Why Most AI Projects Fail Before They Start

[00:00:00] Gwenevere Crary: Welcome to Scaling with People, your weekly playbook for turning chaos into compounding growth. Each week, we go under the hood with battle-tested experts in all areas of business, from marketing to sales, operations, finance, and people, plus product and leadership, to unpack the plays, numbers, and systems that turn chaos into compounding growth. Learn straight from founders and experts who’ve done it and continue to do it successfully. There’s zero fluff, just moves that you can steal immediately. This podcast is brought to you by Guide to HR. Guide to HR: Human expertise, AI-powered impact.

[00:00:39] Gwenevere Crary: Welcome everyone to today’s Scaling with People podcast. I’m Guinevere Currie, your host, founder and CEO of Guide to HR. If AI is the rocket fuel, why are so many companies still crawling? In this episode, we sit down with Ghazenfer Mansoor, CEO of Technology Rivers and author of Beyond the Download: How to Build Mobile Apps that People Love, Use, and Share Every Day. Ghazenfer and his team are helping founders turn clunky operations into high-performance engines with AI and smart software. We’re cutting through the hype to show how real 10X growth actually happens, not through more tools, but through building technology that works with your people and not against them. Well, welcome, Ghazenfer. I’m so excited to have you on the show today. Before we dive in, please introduce yourself to the audience.

[00:01:24] Ghazenfer Mansoor: Thanks for having me. I’m excited to be part of this podcast. My background is in running a software development company, Technology Rivers. We help businesses improve operations through AI and technology. We also help businesses build software products, primarily in the healthcare space. We work with physician entrepreneurs and health tech companies, building HIPAA-compliant software products.

[00:01:48] Gwenevere Crary: Awesome. Well, I can’t wait to dive into the AI conversation because I love it. I’m a junkie. But before we do, as a founder, I love putting my founders on the hot seat. Would you mind sharing a lesson learned from your own experience of building your business?

[00:02:04] Ghazenfer Mansoor: Absolutely. Before starting this business in 2015, I founded a startup in the recruitment space. It was a recruitment software. The lesson learned from my side was that we built a really strong, scalable product. We built the product first, then we started looking for a customer. As we started looking for a customer, we did get customers and went through different iterations, but there was a lot of stuff that we had to discard after that. We went after features, building a strong foundation for something that people were not going to use. A lot of energy and a lot of money was spent on things that were not useful. The lesson learned from my experience, and these are experiences I learned while working with so many founders when I started Technology Rivers, is to build something your customer needs. If I were to start another startup, I would not write a single line of code until I have a customer. So, the lesson learned is finding a customer first before building anything.

[00:03:16] Gwenevere Crary: And that’s really about knowing your audience and the problems they’re trying to solve, instead of thinking a problem is solvable and solving it, and then having no one to sell it to, right?

[00:03:26] Ghazenfer Mansoor: Absolutely, yes.

[00:03:26] Gwenevere Crary: I love that. Thank you so much. That is a big deal. I’ve seen some founders do this really successfully and some that have not. So, I hope for those listening that you definitely take this to heart because it is a way to make sure that you can create a successful business. As we dive into the topic of AI, most founders think AI equals efficiency, but what are they missing that’s actually limiting their growth?

[00:03:51] Ghazenfer Mansoor: AI is definitely considered growth, but at the same time, most people are not using it right. Most people are using it as a search. Many people are just relying on AI and assuming it’s going to give them everything. Unfortunately, AI gives you what you’re giving it. Your data is the key. If you don’t have the right data that you train AI on, you’re not going to get the right results. There’s a lot of hallucination, that’s the term you probably heard a lot. AI is going to make mistakes. AI is going to give you wrong data. AI is going to hallucinate because it does not have the right data. It is important that your data is clean, your workflows are mapped, and you are providing the right context of what you want. This is a common problem even outside AI. You ask a problem in such a way that everybody would understand it differently. You ask 10 people the same question, you’ll get 10 different answers because you’re not giving the right context, or you’re not able to articulate it right. You’re not going to get the right answers.

[00:04:56] Gwenevere Crary: That makes a lot of sense. Where do you see, as you’re working with founders and having conversations with other business owners, companies wasting money on AI right now and not really realizing it?

[00:05:14] Ghazenfer Mansoor: It’s the same part where companies are just assuming that AI is going to solve their problem by just plugging in, and the default assumption is that AI is probably just like ChatGPT or Claude. You just load your documents, and it will give you the right answers. No. Those are lessons learned along the way. That’s where we see the majority of AI projects fail. If 90 percent of projects are failing, what are the reasons behind that? Because most people don’t know how to use AI. There are different strategies that you have to use to build the right AI application. Again, as I said, context is important. Think of it this way: you have a huge set of data hundreds of documents. If you load those, do you think AI is going to give you the right responses right there? Even if you load five documents in ChatGPT and you query, you will quickly realize that you’re getting answers from one or two documents, not all the documents. You have to be a bit more specific because it’s not going to search all of them. You have to prepare your data accordingly and have your queries accordingly. You have to know what kind of responses your users are going to be asking. Prepare your data accordingly. Those are the mistakes I see a lot more commonly. This is similar to even non-AI-based work as well. You build something you don’t know, you made a mistake, you failed, you try again. That’s where you start spending a lot of money without realizing that you’re going down the wrong path. Another mistake is not understanding the amount of data it takes and the cost it would require. If you have tons of data, it definitely will require much bigger servers, and that’s where the cost comes in. Again, you can’t compare everything with ChatGPT.

[00:07:17] Gwenevere Crary: That’s very true. You have to stay on top of it and do your work. When you’re done, don’t just put it aside and think it’s going to work perfectly. You have to constantly be looking at it. When you say 10X growth through AI, what specifically changes inside a company if they’re able to execute on that?

[00:07:42] Ghazenfer Mansoor: Every company has access to different tools—AI and non-AI, whether it’s a CRM, recruitment software, HR software, any software that you’re running in your business. Your competitor also has access to that. Everybody has access. What is your differentiator? Your business, whether it’s HR, plumbing, healthcare, or any business that you’re running you and your competitor are the same. Are you a better service? Do you have better people? Is your cost low? All these differences that your competitor already has access to. What is the difference? The difference is how your workflow is mapped and how much efficiency is in your processes. That’s where technology and AI can make a difference. Your competitor can copy another software, but they cannot clone your workflows. That’s what you can do. How data and processes change hands from your lead coming in to your delivery finishing there are hundreds of steps that happen in between. There are many manual steps. Probably 50 different software pieces are being used. Data is being shared and lost along the way. Many people are involved. There are many manual steps, many bottlenecks, and many slow processes.

[00:09:04] What you can do to optimize those processes is where technology comes in. As you have that technology, that brings more efficiency, meaning you have fewer mistakes, your productivity increases, your profitability increases, your valuation increases.

[00:09:29] Gwenevere Crary: So, I’m thinking about this as a founder, and yes, I need to increase that, and yes, I have these problems and these different workflows I want to put into place. Maybe when you work with a client for the first time, how do you help structure the framework for a founder to know—this is the area you need to focus on first? Obviously, there’s the fast approach where you can implement something and have a quick win versus the longer approach, which has a bigger win but takes more time. How do you help founders find what to focus on? Because there’s a lot of noise in our world, and there’s probably a lot of inefficiencies in every organization. How do they pick what to focus on first?

[00:10:12] Ghazenfer Mansoor: There’s no one solution for everybody. Every company has its own DNA and its own processes. This is where strategy comes in what works for you specifically. Whatever works for you may not work for somebody else, and vice versa. You have to look at your specific processes and flows. In some cases, your people may have different capabilities and you’re really good at one thing versus another. Yes, there’s a bottleneck and improvement are needed, but you may not need something right away versus something that needs to be solved right now. Do you have a problem on the delivery side? Sales side? Nurturing? Whatever that process is. There’s no right or wrong. There’s no first or second. You just have to understand that the most important part is you don’t want to look at it as a whole project because that’s where I’ve seen people get confused. They say, “We want to automate everything,” and then they realize it’s a huge project and suddenly they’re scared. You want to have a very small process. Start with small wins, small processes. It could be starting with just one AI agent that improves one thing. It could be scheduling, automation in your email process, or anything. As long as you’re doing one thing at a time, gradually you’re building your confidence. Change is difficult. Your team will always have pushback on any bigger projects. There’s fear of replacement, fear of capabilities, doubts will we be able to do it? Will there be a mistake? Start with just one small thing at a time, and as you start implementing these small items, you realize they gradually become a bigger project.

[00:11:50] Ghazenfer Mansoor: You don’t even have to worry about integrating those on the first step. All you do is start automating small pieces, and that will eventually become a bigger project, and then gradually you can integrate and bring it into one tech.

[00:12:11] Gwenevere Crary: I love that advice because it also helps bring your employees along, right? The people are able to consume it, work on it, and it’s not so fearful. It’s just a little bit now, learn that, and then keep going. I love that. That’s great advice. But speaking about people and technology, what is the real constraint that you have seen inside a business? Is it technology? Is it leadership and decision-making? Is it the employees? What’s the real bottleneck on companies moving forward and getting AI implemented to help them be more efficient and effective?

[00:13:01] Ghazenfer Mansoor: It’s changed. Change at every level. It’s the fear of bringing any new technology because people have a fear of getting replaced. People have a fear of not doing it right. Some people have a different perspective. Some people don’t want to even learn. They say, “Well, we’re used to doing things the X way, now there’s another thing. “But AI is pushing people to have all these new learning. I would say it’s at a different level and different places. It’s important that you bring those people along with you by doing small things.

[00:13:37] Ghazenfer Mansoor: If you can just improve the productivity of one person with one small thing without having that person fear that this big change is going to impact their job, you realize that they are more excited when they start getting certain things done much better. We all have 10 additional things than we can handle. With AI, it’s not about replacing people it’s about empowering people. As you empower those people by bringing AI, you’re helping them doing things 10 times more than what they were used to. Suddenly they are more excited about it. It’s about how you bring that change into your people so that they are more excited and not pushing back.

[00:14:35] Gwenevere Crary: I think it’s also that you get a couple people doing it and then sharing how they did it, then you’re getting a little more buy-in. You get a couple more people doing it, and all of a sudden, you have this wave of the majority of your employees taking it on. I’ve seen so many times where employees are using AI, but they’re so scared to say that they are and how they’re using it that they don’t share that and help each other learn and grow. A culture that creates the environment of trust and safety so that you can communicate and share that’s where I’ve seen AI really flourish and take off.

[00:15:10] Ghazenfer Mansoor: Absolutely. One of the things I say is that, for example, on the content side, it’s being used the most because whether it’s writing email yes, if that part is even helping, at least that gives people some confidence. Beyond email, then you start getting into maybe proposal writing, maybe optimizing some other documents, or creating some steps and plans. Gradually you’re doing it without even having the fear oh, this is going to replace or become a big project. It’s about this one small thing that previously was taking three days, now you can do it in two hours or one hour. Suddenly that thing is much better and they can do a lot more things. You want to have their buy-in, and that is the most important part how people are having buy-in and they are more excited about it.

[00:16:04] Gwenevere Crary: There are so many pros to AI, but on the flip side, what have you seen as being the most expensive mistake that founders make when investing in software or AI into their business?

[00:16:21] Ghazenfer Mansoor: The most expensive mistake is not working on the foundation or the strategy, but just getting into the weeds and implementing something without knowing if this is going to work. Look at it this way: if we all look at our company and say, “Well, how many subscription software do we have?” You’ll be shocked if you do the search. I did when I looked at my QuickBooks I saw how much money we’re spending on those. Most of this software will have one or two overlapping features among the others. That means we are using so many different software, and we are not using all the features.

[00:17:04] Ghazenfer Mansoor: How do you sync the data among all of those? There’s a lot of integration effort. Do we really need all of those? Do we need one of them? Thinking through and knowing before what you want to build versus just getting it and then trying to make it work those are two different things. What I realize people are not doing and this is a common problem across many industries is we’re not doing the right discovery. We’re not really figuring out what we really need. What problems are we trying to solve? We need another HR software, recruitment software, or CRM. Why do we need it? What problem are we trying to solve? Not saying that you don’t need it, but what features do you need? Each will have some overlap. By looking at your specific needs and workload, you realize what you need, which software you need, and where you need AI. Even considering do you need Claude? Do you need OpenAI? Do you need Gemini?

[00:18:06] Ghazenfer Mansoor: Each will have some features and each of these are good in certain things, not everything. Do you need on-premise? Do you need cloud? Once you start building and investing in cloud, and then you realize we’re not even allowed to deploy on the cloud and now we need an on-premise solution, suddenly all of those things are gone. All of your R&D effort is gone, or vice versa. Knowing your boundaries, knowing your regulations, compliance requirements in your industry all of those make a difference. In our business, we work a lot with healthcare, so obviously HIPAA compliance is a big part. Any tool we use, we have to look at the main thing: Can they sign a BAA? What is their HIPAA-compliant versions for whatever whether it’s analytics, whether it’s any tracking thing? All the tools we are integrating, and you realize later that suddenly the cost is too high.

[00:19:09] Gwenevere Crary: I completely agree. I see, on the flip side with the people, that founders and CEOs will go out and say, “Everyone use AI or get off the boat, get off the ship,” whatever you want to call it. But they don’t actually have the strategy behind well, what does that mean? What are the guardrails? What are the systems and tools you’re giving your people that they can execute on? What does good look like for them? What does that mean regarding—well, everyone’s using AI in some capacity to write their emails. Okay, is that sufficient? Is that enough to stay on? Or are you talking about workflows and agents and all these other things? Part of what you were sharing is the strategy of what are you trying to solve, and how are you going to solve it? And then what does that mean for your employees? What do you expect your employees to do? Give them that roadmap and those guardrails, and then let them go and execute on it.

[00:19:51] Ghazenfer Mansoor: You raised a good point about guardrails the governance and ethics part, because people are blindly uploading data into AI without realizing the consequences. It’s important that as your business is growing and you’re touching your customer’s data and using AI; how do you use it? You have to have that governance. You have to have those policies so that the data is not misused.

[00:20:26] Gwenevere Crary: I’m curious from your perspective if a listener was to go out and execute an automation today, and I know every company and business is a little different, but is there one process that you see consistently over and over again that every company could automate, and maybe even automate it this quarter or this year?

[00:20:46] Ghazenfer Mansoor: That’s a hard question. I can’t think of anything on top of my head. But I think the basic business operations things that everybody can do you don’t even need to write code. Many of those could be simple AI agents that you can use with cloud code or any of those. For us, we use it everywhere in our email, in our proposal writing. Those are the areas where, when your lead generation is doing some research or any of those things, those are the ones that you can simply start with, and there are many that you can find available online.

[00:21:23] Gwenevere Crary: I would say start with very simple, basic ones where you can really see the value and you start seeing the time difference.

[00:21:34] Ghazenfer Mansoor: Mm-hmm. Yeah. I was thinking maybe something like your engagement with your clients, with your employees, with your vendors. I know so many founders one of their hot buttons is the finance side, invoicing and getting the invoicing right, and then number two is following up with those clients that don’t pay on time to get them to pay. I know so many finance teams that just spin their wheels constantly reaching out and trying to get people to pay those invoices. There’s got to be a way to somewhat automate that.

[00:22:12] Ghazenfer Mansoor: There are a lot of those smaller agents being created. We build so many internally some for our own use, some for our customers, primarily more on the healthcare side. You’re right. Those are the small use cases that you have to identify. What is your bottleneck? For example, in our case, that may not be a problem. In your case, it may be. If you need that, what are the things? How different is it? In some cases, your automation is already there. In some cases, you do need more custom work to do certain things.

[00:22:48] Gwenevere Crary: That makes sense. I know you are the author of Beyond the Download. What did you learn while writing that book that most founders ignore?

[00:23:00] Ghazenfer Mansoor: Beyond the Download: How to Build Mobile Apps That People Love, Use, and Share Every Day I think this is one of my passion projects that I started three years ago. It’s based on lessons learned that we created over 60 different applications as part of my business and my own experience in mobile even before starting this business. Over the years, I’ve built so many different mobile apps and seen them grow and fail. This book talks about different strategies for how to build a mobile app not just another app that’s downloaded on your phone, but one that gets retention and people come back to it again and again. How do you engage those users?

[00:23:41] Ghazenfer Mansoor: Those are the different strategies that I talk about. I think the bottom line of all of this is you want to build an app that is remarkable, which has greater user experience and greater design, but at the same time is easy to use and keeps people engaged. That’s the most important. How do you have different ways in the app that keep it exciting for users? It could be some of those online strategies and even offline strategies that may not be part of your app, but they’re related to the app. A lot of those principles can be applied to non-mobile applications as well.

[00:24:23] Gwenevere Crary: That’s so true. I didn’t even think about that. As we wrap up today’s episode, if we were to give a founder 30 days to complete some AI workflow of some sort, where should they start with AI to see real results in their business?

[00:24:40] Ghazenfer Mansoor: I would say go back to AI for that advice as well. Maybe ask Claude, and Claude can give you a roadmap and a plan to start your first workflow. You have to first give your own context who you are. If I’m putting my context in ChatGPT and my Claude knows about me and your case, a lot of that information about okay, who I am, what company I work with, what kind of stuff I’ve done, what can I do? Having that right context, then it will give you direction. There are certain things you want to do, certain things you don’t want to do. This goes back to some of the mistakes we’re talking about, because AI is not about just asking a question it’s about giving the right context.

[00:25:27] Ghazenfer Mansoor: Sometimes it takes time to tweak that. I treat AI as my thinking partner or my assistant, where I keep giving instructions, I get something, and then I say, no, I want this thing. I keep optimizing based on that. Your plan is also one of those that could be optimized. You want to know what workflow because your AI knows and has all the memory about what your business does, so it’ll give you some ideas of that. I did that for my own business as well. But that doesn’t mean I’ll just implement all of it. I narrow down based on my preferences: here’s what I want, here’s what I don’t want. Can you remove this? Can you change this? Going through this iteration helped me get to what I wanted.

[00:26:08] Gwenevere Crary: That’s so smart. I never really thought about it from that perspective. I use that all the time how do I market this? How should I communicate this? I use it as a thought partner. For those listening, if you haven’t used AI as a thought partner, that’s what I would push you to start doing in your first 30 days: just give it the context. You use Claude, it sounds like. I use ChatGPT. I love the project functionality because you can really narrow the focus of what it’s looking through in terms of how it’s going to answer you. But then don’t just throw in the same information to Claude or Gemini or Perplexity or whatever you use. Maybe try using two or three AI and get maybe different results, but ultimately combine that together, and that could get you your roadmap.

[00:26:59] Gwenevere Crary: Great suggestion. Really appreciate that. For those listening, I hope you got some great tips on how to get AI flowing into your business. Don’t be scared. It is a great tool that can help make you and your team more efficient and effective so you can grow your business 10 times faster. Thanks so much for joining us today, and we look forward to having you on the next podcast. Until then, have a great afternoon and day.

[00:27:25] Gwenevere Crary: If you got value from this conversation, do me a favor and share it with someone building something big. I’d love to hear your take. Drop a comment, shoot me a message, or start a conversation. Don’t forget to subscribe so you never miss the bold, unfiltered strategies we drop every week. I’m Guinevere Quarry, founder and CEO of Guide to HR, where we help high-growth companies scale smart with people-first strategies and AI-powered systems that don’t just keep up they lead. If you’re billing fast and want your HR to move faster, head to guidetohr.com and let’s talk. Remember, scale isn’t just about speed it’s about people. Until next time, have a great one.