even before this whole Gen AI boom, right?
[00:05:33] Al Shanmugam: So again, in all of these departments, AI plays a different role. And automation is one big thing where, you can automate a lot of things using AI because it's very contextual, right? In the past, like we used software and systems to do that. It is very rules, but now we are not anymore. Continuously from the mistakes that that is being made and then, continuously improve yourself. Yeah, how do you say a self understanding or a goal-driven system. That's how I would describe AI. And, outside of [00:06:00] automation we think about personalization. So these are the new things which we think about where in the past, if you think about it, we do segment based personalization hey these customers love this.
[00:06:09] Al Shanmugam: So let's market that to the customer, right? And so we put all these customers into a single segment, and then we send out these offers. It applies to every business, right? Think about retails every business now you can do one-to-one communication where this hyper-personalization comes into the picture where AI can learn.
[00:06:28] Al Shanmugam: About you as a person, not as a cohort or not as a segment, right? So instead, you might have some nuanced, liking or nuanced interest. So intent deduction is another area which where we use, right? So we talked about automation, personalization, and then the intent deduction. So intent deduction is useful in product in multiple ways, right?
[00:06:46] Al Shanmugam: So if I know, what you're looking for. Without you explicitly saying, I can actually, customize my product for you, right? I can tailor made every screen for you. Think about homepage a same, dot com page looks different for you versus [00:07:00] me. That's where we are going, right?
[00:07:01] Al Shanmugam: So there is like a e-commerce aspect to it as well. So yeah, so all of these areas are very AI driven and there are different maturity levels, of where I am and where, where we are doing things. We can talk about that.
[00:07:14] Christina Ellwood: Yeah. You just anticipated my next question. So it sounds like the autonomous agents is the area that you've been doing for the longest, just based on your, the implication of the way you described it.
[00:07:25] Christina Ellwood: Is that correct?
[00:07:26] Al Shanmugam: Yes. So agent ai is very, is the next step, right? And currently we are doing things, to achieve that. So far, again, with, OpenAI and, other, LM based models. You can generate text, you can generate images, right? So now you wanna achieve actions, right?
[00:07:41] Al Shanmugam: You want, AI to do things for you, right? Rather than, asking questions and, giving back the responses. So we are moving ahead of the chat bot based, applications, right? So again, chat bot is one important, place where we, where most of the companies now use ai.
[00:07:55] Al Shanmugam: But, what is the next step, right? [00:08:00] Is.
[00:08:01] Al Shanmugam: Let me give you an example, right? So where, think of marketing. So now we go land on a homepage. Now, you're looking for certain. You, search for it. You saw the product. Now you wanna understand like what are the variations in it? What are the different colors you have available? And things like that.
[00:08:18] Al Shanmugam: When it comes to our businesses what are the different channels you're looking for? Things like that, right? So when you go through that, flow, and then say for example, you stopped and, you had some, some of the work you moved on. Now I wanna reach out, back to you.
[00:08:31] Al Shanmugam: So now with ai, with agent ai, what we can do is like technically, understand all the customers, whatever, customers doing on the side, and also predict what. Do as a next step and, are they actually looking for a better offer? Things like that. And then also reach out, back to the customer through some channels like push notification, text messages, email, whatnot, and, get them back to the website and, get them, convert them, as a paid subscriber.[00:09:00]
[00:09:00] Al Shanmugam: So you can do all of that. So in the past, even currently, a lot of companies do all of these steps using like different teams, right? So now AI can do all of this as a single agent, right? So you can have it as a single agent or connected agents. Depends on how complex your systems are. You can have connected agents and then the whole AI workflow will help you achieve.
[00:09:20] Al Shanmugam: Interesting part.
[00:09:21] Christina Ellwood: And are you writing your own agents, Al Do you guys write your own agents and manage them and and maintain them and so forth?
[00:09:28] Al Shanmugam: Yes. So right now, again, if it is very, preparatory, right? So then, we use our own agents, right? Due to, privacy reasons. And also, and most of 'em are right now related to systems, right?
[00:09:40] Al Shanmugam: Talking to one, every system in that specific, area when it reaches consumer. So then, we have human in the loop, right? So again, designing this you have to think about a human in the loop, a design, because guardrails are very important. So you have to implement all these guardrails so you don't really let AI do on its own.
[00:09:58] Al Shanmugam: And also reach out to like [00:10:00] all sorts of, external websites because it has to be trained using your brand. So Rag that is where RAG comes into picture. We can talk about Rag if you're interested. So where, it has to be aligned with your brand. So it should not, communicate to customers without, brand validation.
[00:10:14] Al Shanmugam: So there are certain things, for example, our brand is very family based, right? Very sports centric. So where your communications are very different compared to compared to other brands, right? So now your communication style, your colors, your creatives, all of that should match with your brand.
[00:10:30] Al Shanmugam: And and that, that is the, I'm talking about.
[00:10:34] Christina Ellwood: So let me just we just did a a round table on agents and one of the takeaways from that was that closing gaps between systems is a primary use case for agents, and the other is reducing toil reducing. Dreaded work drudgery. So it sounds like you're using agents between your systems with a human in the lube which is the closing the gaps.
[00:10:58] Christina Ellwood: Are you also doing it to [00:11:00] reduce drudgery? And what are some of those areas? I.
[00:11:04] Al Shanmugam: Yeah. Yeah, think about some testing use cases, right? Before you release something to the customer or before you wanna release a, product experience to the customer. So you wanna do a intense amount of testing.
[00:11:14] Al Shanmugam: We all spend days, sometimes months in testing some of these, releases where now you don't have to.
[00:11:24] Al Shanmugam: Test cases and also do, AI based testing.
[00:11:28] Christina Ellwood: So let's take these, the agent use case where you're closing the gap between the system. So we have somebody's coming to the website, you're gonna do this intent understanding, you're gonna then reengage them. That whole workflow where you're going from system to system with the agent and the chat bot is gonna be the primary interface back to the customer.
[00:11:46] Christina Ellwood: How are you judging the ROI of that use case?
[00:11:51] Al Shanmugam: Yeah, in this case. Okay, so there are like Northstar, KPI, KPIs, right? And then there are like system level technical KPIs, right? So Northstar is [00:12:00] like basically how much of, human involvement reduction that we are having, right? So basically you save dollar value there, right?
[00:12:06] Al Shanmugam: If, it's been the same case, from customer support for a very long time, right? So instead of, routing it to a call center with human if you can solve it in the chat bot itself, so you save a ton of money. So you save for every call, and if your call volume is high so then technically, you save, dollar value valued right there.
[00:12:23] Al Shanmugam: But, in the the chat bots are not that, because. It responds with like series of te texts that are programmed, right? It says the same thing over and over, but now with ai so now it becomes very natural, right? Very organic, right? So that's one. The second thing is, again, connecting all these systems and maintaining these systems, right?
[00:12:42] Al Shanmugam: So now. Say, for example, certain systems you can skip so you don't have to, have multiple different UIs, right? Where agents actually go in, log in and see a hundred things before actually talking to the customer, right? So now you can have it all automated because you already, train [00:13:00] the AI with all that data.
[00:13:02] Al Shanmugam: Your, own proprietary data. So now the AI can actually, don't have to look it up because with human, we have that limitation where, you cannot remember that level of stuff, right? What this means is like basically you're reducing some systems. So this, means infrastructure costs.
[00:13:17] Al Shanmugam: You're talking. Number of, people, help in maintaining it. And number of even like smaller partners. So we, every company uses like smaller partners to actually, connect between different systems as well. They act less a, they act as a pipeline between these devices or between these, systems.
[00:13:32] Al Shanmugam: So now all that has gone, so you will see that, benefit. And, on top of it, like a Northstar metric is how much of customer conversion that is happening, right? Say for example, if customer churn. Churn rate is very important, right? So churn rate or retention rate, however you measure it in the past customer reach out to the customer support and then, end up leaving the service.
[00:13:55] Al Shanmugam: Right now if you are seeing that going down, the [00:14:00] number of customers leaving the service. So then that is a huge cost saving for the company. So that is another lever from customer support standpoint. But, from the previous personalization use cases I was talking about their conversion rate address that, KPI.
[00:14:12] Al Shanmugam: Where if I can get say for example, if I can get a thousand more customers convert because of ai so then it's like a thousand new subscribers for the business, right?
[00:14:22] Christina Ellwood: And when you're doing that personalization, obviously you're gonna convert more if you personalize, to, the personalization resonates with the audience that you're targeting.
[00:14:33] Christina Ellwood: So would, if I were getting that personalized message, would it have personalized. Images as well as personalized messages.
[00:14:42] Al Shanmugam: Yes. That is why I'm very excited about this next stage, of multimodal, convergence, right? Where it's not just just text, right? We are talking about a model, that can understand text, image, video, audio, and also user intent detection.
[00:14:56] Al Shanmugam: So if all of that is in the same model then that would be perfect, right? [00:15:00] So instead of using all these multiple models. But, so to answer your question, yes. So we are talking about personalized creatives, personalized messages, personalized notifications in app notifications. Also, reaching out back using your own preferred channel, in a personalized way.
[00:15:15] Al Shanmugam: Because, you might be an email person. I might be a text person who see the text and interact with it, but not email at all. So there are different channels customers prefer. So how are you personalizing that channel level communication would be the last stage.
[00:15:27] Christina Ellwood: In some of the recent reports from analysts on AI adoption in the enterprise, they're reporting, companies with high ROI, which they're defining as over 25%.
[00:15:38] Christina Ellwood: ROI. This is a box, report I'm thinking of. But I think the Mary Meer report might have some of this in it as well. That, companies that have high ROI. For their AI deployments are using multiple models, so they use a public model, like an open AI for some things. But they use a custom, ai, which might [00:16:00] be something that they've trained themselves, but they also are using these I think of them as vertical ais where they've been trained on industry information or some type of niche information so that those, and they're usually a small language model so that those are.
[00:16:14] Christina Ellwood: Sort of boutique models is, has that been your experience as well? That you need different models for different cases? And if so, which of those are you using that are behind the firewall and which are the public side? How do you make that distinction or separation about which to use?
[00:16:32] Al Shanmugam: So today that is the reality, right?
[00:16:34] Al Shanmugam: So where, you have to use multiple models to achieve, certain things, and we do the same, right? So now if you wanna, use it for recommendations, you have to use a separate model. And then if you wanna use for advertising then you have to use a separate model that understands ads, which connects with all of the partner informations, right?
[00:16:51] Al Shanmugam: Yeah. So the creatives and all that. And also it has to be very dynamic, right? Very real time. Some models doesn't have to be real time. It can train overnight [00:17:00] and then give you results in the morning, right? So yeah, now there are that nuances and differences. That is why, I'm super excited about the multimodal thing where, we have to get away from this because at this point of time what'll happen is, the new models will be released like every week.
[00:17:13] Al Shanmugam: Nowadays, like every day, every week. So now consistent catching up, will become a problem. And, again, you don't have to, right? In my experience, you don't have to, right? As long as your model is actually working with your own proprietary information, you're good. And if you have a good rag, like a retrieval augmented generation in your in-house, based off of your own data so then you are good.
[00:17:33] Al Shanmugam: So because you know your embeddings are in your in your systems, and then your model is actually trained on your own embedding.
[00:17:42] Al Shanmugam: So public use cases comes into the picture where you know, if you are doing like partner marketing, things like that's where, public information or public hosting of models will help. Again, most of the cases that we use are LLMs, right? Not, the smaller, models. But for companies like Apple and others, it [00:18:00] might be useful because, they're trying to deploy it inside a device.
[00:18:02] Al Shanmugam: With a limited, memory capacity and all that. So all of the other bigger enterprise companies, most of the use cases will be on the server side, so we will have a lot more compute to do that with AWS and other cloud, computing. So you'll be able to do it. Using LLMs. Yeah, so in our case, at least in I've, used mostly LLMs rather than, the smaller ones.
[00:18:22] Al Shanmugam: But again, even in LLMs, right? So there are like different, different sizes or different, processing that you might want to do based off of your use case. So again, you start with the use case and then think about, kPIs because usually business, they don't really measure KPIs.
[00:18:35] Al Shanmugam: Again, that is very important, right? It's not just about the cool factor of having it, but it's whether it's actually yielding results or not. In this case, say for example, if your maintenance cost is like really high, for maintaining the models and training it, but big data. Rather, and the results are like, you're seeing only like 3%, 4% increase.
[00:18:53] Al Shanmugam: So then it's not worth it, right? Because you're spending a lot more money to get that 5% increase. So then I would say, go with a partner, [00:19:00] right? A. There are a lot of companies in the market who could do it, for you. And then, just like increase that. So KPI matters, but in some cases actually, hosting them, training them in your, in-house, due to privacy reasons as well as, again so I worked on most of the privacy related in share ups.
[00:19:15] Al Shanmugam: So yeah, privacy control. Maintaining a clean room is very important because, that's where you transact all your partner, data, and also your first party data in a very clean way with all privacy compliant infrastructure. And your AI is trained on top of that in a very hosted way, right?
[00:19:29] Al Shanmugam: So in your controlled way, which yield better results. Yeah. So in this case, again, it's use case by use case. Again, I, advertising recommendations, all of that is use, using LLMs in-house. Some are open source models and some are not. Again, you could use, models like Lama.
[00:19:44] Al Shanmugam: If you are if you don't really wanna rely on a company and, do your own training, do your own maintenance and all that and in some cases, we use, things like cloud where, you know, the full support, is given by Anthropic. So they'll take care of all of that, patching, model updates and all that stuff.
[00:19:59] Al Shanmugam: I know you don't have to worry about [00:20:00] it. And then you can just keep doing your own business innovation rather than like the model level,
[00:20:06] Christina Ellwood: maintenance. And when you need a new model, like you're interested in using these multimodal models how is that decided internally?
[00:20:12] Christina Ellwood: Do you have a committee that is handling your ai decisions across the organization? Or is there a, how's that done? Because that's different it seems like for every organization and given how complicated your organization is. You have three entities that have been brought together into a single entity.
[00:20:31] Christina Ellwood: They had their own processes in the first place. They might have already been using ai. So how does that decision process look in your organization?
[00:20:39] Al Shanmugam: Yeah, so it is so you already may, answered the question, right? It is a committee, right? So you need an enterprise wide committee because, your teams engineers will.
[00:20:49] Al Shanmugam: AI tool should I use because there are too many variations and too many, different options available today. So what are the tools I should use and they know, what is the right, way to, do this, right? Best practices, under the [00:21:00] policy itself, right? What data I could use for training and what data I should not.
[00:21:04] Al Shanmugam: All of that stuff, has to come from a single, place when it when we're talking about an enterprise in our case, yeah. So we have a gen AI committee who actually, meet every week talk about new things, new use cases and also, talk about the progress that we make, right?
[00:21:16] Al Shanmugam: Because, we also have to track. Track it very similar to how you manage any project, in the ecosystem. So yeah. So that committee will help us, drive. And also it is good to have a legal person as part of, the whole JI as well. Maybe in the past, like technology committees don't have.
[00:21:31] Al Shanmugam: Legal, but now that you do have to have a legal so you have to think about every single data set that you're exposing or sharing with, with the model and which APIs you are contacting, right? That also makes, makes a big difference, right?
[00:21:44] Al Shanmugam: Are you allowed to use those APIs? Is it like, proprietary and all that stuff? Yeah. So the this centralized team is very helpful.
[00:21:50] Christina Ellwood: And where does that team report in?
[00:21:53] Al Shanmugam: Oh, it, okay, so we don't have, again based of the current, experience and even my past experience with MAD and, Amazon.
[00:21:59] Al Shanmugam: So we [00:22:00] never had a single reporting structure, right? So this, these people, or this committee is formed by different people, from different side of the business, so that, we talk about all different use cases, right? And also they.
[00:22:15] Al Shanmugam: Departments, I would say engineering, right? Product, marketing. Think of customer support, right? Customer experience teams. It depends on how the organization is structured. So having one person from all of these teams, or one leader from all of these teams is very important. Otherwise what will happen is like we'll end up solving only one set of problems, and we are not addressing like the whole full enterprise.
[00:22:35] Christina Ellwood: Is there someone who leads the whole committee?
[00:22:37] Al Shanmugam: Technically end of the day, all of these all of these, people will, report into CEO, right? So end of the day, these are big leaders we're talking about. So they will be reporting into A CEO and in some ca cases, CIO, it depends on different companies. But in the, in our case, it is like it is all they all go into CE.
[00:22:55] Christina Ellwood: And they're just, they're able to make decisions at the speed that people need for the yeah. [00:23:00] Model decisions and so forth.
[00:23:01] Al Shanmugam: Exactly.
[00:23:01] Christina Ellwood: So I wanna shift gears just a little bit and talk about what you envision coming.
[00:23:08] Christina Ellwood: In the future. Actually before I ask that question, let me ask one remaining question on the agents. Do you have unique KPIs for the agents or did you take the KPIs that you had for the workflow in the first place and then just split them between the agent and the human in the loop?
[00:23:25] Al Shanmugam: Yeah.
[00:23:26] Al Shanmugam: Again I'm using same KPIs like, which I was talking about, the top level business KPIs. But from granular standpoint, yeah. So we have, different KPIs that feed into the, bigger KPI. Agent performance, what is the reduction and all that, goes into the agent area.
[00:23:39] Al Shanmugam: And then from human standpoint what are the guard rails? How is it function functioning, every single part of it. How's your error rate? Are you getting,
[00:23:47] Christina Ellwood: okay, so there are KPIs that are unique to the
[00:23:50] Al Shanmugam: right?
[00:23:51] Christina Ellwood: Yes. Okay. Alright, so now I'd like to talk about what coming.
[00:23:55] Christina Ellwood: Where do you see AI going in the coming years [00:24:00] for the enterprise? On the commercialization side, marketing, sales product, that side of the house. Yeah. I see most of in the future most of the enterprise companies will have, an AI driven goals, right? If your company is following some kind of OKRs, they will all have okr, AI driven OKRs.
[00:24:20] Al Shanmugam: So where, hey, this much percentage of, productivity has to come from ai. So you will have certain goals set for AI itself. And also enterprises will start thinking about, the AI as an investment, right? So where, how much, you can invest in an ai. So right now, all of the AI companies do that and, the big, companies do that.
[00:24:39] Al Shanmugam: But enterprise companies, the adoption level is not that great, as of now. So now what will happen is like the adoption rate will increase. But instead of relying on, previously, design systems so now, the systems will be replaced by ai, right? So you'll see AI agents in sort of systems, right?
[00:24:55] Al Shanmugam: So you'll still have APIs and all that, APIs and services and all that, but it'll be driven [00:25:00] by, agent AI rather than by rules-based systems, or do
[00:25:04] Christina Ellwood: you see any of those systems going away?
[00:25:07] Al Shanmugam: Yeah. Yeah. I see a lot of systems, going away,
[00:25:09] Christina Ellwood: that make some forecast for us here. What systems are gonna go away.
[00:25:13] Al Shanmugam: Yeah. So okay, so let me start from the systems that's gonna stay, right? So
[00:25:17] Christina Ellwood: Okay.
[00:25:18] Al Shanmugam: Make systems like, CRMs will stay because the data storage always has to have, has to be there and you need a single source of truth. So the source of tool systems will be, will be here to stay, but then the systems that are gonna go away are integrators, right?
[00:25:31] Al Shanmugam: The first, right? So the systems that we're trying to integrate between one system to another will go away. So now what'll happen is, all of these systems will lead, replaced by some form of AI or agent, tools that will start, doing this job. And also the, the systems like, you know, how do you say the monolith, systems today that we use, will go away? Very similar to, all of the, supply chain, inventories things like that, all of that will go away. So now what will happen is like everything [00:26:00] will be real time. So instead of you having your own replica you will read it directly from the source of truth.
[00:26:07] Al Shanmugam: Which is like your supplier, right? So you don't really, wait for, system catch up database upgrades, ca you know, move the data from here to here. All of that, will never happen. So now you'll be reading it live and everyone will have, have their systems upgraded to do that.
[00:26:21] Al Shanmugam: So all of those, interim databases and database teams, database systems, all of that will be gone unless it's proprietary for you. So proprietary will be here to stay because, you do have to have systems and services to manage that.
[00:26:33] Christina Ellwood: So you'll still have any ERP, it just won't have the same
[00:26:36] Al Shanmugam: Right
[00:26:36] Christina Ellwood: span.
[00:26:37] Christina Ellwood: It's span will narrow.
[00:26:39] Al Shanmugam: Exactly. Exactly.
[00:26:40] Christina Ellwood: Okay. So you envision we'll have fewer monolithic systems. A few sources of truth, do you think? There's some, speculation that SaaS is dead. I think that you've answered that really? Because if it's a source of truth and it's SaaS, it'll stay.
[00:26:57] Al Shanmugam: Exactly.
[00:26:58] Al Shanmugam: Yeah. So sas, but [00:27:00] we cannot cover everything as a SaaS, right? So like in this case, all SaaS will not go away. Yeah, the SaaS with, source truth will be there. But then, if you are using a SaaS for just emailing customers, right? Or just, some kind of, social media marketing tool, right?
[00:27:14] Al Shanmugam: So those kinda SaaS are in trouble. But again, these SaaS companies will innovate. If they innovate, they'll be able to stay, right? So what is the additional value they are bringing to the table? Which is, which is where, their life is gonna be or their growth is gonna be.
[00:27:27] Christina Ellwood: So for other product and marketing executives like yourself, what's your guidance for those that are really early in their AI journey? You're quite far along in your AI journey for those that are early, what's your advice? Or guidance for them.
[00:27:43] Al Shanmugam: Yeah I would say start with some kind of, building out a source of truth systems, right?
[00:27:47] Al Shanmugam: Where you have, a single source truth for understanding your customers, right? Any business you could be, so there are similar, like CDPs we call it as CDPs. And there are different, structures to it. So have some [00:28:00] systems like that. So where you understand the customers from all different touch.
[00:28:04] Al Shanmugam: Customer might be, interact with your app, interact with your website, interact with your customer teams. So now central, very important ai. Do anything, from personalization or from, tackling problem standpoint. So it needs data, right? It needs data to understand about, the customer in order to do the one-on-one personalization, right?
[00:28:27] Al Shanmugam: From customer support or any angle that you take, that is very important. So start with some form of CDP or some form of centralization. Second is like start using LLMs as a, experiment, right? Again, most of us are doing everything as experiments, right? So now the product building itself is becoming an experiment, right?
[00:28:45] Al Shanmugam: So it might work, it might not, right? So then you keep by trading on top of it. So yeah, start doing experiments with LLM and don't train your models by yourself. So use, the pre-trained models and, develop a rag, right? In in-house. So where you will have a vector database think [00:29:00] of databases like chroma some vector database, anything is fine. Open source or per vector databases, have your own drag have your own embeddings because AI can, ai, reads text, images, everything using an embedding, right? So vector database is very important for AI to actually interact with it. So AI is not gonna directly go into your Oracle database and, do things with it, right?
[00:29:21] Al Shanmugam: So you have to have this, property rag in order for AI to be better. So start experimenting. The third thing is definitely have KPIs, right? I've seen, people doing experiments without KPIs. So in this case, we are not just doing this or using AI just for the sake of using ai. We are using AI to actually, improve our business, right?
[00:29:40] Al Shanmugam: To move faster. And, that's how it's gonna be and that's how your competitors gonna do it. So you should also do the same, right? So now getting to market or go to market is.
[00:29:52] Al Shanmugam: That's exactly where you should, think about.
[00:29:54] Christina Ellwood: Yeah that's great guidance. What resources do you recommend to listeners who wanna learn more about you [00:30:00] or about EchoStar? Or about doing, using AI in product marketing? I.
[00:30:06] Al Shanmugam: Yeah, to learn about me, I use LinkedIn quite often LinkedIn is the best place to reach me.
[00:30:09] Al Shanmugam: I can give you the link, from my profile and, they can, message me there. Because compared to X, I use LinkedIn more. So that's the best place to reach me. And if you wanna talk about LLM based personalization, happy to talk about it, in a chat or in a call. And you can reach me.
[00:30:22] Al Shanmugam: Second is to learn about my company and what we are doing. Yeah. So we have our own website and, which links to like multiple other websites. So you.
[00:30:33] Al Shanmugam: And, and also some of the things that I was talking about like rag, your personalized, experimentations, LLMs and all of that. So probably, I can give you some links. For example, I love Codex, which is, launched by OpenAI. So I met OpenAI a couple weeks, three, four weeks back to talk about that.
[00:30:48] Al Shanmugam: So yeah, so Corex is pretty cool. We should try that out, because that is now you can understand the code base without you being a coder, you don't have to be, which is beautiful,
[00:30:57] Christina Ellwood: isn't it?
[00:30:57] Al Shanmugam: Yeah. You don't have to be a computer science expert to do [00:31:00] that anymore. Yeah. And also now you don't have to worry about, if your company has like a lot of legacy code.
[00:31:04] Al Shanmugam: So now you know, AI will read the code and, give you whatever you want. So it's amazing. And, and also yeah the things that I was talking about, the projects and other LLM projects, hugging face as this, transformers, which is very similar to GitHub in the past, like if you're used to GitHub, this is like GitHub for all of the AI projects and AI apps.
[00:31:22] Al Shanmugam: So hugging Face Transformers is another place where you should go and look at what projects currently going on. And yeah, some of these vector databases, again, I can give you some links as. You should take a look at it because, that is a base of how do you build out a rag. I also use something called as a lang chain and LAMA Index.
[00:31:40] Al Shanmugam: Again, those are pretty cool. Take a look. You don't have to have a lot of insights about these technologies to run it because that is the core part of, ai, right? To make it easy for everyone, right? If it is complex so then you. So basically you should write a comment to that, service that in the US is not three W anymore because, in the past, like technical systems are [00:32:00] very complex, not anymore.
[00:32:01] Al Shanmugam: All of these AI services and AI systems that I'm talking about are actually made easier for every enterprise or every customer, right? To become an AI expert. Yeah, so go for that.
[00:32:11] Christina Ellwood: We will make sure to include the links in the episode summary and so people can take advantage of those.
[00:32:18] Christina Ellwood: As we wrap up, what would you like our listeners to take away from our conversation today?
[00:32:24] Al Shanmugam: Yeah my takeaway would be yeah, think about improving, the product and customer experience using, some of these goal-driven systems, right? Because in the past we try to reach a goal, in a human way, right?
[00:32:37] Al Shanmugam: We set a goal and then we come up with some strategies, try to achieve it. Right now, I would say use AI to do that. So that what'll happen is AI will see the goal is reach or not. And then it'll change its strategy and then start doing it again. So basically for us, if we wanna do implement multiple strategies and, try to reach the goal in six months, ai, I can continuously do [00:33:00] that and reach it in 10, 15 days, right?
[00:33:02] Al Shanmugam: So think about, a goal setting and then using AI to reach the goal.
[00:33:07] Christina Ellwood: Love it. That's great. They got good advice. So Al Shana gum, head of product for EchoStar, thank you so much for joining us today on AI Realized and we are looking forward to having you on the stage at AI Realized Summit in November.
[00:33:23] Al Shanmugam: Absolutely. I love the conversation. Thanks for having me.
[00:33:26] Christina Ellwood: Thank you for doing this. It's been great.
[00:33:29] Al Shanmugam: Cool, [00:34:00] awesome.