Step Zero: What Comes Before Every AI Rollout in B2B

And one of the things that we've discussed is how there's a real gap between how AI gets talked about online, and what manufacturers actually ask you to build. I know you're in those conversations. What's the biggest disconnect that you see?

Kulmohan Makhija: Sure. Before I answer that, Aaron, thanks for having me on the podcast. This has been a long-standing thing. I've been following the way you have been trying to create genuine content for business owners, and this was the most intriguing space for me as well, to understand and discover newer findings in the market.

Aaron Sheehan: Thank you for being here.

Kulmohan Makhija: Coming to the question, I think the biggest disconnect is that online, AI is often discussed as a destination, right? In customer conversations, it's simply the other way around. They're looking to solve a business problem, and for them, AI is sort of a tool.

So nobody has asked me for an AI agent because it's more fashionable. They come up with something which is much more practical. Can I reduce the time it takes to onboard a supplier? Can I stop my operations team manually reconciling inventory? Can my customer service team stop searching five different systems for one answer?

So AI usually enters the conversation once we have understood the operational bottleneck.

Aaron Sheehan: Perfect. So, the CEO of Azilen, Naresh, has a line I read yesterday in a Forbes article. He writes quite a lot for Forbes Business Council. And he had a quote that stood out to me, and he said, interfaces attract attention, but systems create enduring value.

So related to that: when someone is coming to you excited about AI, are they chasing the interface or the system? And probably, for a services business, the important question is which one are they actually willing to pay for?

Kulmohan Makhija: This is a line we go by in our consulting business, largely, when it comes to sketching solutions for our customers. So initially, clients are usually attracted by the interface, because that's what they experience, that's what they see. They see copilots, conversational interfaces, AI search. But very quickly, the discussion shifts to the actual outcome.

They realize that a brilliant interface sitting on fragmented data doesn't solve anything. The organizations that are investing seriously are asking how much AI fits into their operational system, because that's where the long-term return comes from.

And in the majority of the cases that I have seen, the most complex AI solutions sit behind the scenes, more silently. You don't look at their interface, you don't open them until something breaks. So if everything is working fine, you just enjoy the automation and the intelligence it brings, rather than looking at the interface.

Aaron Sheehan: Yes, I completely agree with that. That mirrors very much our philosophy and our roadmap around where we are adding AI. He also sort of framed AI as working alongside engineers, rather than replacing them.

Certainly, I guess a hot topic is, is AI making people more productive, or are your customers buying AI hoping to reduce headcount? And that goes to the where is AI fitting in the organization question.

Kulmohan Makhija: Sure. So I'll come from our point of view. What we have seen in the business lately, in the last two years or more, since the AI bubble has started to expand and is now getting into our day-to-day lives — interestingly, very few companies actually begin with headcount reduction. That's probably sort of the media narrative and the PR for a lot of AI platforms, more than the boardroom narrative.

What they talk about is scalability. They want the same team to handle more complexity, more orders, more product information, more suppliers, without increasing operational overhead. So AI becomes an augmentation story, rather than it becoming an automation story.

Aaron Sheehan: Exactly. Yeah, growth without the overhead is definitely a part of the conversation that we hear and have regularly. I'm curious, how long has Azilen been working in the AI space?

Kulmohan Makhija: I would say AI, probably, for the last four to five years. And it has been an evolution, starting from data first to ML, then to generative AI, and now to agentic AI. But yeah, it has grown over the last four to five years now.

Aaron Sheehan: Right, yes, there are many different flavors of AI, we've talked about that on previous episodes. So you've been in it for long enough. I'm curious, have you ever seen an AI project disappoint a customer? And if so, was it the technology, or was it something more human-related or operational?

Kulmohan Makhija: Usually it isn't because the model wasn't clever enough, right? There are certain cases where we see that the implementation did not fulfill the business KPIs or the OKRs.

But in our finding, it is never usually the model to be blamed. It's because the organization expected AI to compensate for years of fragmented process and inconsistent data. So the KPIs and OKRs you set as an expectation from AI have to be more realistic, and they have to increase gradually.

So AI accelerates whatever environment you put it into. If the underlying process isn't mature, AI simply exposes that faster, rather than solving it on the very next day.

Aaron Sheehan: Yeah, totally agree. You mentioned agentic as sort of the latest iteration of what AI is, right, through all of the different generations of that.

AI is a big term, and what we found is that it used to be sort of a weird black box. You could say the word algorithm, and people would just, like, their eyes would glaze over, they're not sure what you were talking about. It feels like, to me, the maturity of what buyers want is improving. You've said that AI is no longer kind of a black box, so they know what kind of automation they're looking for. They know what it's capable of, to some extent.

But agentic, on the other hand, nobody yet that I have seen can quite agree on what the word means. So if people don't know what they don't know, what is agentic capable of? What is it that they should be asking for that they don't know how to ask for?

Kulmohan Makhija: I would, then again, re-emphasize that mature conversations today are around outcomes and automation. Things like document processing, customer support, inventory visibility, pricing assistance. Businesses understand that these are the areas to be solved, because they feel the pain every day.

Now, agentic AI is a bit more different. Most customers don't arrive asking for agents. They describe an operational challenge, and together we realize that an agent could probably orchestrate this better, because it has to deal with multiple systems behind the scene.

So if I was to put it from a comparison standpoint, consider this as generative AI in 2021, when people were not very sure of what this can do, and the only capacity everyone was considering was that, okay, it might produce a few images and a bit of content here and there, but nothing beyond it. So agentic AI currently is being taken into account as, okay, this might automate one workflow in one system.

But the capacity goes beyond that. Making decisions, working with multiple systems, even where connections are not very well established. And then also taking augmented decisions on top of the data there.

Aaron Sheehan: No, I completely agree, and the level of trust that I've seen manufacturers have in agents to make decisions… we'd asked a question of our customer advisory board a few months ago. In person, we had everybody in a room, and asked them, how many of you would be comfortable with AI or an agent writing data to your ERP?

Nobody, right? So it's very much an augmentation story and an automation story, but it's not yet a place where, I think, on the manufacturing side or the distribution side, companies are quite comfortable with fully autonomous robots creating orders, creating fulfillments, and all the rest of it.

You'd mentioned, we had spoken previously, you had a story about a customer running eight different source systems, trying to put a dashboard on top of that. And obviously there was some AI framing in there. To some extent, dashboards have been around a very long time. We've had REST APIs and GraphQL APIs and SOAP APIs for a very long time.

What is it about sort of the need to integrate that is now coming up as, well, we need MCP? There's an MCP problem now. Because to some extent, it's the same API problem that has existed for a long time. What's different about the AI lens on this integration story, and the specific ask around MCP, Model Context Protocol?

Kulmohan Makhija: Sure. So let us take the example that you quoted just now. You were talking to people, and they are not very confident that they would want autonomous agents to write into their ERP, or even take certain guided decisions, pre-configured decision-making steps, to reach there. Because the governance framework does not exist, and you probably don't know whether the action that will be taken can be controlled, can be supervised, can be audited.

So MCP helps in all of that, in a nutshell. And I would not say it is an answer to everything, but it is the first step to build on the later stages, right? So I don't think customers wake up and come asking for MCP. What they are really asking for is a secure orchestration process across multiple source systems.

MCP becomes an important conversation because it gives organizations confidence that AI can interact with business systems in a governed, secure, and auditable way. It's less about replacing APIs and more about creating a structured way for intelligent systems to use them.

Aaron Sheehan: That's exactly right. We have released MCP coverage for Oro in the last several months, and the way that we're describing it is we're giving something for an agent to reason across. It's like building a knowledge graph for a system, whether that's a very single-threaded agent trying to pull orders out, or create orders, or pull customer data out, learn something about the business.

The MCP server is providing, as you said, a schema and a structure for them to do it that doesn't require making lots of API calls, and then taking payloads, and then guessing what it's looking at. It's sort of telling the third-party system what you're looking at. And it does make, I would say, advanced integrations probably simpler and more effective, and with less custom development, I think, required.

It's interesting, we both, I think, started hearing it only from customers in the last few months, the MCP acronym. You said people aren't really asking for it, but are you finding that buyers' expectations have changed with what is possible?

Kulmohan Makhija: I think so. There is a lot of education nowadays on AI and the possibilities and the newer use cases, right? We see content being published every day, in millions and billions, around it, and now the buyers are equally educated. I would not say that everybody understands MCP and agentic AI, but usually, let's say, the innovation teams do.

Most customers do not anyway buy MCP, right? They buy outcomes. They want AI to work reliably across their ERP, CRM, commerce systems, operational systems. Whatever orchestration happens through MCP or any other architecture is often secondary to them.

Aaron Sheehan: Yep, completely agree. I am on LinkedIn a lot. You may be too, I don't know, possibly. I think it's part of the industry that we're in. And certainly, I consume a lot of thought leadership and speculation and analyst coverage around agents that are buying and selling autonomously. I've been on multiple panels and webinars this year talking about this exact topic, talking with our customers.

A lot of what you're seeing is not necessarily that. You've talked about integration, you've talked about automation. It's the agents that are replacing manual work between systems, but they're not running a full end-to-end transactional cycle completely on their own.

When you see a mature business, let's say a manufacturer, deploying an agent inside their stack, what kinds of things do you see the agents doing? What kinds of tasks are they appropriate for?

Kulmohan Makhija: Sure. No, I would agree that most enterprises are not asking for a fully autonomous or entirely autonomous commerce, largely, let's say, for that example. They are asking to eliminate repetitive coordination if an order is delayed, if inventory changes, if pricing needs validation. Those are the exact kinds of decisions where an intelligent agent can support. Now, this is not just exchange of data between two systems. It also requires a certain sort of intelligence on top of it.

And these are sort of minor decision-making that you would want to probably send to an AI agent today. I'm not going to say that the world will definitely not move towards an area where everything is more headless and agents will communicate and do an autonomous process, but that's too far-fetched of a future, at least in my opinion.

So currently, what customers are wanting is not just interaction between the systems, but then an intelligent decision layer to at least low-risk decision profilings, right? So there are two levels of decision profilings in the majority of the enterprises, which is low risk and high risk. Wherever high risk is involved, there is a human in the loop, and then you have somebody validating what the agent wants to do, what the agent wants to automate. And for low risk, you definitely believe that, okay, this can probably be passed through, if you have your governance framework, if you have your auditability and scalability all placed right in the MCP or the agent development.

Aaron Sheehan: No, completely. I think that makes a lot of sense. Thinking through that sort of shipping and fulfillment and inventory management example, I like the framing of low risk, high risk. Low risk, I can maybe let the AI, the agent, reason on its own and make a decision on its own. Anything high risk, I flag, I alert for a person to do something.

We have very similar processes for SmartOrder, for instance, in Oro. Works that way. And that's all configurable to some extent. Businesses can make their own decisions about what is low risk and what is high risk.

Do you have any examples of a specific flow that you've seen that someone would describe as either low risk or high risk? If we can make it concrete, what's a good example of that?

Kulmohan Makhija: Sure. Let me take a reference of a recent client, right? So we believe that wherever judgment matters, that is more high-risk situations. Where a judgment decision can probably change the outcome of the business, could have commercial impact, could have probably a compliance impact. That is where judgment is most crucial.

AI is very good at gathering information, applying rules, and recommending actions. So humans are still better in handling exceptions, commercial negotiations, and decisions involving risk. The organizations we are working with aren't trying to remove people, they're trying to let people spend more time where human judgment adds value.

Now, take an example if probably a manufacturing hub is trying to automate a cycle, based on the number of orders, of how the shipping and inventory control needs to be done. You probably would want a certain level of agentic decision-making happening here. If, let's say, the order volume is exceeding a certain threshold, what is the inventory threshold that you should be managing internally? And how is shipping to be alerted to ensure that the delivery happens within the customer lifecycle values?

Now, in these particular cases, there are two areas where humans will have to make decisions, because this will have commercial impact. You can't hold a lot of inventory, you can't spend on shipping, until the order reaches a certain stage and you are holding a certain threshold of inventory. So these sort of mixed scenarios is where we see automation coming in, and also human in the loop coming in.

Aaron Sheehan: Completely. And it's interesting, too, because some of those examples have been common in what you might call B2C eCommerce for some time, where there's a free shipping threshold, or there's some kind of logic around a free gift when I add. But it's challenging in a B2B sales scenario, because often the shipping cost, let's say the fulfillment cost, isn't known at the time that the items are going into a basket and then going into an order. So it's downstream that that information is known, and then a decision can be made about… yeah, completely agree with that.

Aaron Sheehan: I know you've mentioned, actually, I think from the very beginning of this episode, data quality as a barrier to doing agentic anything, let's say, or automating anything. And I'm curious. A line I use often is that garbage data, automated, creates a high-speed garbage cannon.

How do you and Azilen think about data quality on an AI project? Is that sort of a step zero or step one of your process? Do you ask those questions up front, and do your customers maybe have an idea of what their data quality is before they try to automate it? You can take that from whatever angle you want.

Kulmohan Makhija: Sure, yeah. No, I would agree with you, data quality isn't step one, is it? It is a step zero. Anything that you want to build in terms of AI, automation, machine learning, NLPs, everything comes from structured, clean data.

I would say that now, clients are much more matured and much more educated in terms of why data quality is more important, and you see that within the conversation as well. Of course, they're not very well aware of how to assess it by themselves, and that's where companies like us, or consultants like us, help them understand, with a certain audit and a certain review.

But we have seen organizations with fantastic AI ambition discover that product data exists in six different systems. Customer records don't match, supplier information hasn't been updated for years. AI doesn't solve that, it simply encounters it faster. So the companies making the fastest progress are the ones treating data as a strategy, rather than treating AI as a strategy.

Aaron Sheehan: That's so true. That's so true. We've talked to many people around how data governance is a program, not a project. Like security. It's not something that is a one-time project to clean up a particular set of spreadsheets, and then you're done, or a particular database, and then you're done. There needs to be oversight continually going into improving and automating it.

This is one place where, on the product side, I think Europe in general is taking a more… well, it's common for Europe to be more regulatory than the US, I'll say. I think that's a fair statement. But there's something that went online this month, I think. So the EU introduced a concept a few years ago called the DPP, or Digital Product Passport.

It's been talked about for a while, but the point of it is that it's a sort of comprehensive digital record that talks about a product's lifecycle, what materials are in it, ingredients or components, and environmental impact. Obviously this is part of eco-friendly regulation, but it has a huge impact on how manufacturers know what they're producing, and how they describe what they're producing for use further down the supply chain, which is something that is often not well-resourced, I think, at least in some industries.

So I think the initial registries were set up this month, and then it's gonna roll out by industry, I believe, over the next couple of years.

Often when we're having conversations with people around Digital Product Passport and data, it gets treated like compliance theater a little bit, like a lot of, I think, Brussels regulations. Arguably, it's really a data problem. It's forcing one authoritative record of what a product is. And most of them, many of them don't have that data. Is that true? And do you guys see Digital Product Passport showing up in your practice at all?

Kulmohan Makhija: No, I would agree with you, it is 100% a data problem, right? So if you will talk to anybody in Europe right now, because the compliance is going to hit starting February 2027, and the framework for it, and the test environment, and everything is out now. Of course, this is going to be sort of a phase-wise rollout. A couple of industries would be hit first, and then the others.

But the idea is that all these sort of data sets were already existing with large-scale manufacturers, but not centralized in a place, not clean in a place. Certain with suppliers, certain at the warehouse level and the shipping providers, certain at their own manufacturing side, and certain with their retail partners.

Now, compliance is simply the trigger here. DPP is forcing organizations to answer a much bigger question. Do we actually know our products well enough? That includes origin, composition, suppliers, lifecycle, sustainability attributes. Those questions existed before DPP. The regulation is just making them unavoidable.

And because this is going to hit them very soon, we see a lot of conversations coming around this, and people equally being confused that all of this data is there, scattered enough in different places, different people, different parties. Now, how do I get it together and start making sense out of this?

Aaron Sheehan: No, that makes sense. I think you ran a DPP event last month, was it?

Kulmohan Makhija: Yes, in June, yeah.

Aaron Sheehan: In June, yeah. I know the EU registry just went live. Are the conversations changing at all as a result of the calendar and kind of the regulations? And what are you hearing?

And apologies for all of our US listeners, this probably doesn't apply yet to us. I'm sure it'll eventually get here like most regulations do, but right now, I would say, from a regulatory standpoint, this is relevant in the EU. However, I would say, hey, US listeners, if you're a manufacturer and you're listening to this, it's still a good idea to have this data and have an authoritative record of what you're maintaining. There are a lot of benefits to your future automation and expansion from doing this. But you're not going to get fined for it… yet.

So, sorry, the question was, has the tone of the conversation shifted in the EU for you guys around DPP?

Kulmohan Makhija: Definitely. I think we have been very actively understanding and watching the DPP conversation as it has shaped up in that part of the world. And I would say 12 months ago, the conversations were more educational.

Today, they are much more practical. Organizations aren't asking, what is DPP? They're asking, where do we begin? How much of our existing infrastructure can we reuse? That's a much healthier discussion, because now it has moved to an adoption level, rather than an exploration level.

Aaron Sheehan: I'm curious, do you think that… I kind of made this assumption, but maybe you can tell me if it's true or not. If a company does the work necessary to implement a DPP platform of some kind, and gets their data in order, does that open more agentic use cases for them, in terms of what they can do with that data?

Kulmohan Makhija: I actually think that's one of the biggest hidden benefits, right? Once you're invested in structured, trusted product data for DPP, you have already created the foundation for AI models, for AI needs. Because now you have all the data needed for a product lifecycle, starting from the sourcing of material, to the manufacturing, to warehousing, to the retail supply chain, to the end of life cycle. Now, that could be recycling or otherwise.

While you have all of this data set for a particular product, and you have created a product passport, imagine the amount of automation that can come in. Now, be it in any form. This could be simple RPA-level automation that you can generate, intelligence that you can generate from this for your business, and also AI-level use cases that you can create out of this. This can help you in better supplier collaboration, better automation, better recommendations, better search. So there are a lot of use cases this will uncover as we go along the journey.

Aaron Sheehan: No, I think that's exactly right. I'm thinking back to something you'd said earlier around the low trust, high trust framing. One of the big levers a business can pull to move an automation from low trust — meaning someone has to be involved, we simply don't trust the machine to make these decisions — is data quality.

I have seen many, many times, especially for B2B, where product data is understood in people's heads. And maybe in paper catalogs, or paper sheets, or spreadsheets, or digital files, they're scattered in a lot of places, and there's a real hesitancy to automate any kind of quoting, or sales, or fulfillment, or customer service inquiry, simply because the data is not accessible to a digital system to automate it at all.

And so what the automation becomes is a request comes in, and I send an email to a person to pick up the phone and call that potential customer back, and say, yes, this product will fit your use case, no, this product won't fit your use case. And the reason that has to hit a person is simply because the data quality is not high enough to then, frankly, in a high-trust way, automate any kind of processing around it. So I can absolutely see that.

So, I want to go to some inside baseball here. Azilen is a solutions integrator, you're a digital engineering firm, you do a lot of work for companies around the globe. Because you're a services firm, you're selling AI that automates system-to-system integration, which — SI means System Integrator — I am curious, how do you sell "we'll do less of what you pay us for," and have people believe that?

This is a question I have not heard asked on an industry podcast. It's not a question that I think a lot of people are openly talking about, but it's absolutely a conversation that is happening, I think, in every boardroom, in every meeting. How does that work for Azilen? How do you sell "we'll do less of what you're asking us to do"?

Kulmohan Makhija: Right. I think you have asked the very critical question that the majority of my SI partners and SI colleagues are facing every day. So the nature of implementation work is changing. Customers aren't looking for more development hours, they're looking for measurable business outcomes.

Times have gone when people come up and ask you for tech resources, and then they will build it. They come up with a very specific business challenge. Either they have a KPI mapped to it, an OKR mapped to it, and our role is becoming less about writing code, and more about helping clients build an operating model that's intelligent, scalable, and adaptable. That's a much more valuable relationship, in my opinion, because now we are trying to solve a business outcome that will, in turn, give them a particular green signal, be it in terms of top line, in terms of profitability, in terms of whatever bottleneck it was solving. So now, SIs are also measured from a standpoint of business outcomes.

Aaron Sheehan: Yes, exactly. I see this playing out in terms of billing for value delivery instead of hours logged, as a big piece of it. Does leading with AI change your pricing model and how you're selling? Are you seeing the same thing?

Kulmohan Makhija: I would say it has definitely changed it. The conversation increasingly now involves operations, digital transformation, and business leadership, not just IT. With IT, it was a plain vanilla pricing model that used to exist, and now it is much more complicated, because you have to be a subject matter expert, you have to understand their process, you have to understand the technology expertise as well. 

The clients will no longer come and tell you what technology to build it on, what model to use, and otherwise. So AI touches business processes, and naturally the discussion becomes much broader.

Aaron Sheehan: Yeah, I totally see that. I'm curious, when you're in the room with a firm, and you say, hey, AI can handle this problem — they come to you with a problem, you say, AI can do that — what kind of reaction are you getting from people? Are they excited about that, or are they anxious about that?

Kulmohan Makhija: I would say a mix of both. You usually see both reactions in different situations and scenarios. The first question is often, what happens to my team? Because that sort of scare is still very prevalent in the market in terms of AI adoption.

And the second question is, how quickly can we start? Organizations that do this well involve their people early and position AI as removing repetitive work rather than replacing expertise. So currently, the initial starting point is a mix of both reactions, but it usually transforms into, okay, how early can we start now, because they understand the value it will bring to the table.

Aaron Sheehan: Absolutely, absolutely. It's interesting, it's always the human — I think your CEO actually said that in an article, which was, the biggest variable in AI disruption isn't the AI, it's the human reaction to it. Clearly, you're seeing that.

If an executive is listening and feeling like they are falling behind, and a lot of companies feel like they need to be doing something about AI, but they're not sure what the something is, other than buying software — what's a good first step for someone who wants to be prepared for this more automated future?

Kulmohan Makhija: Let me start with an example. What I have seen largely in 2025, when the AI wave was riding and the use cases were going, people crazily hopped onto buying licenses of AI models. Somebody's using Copilot, somebody's using Claude, and gave access to larger teams — okay, build, build, build, and let's see what automation can happen. And the majority of those internal proofs of concept or MVPs failed horrendously.

And that also scared off a lot of enterprises into thinking, okay, AI might not be right for me. So my recommendation is: do not buy another AI tool immediately. Start by understanding your data landscape, and identifying one operational process that is creating the most friction every day.

If you can improve one meaningful business process using AI and trusted data, you'll build far more momentum than trying to transform the entire organization at once, which is very random.

Aaron Sheehan: That is really good advice. And I would say the other benefit is that if you can solve one workflow or business outcome well, you've likely built the foundation to solve the next one in less time.

Kulmohan Makhija: Decades faster, yeah.

Aaron Sheehan: Yes. So, okay, one prediction then. Twelve months from now, what do you think we will have stopped arguing about? Because it just became so totally obvious.

Kulmohan Makhija: I think we will stop debating whether AI belongs in enterprise commerce or not, which is still a larger conversation that I see with good quality folks, right? And they still believe that tech is just an enabler to the business, but I'm telling them that tech is now going to be driving the business. As much value as you put an emphasis on your manufacturing side of the business, you'll have to put it on technology, and largely AI. So I think that's a debate that will end in the next twelve months.

The question will disappear, the conversation will become much more practical. Which process should remain human? Which one should go autonomous? And how do we govern both? AI itself won't be a differentiator anymore. It's how well organizations operationalize the technology.

Aaron Sheehan: I think that's so true. We saw this with computing and the internet. We've seen this with the rise of cloud solutions. We've seen this with adoption of API-based integrations and all the rest of it. Things that started off as opinions, and one way that you could do it, have slowly over time, and sometimes very quickly over time, become norms that no one questions.

So, I really appreciate that, Kulmohan. I really appreciate your time with us. This has been very interesting. I always end each podcast with a surprise question to all my guests, which is, what is something that you have read, or watched — a book, a novel, a movie, a TV show, a podcast, anything that you like — that you would recommend to our listeners? And it does not have to be work-related.

Kulmohan Makhija: Yeah. To be really honest, something that pops up to my mind is a book called Everyday Creativity, which I have just started reading, and it gives us non-technical ways of solving everyday challenges. Now, this doesn't have to be business, this doesn't have to be your operations, this doesn't have to be anything. Anything in life that you face a challenge with, there are tools, and this comes from certain ancient methods that were used in Japan in the manufacturing era, certain methods that are now being used in the US as board-level decision-making. And then you apply it to your everyday problem statements, and probably be a bit more creative and objective in terms of solving and learning something. So I think that's one thing I would recommend.

Aaron Sheehan: That sounds very rewarding. If you can, send me the link, I will put it in the show notes, and maybe we can move some product while we're at it. Kulmohan, I really appreciate it. Thank you very much for your time. And if people want to contact you and Azilen, what's the best way for them to do that?

Kulmohan Makhija: The best is you probably can come to our website, or drop me an email. You can find me at [[email protected]] — while it is very hard to spell and pronounce, I'll probably have that attached somewhere in the description or the notes, and it becomes easier for them to contact us.

Aaron Sheehan: Absolutely, we will put all of those links in the show notes. Thank you so much, really appreciate your time. And listeners, have a great day.

Kulmohan Makhija: Thanks, Aaron, and thanks for everyone listening. Have a good one.