What you'll learn:
- Understand AI technology and what all of the hype is about
- How to assess different types of AI technologies and vendors
- A prioritization framework for building your own roadmap
Apoorv Dwivedi (00:12):
My name's Apoorv, and we're going to talk about building and AI roadmap. Oh, there we go. There's a sound, big booming voice. So this past weekend there was the Berkshire Hathaway annual shareholder meeting took place, and at the Berkshire Hathaway AGM, both Warren Buffett and Charlie Munger talked about AI and in fact, Warren Buffett. And if you know Warren Buffett, if you follow him, if you read his letters to the shareholders and those types of things, you'll know that he is not someone for hyperbole. He's the type of person that generally cuts through hype. But Warren Buffett said that AI is like an atom bomb. He said that it is already capable of so much and it has so much more potential. Charlie talked about it as well. In fact, if we go back through today, Heather, that was just on stage here. She talked about AI. Alan in his session, talked about Ai. Dan in his session, talked about AI. Gene Ger, who kicked out the day off with the workshop, talked about AI. So it's on everyone's radar. Everyone is talking about AI, but the question is, what is AI? What do we do with it and how do we make sense of it and actually take it and apply it to our firms? So first question here, just a show of hands. Who today is using AI in some way, shape, or form within their firms? Okay, so that's probably, I'm going to say that's maybe about 20% in the AI CPA, CAS benchmarking study. There's a copy available out there at the booth, but if you take a look at that, they said that I think it was roughly a quarter of the firms in the benchmarking survey are using AI already today in some way, shape, or form within their firms. A smaller subset of those, it's probably about a ninth of the firms overall are using some form of robotic process automation or RPAs or robots. Now, some of you may be saying, what is that? I don't know what that means. There's probably some of you. I don't even know how to spell AI, right? So what we're going to talk about today, what we're going to get into is I'm just going to give you an overview. What is AI? Understand the principles of how AI works and why it's important for you and how you can start looking at it and basically putting together a roadmap and a strategy for your firm. What I am going to get into is having some type of ethical debate about Skynet taking over the world. At some point in the future, we can get into the potential of what AI can do, and of course there's all sorts of legal ethical ramifications, but that's well beyond the scope of today.
(03:23)
So a hundred sixty eight, a hundred sixty eight times in. Now this seems like really old slides because they're from two weeks ago, and in the world of AI, things changed really quickly, but 168 times large enterprise tech firms. So basically your alphabet, meta apple, Microsoft mentioned AI on their earnings calls for Q1 of this year, 6 trillion. Now, that is the number that McKinsey puts on the value of the total economic impact of AI by 2030. Of course, they tend to inflate numbers. 80% though was a separate study that looked at the number of professional services firm tasks that will be in some way affected by AI. So I mean big numbers out there. Sorry about the small text here, but questions that you might have. What's all the hype about? What exactly is ai? Is it just a passing trend, right? Is it a threat? Where are the opportunities? And probably the most important is what is the impact potentially going to be for your firm and how can you get ahead of that?
(04:51)
Now, there's a picture of me there that is an AI enhanced photo. You can see me, you can see the photo, you can see that in real life, I don't look anywhere near as good as I do in that photo. So if AI can do that for me in my photo, just imagine what it can do for you and your firms. There's a QR code there if you want to connect with me or just look me up. Actually, I should add one thing. So a little bit about myself, the background there, whole bunch of background in accounting and financial services, product development, marketing market research, et cetera. But just purely by coincidence, we actually developed an AI power adapt using open AI for the accounting industry in 2021. And so we have a little bit of experience with us and said, purely by coincidence, this was just off the side of the desk during COVID looking for something to do, came across someone thought, Hey, why don't we play around with this? See what it ends up doing. So here's the thing, right? AI is all over the news. Everyone's talking about it, but it's not new. You're all using AI today. You all have some type of electronic devices and you've been using AI and you have been for some time. In fact, unless you're like living in a cabin in the woods, you've been using AI for a long period of time. Search engines, every single search engine you could imagine, if you've ever used Siri or Alexa, Bixby is the Samsung one. No one really knows what that is, but it's their version of that. Google Home, Spotify, Netflix, blah, blah, blah. You go down the list if you've ever used a credit card, right? Ever done any form of online shopping, ever been in a self-driving automobile? If you've ever crossed a border in the last probably, I don't know, nine to 12 years, you've used an AI app of some way, shape, or form. Now, in addition to the Warren Buffet quote, this is an older quote because this is from a couple months ago, but Sundar Pichai, the CEO of Alphabet, said, AI is the most profound technology humanity is working on, and he compared it to fire and electricity. Now of course, he's the CEO of Alphabet and he needs to jump onto the hype wagon to pump up the stock. But it is a big deal and folks like Google have actually been working on AI for a very long period of time.
(07:35)
Like I said, it's not the latest greatest. In fact, it's not even something new. Not going to go through all the dates, but really when you take a look at the timing, the term AI was actually first coined back in 1955, right? That's how far back it goes. The first chatbot was 1964, voice activated technology. If you think of where your Siri and those things are at today, 1978, 1988, this is actually probably the most significant development here. IBM published a paper on statistical methodology, and I'll explain in a little bit why that's important, but dates like 97, IBM, big blue beat Kasparov in chess after that came IBM Watson, which beat Brad Jennings in jeopardy and et cetera, et cetera. So what's changed with AI? Why are we going from the 1950s and 1960s to now? And really what it's come down to is, I mean there is the technology there, the quality of the output has increased exponentially, but that's not really what it really comes down to is the costs have come down at the same time. So on one side you've got the quality going up the other side, you've got the cost coming down. And I give a story to folks not to do with AI, but probably about 15 years ago I was at a financial institution and I led the institution in building a business case on customer database marketing and email marketing. So we went out and we looked at what was the potential applications out there for it, and we came across an application put out by IBM Eloqua that was used by all the large FI's in the world. We put together the business case and implemented that technology. At that time, email marketing database cost 1.8 million to implement, and that was before the first five years of support and maintenance cost. That exact same technology is available today for less than $30 a month. Now, I mean, has the technology changed? Sure, it's improved. Not a whole lot, but it has improved. But the big thing really has been that the costs have come down so dramatically, and that really is the same story. What we're seeing with AI, what was previously in the realm of super expensive supercomputing, I mean it still is, but the cost on that have really come down. So Heather was just talking about Jeffrey Moore previously. Jeffrey Moore is was a professor, Clayton Crenshaw and Jeffrey Moore at Harvard developed this theory around technology adoption. Some technologies make it, some don't. But really kind of what's happened in the last year is there's kind of been this leap made, primarily driven by the folks that open AI. So the folks that are behind chat, GPT when they've so-called crossed the chasm or back to the days of the fawns, jumped the sharks, so to speak. So when you take a look at the timelines, pre 1990s, it was really a lot of really high end academic military type applications. And then in the nineties you started seeing a small handful of super high-end technology firms, the IBMs of the world, those types of folks implementing and leveraging AI. In the two thousands, you had a whole bunch of developments and you started seeing folks like the Googles and those types using and implementing AI. In the 2010s. It basically went jumped from there to big tech. So at the slide I had at the beginning, all the big tech firms started using AI, and that's you've been seeing for the last, say, five to 10 years, the last year or so, what's happened now is AI's become accessible to the mass market, But I mean this is just the beginning. So Gartner puts out this what they call the hype cycle, and AI is still very much going up that cycle. So there is still a ton of room for it to improve, and just as people start to understand the applications behind it and where it can be edit, embedded and leveraged, it's going to increase even more. Now it's going to come to a point where it's like over-hyped, right? No question about it. And there'll probably be a little bit of a dropdown from that before it stabilizes again, but we're not anywhere near that peak point yet. So what is AI, I won't read this off, but essentially bit of a description there around intelligence systems performing tasks that were normally or previously done by humans. There's a whole bunch of stuff that falls under the realm of AI So you might have heard of terms, machine learning, deep learning, et cetera. AI is kind of an umbrella term for a lot of those different things. I like to kind of think of it in terms of a model. So if you look previously at, say for example, a spreadsheet, you have this model where you have input. So you put something into the spreadsheet or piece of software calculator, whatever it was. The machine does some type of processing and then it spits out in output. So that's kind of your traditional software model. So you have an input, you have processing, you have output. Usually it was kind of in the realm of you input some type of text. It does the processing, it fits out some form of text. Now, if you take that and take a look at what's going on with AI, you have that same idea. You have input, say if we use the example of chat GPT, it's being done in the form of text, but there's a whole bunch of processing that takes place and I'll explain that in a sec. And then it spits out a form of output in the form of tax. What's really different though is that middle part, the processing. So what happens in that middle part? There's two important terms to understand with AI.
(14:51)
So there's two important terms that you need to understand here. So first one is fuzzy logic, and the second one is machine learning. So when we look at that model with the spreadsheet where you put in the input, it does some processing. You have the output in that model. If you take a look at it, you have to be very precise about the information that you enter in terms of the input it has to meet a certain format. And the formula that you're putting in there calculates based on how you're putting things, how you've set it up and how it's looking at the data. With fuzzy logic, it goes back to actually that paper that IBM published around statistical probability models. So it's not a matter of a right or wrong answer, yes or no answer. It's a whole series of probabilities that are stacked up. So that's the important thing to keep in mind with AI when it's looking at data is it's literally hundreds, thousands, tens of thousands, millions of probability statements stacked up on top of each other. And so what the AI is doing is it's looking at a whole bunch of different data points and saying, okay, I think there's a 80% chance that this is what you're looking for. I think this data point is say 90% valid or associated with what you're looking for, and it does that with millions upon millions of data points. So what that allows for is processing of vague or imprecise data. Now the second thing is what's called machine learning. So if you take a look at those probability models, the systems can actually learn based on the output, based on how you react to the output that okay, this person liked the quality of the output and then asked a subsequent follow-up question, or no, they went back and they restated a question. So that means that my probability model was either correct or was either wrong and I need to adjust something. And so it's taking the machine learning and it's improving on those probability models and it's going through these loops over and over and over and over. Fuzzy logic machine learning. So now you take those fuzzy logic and machine learning and you say, okay, we don't just need exact text. We can have imprecise text. We can actually take voice, turn it into a data file, assess based on the probability model, what we think that voice is trying to say. We'll process it using the fuzzy logic and the machine learning and we'll have some form of output, and that output can take the form of text or we can just turn that into another voice file, right? We'll translate that text into voice or audio. You can actually even take it a step further. And so there's a lot of development going on right now with images. Images are just another form of data file. So whether it's a static photo or it's a video file or something like that, basically what they're doing is they're taking that as a form of input, turning it into a data file, assessing that file based on the probability models, and then creating a form of output. And the output can be text, it can be audio, it can be video. Right? Now, going back to the spreadsheet model, think about spreadsheets, right? It's not a single formula in a single cell. It's hundreds or thousands of cells connected, and that's just a single worksheet and there's multiple worksheets and there's multiple workbooks and they're all connected together. So you can have these different AI input processing outputs, you think literally of thousands upon thousands of these connected together, connected with automation triggers. So the output from one becomes the input to the next and so on and so on. So that's where it starts to get a lot more complicated and sophisticated. Now, what that allows you to do is what's called robotic process automation. So with robotic process automation or RPA, essentially what you're doing is you're taking that combination of automation and triggers and you're using that to repeat high volume tasks using the rules and triggers. So I mean that's really what it does. How it does it. Think of it just as a piece of software that's in your browser or that sits on your desktop. You do something on say, in your browser, and it's going to mimic that over and over and over again. And it's smart enough using the fuzzy logic and the machine learning to figure out what it is that you're trying to do. And it will continually improve upon that. So the robotic process automation, the RPA is basically just taking the same task over and over and over, but rather than it has to be in the form of that spreadsheet where you're inputting very precise text. It is using that fuzzy logic to do stuff on your desktop or through a browser or something like that. Excuse me, I'll just grab some water.
(20:58)
So I mentioned the app that we had developed. So quick side note on that, all it really did was we went out to the top 100 accounting firm websites with a RPA app and scraped all the publicly available data from them. And then we did that with all their social media profiles and their YouTube videos and things like that as well. And then we aggregated all of that. Now, could you have done that manually? Sure, but very time intensive, very low value to actually do that. So we have these robots, robots that are essentially just doing this over and over and over again all day long, every day, month after month after month. And so that's just a very simple application, but that's allowed us to aggregate over a hundred thousand pieces of accounting industry content that we can then take a look at and apply other label, other levels of analytics and insights to too. So thinking about the model top takeaways, you've got input on one level, which can be text, audio, video, trigger automation. You've got processing. Processing is being done through fuzzy logic, robotic process automation. You have output, which can also be text, audio, video or some type of automation. And that whole model is continually getting better and better with machine learning. So in a nutshell, in layperson's terms, that's how AI works. Great. So what does this mean for my firm?
(22:49)
Here's the thing, and here's what everyone was talking about earlier. This is not a check the box activity. We went away and we did this for our firm. This is something that you're going to have to get used to. AI is here to stay. It's going to be an ongoing iterative process for the future. Just get used to it, right? Heather was talking about change, right? No one likes change. They say the only person that likes change is a baby with a dirty diaper. Well, guess what? We all have dirty diapers now because of AI and what it can do. So when you think about implementing AI within your firm, there's a whole bunch of stuff there that needs to be done. And it's not just about the technology, it's about people, systems, processes, policies, all of that.
(23:41)
So let's look at what you need to do to put together your roadmap. Now, I apologize for this. I was trying to figure out different diagrams I had actually put in to, there's a tool called Dolly, which generates images from AI, and I actually got garbage out of that, so I wasn't actually going to use that. But in terms of putting together something, now, if you're within a larger firm and you want to be the champion of AI within your firm, this is for you. So build your foundational knowledge, build your support team, get a business case secure, buy-in and measure. Now, I think for most of us, I think a lot of us here are running firms of our own or are managing partners a little bit different. So the first thing is we need to make sure we're managing our risk exposure. So very first thing is you're going to want to go back and develop an organizational policy on the use of AI And I'm not talking about chat GPT, I'm talking about AI, right? There's an important distinction there. Cause everyone's focused around chat GPT, this is so much larger than that. So work on developing that organizational policy for what that is. I would say you probably also want to provide a foundation level of training for everyone within your firm. So basically the type of stuff that we're talking about here today, what is AI? How does it work? Kind of understanding the basics of it.
(25:21)
Also, make sure you're tying in your firm's policies. So you may already have policies around confidentiality, data privacy, client information, et cetera. This is the time to be reinforcing that, and it's a very rapidly evolving field. So you want to keep your finger on the pulse of any updates from legal and regulatory bodies. There's a whole bunch of task forces that are out there that are underway right now around AI So it, it's going to be involving, keep your finger on the pulse, but I would say don't wait for them to complete their work. Build your delivery team now, I mean it makes the most sense right off the bat. Make sure you're including operations, HR, legal, marketing, comms, folks like that as well. The other thing I'd say is this is something that you want to have senior leadership direction on. There was a time for those of you that remember back when social media was the next greatest thing, where it was like, okay, let's find the youngest person on the team and they're going to be responsible for social media. This is different. So this is not social media. Sure there, there's all sorts of potential applications, but the implications are much broader and deeper. And so you need to make sure that what you're doing on the AI side is aligning with your firm's strategy and direction, developing business case to secure buy-in. So Heather talked about OKRs. So organizational key results. Take a look at basically your firm's business objectives and where you want to go and make sure that what you're doing on the AI side aligns with those.
(27:15)
Probably the best place to start is to go through and do a brainstorming exercise. So have each of your departments identify a number of potential quick win projects. Right now, I would say with AI, this is not the time to be working on brand new product development, greenfield stuff. This is not the time to be looking at projects that have 24, 36 months payback periods just because things are changing so quickly. So it's probably best to start off with quick implementation projects. Put together a list, the brainstorm list, assess, pick and prioritize a handful of projects that you can implement. We've got a spreadsheet that we use for this that we'd be happy to share. So just shoot me an email on that. And then I'd say as well, risk mitigation, you want to make sure you're controlling your risks. So have a mitigation plan in place. Make sure you have project risks and controls in place as well. Look for that agile cadence. So like I said, don't look at 24, 36 month projects. Look for projects that you can implement in four weeks or less. Pick two or three of those, implement them, measure them. Look at integrating with the existing tech stack if you can, and make a decision afterwards to kill or continue, decide quickly and move on. Now the one thing I'll say on these quick win projects is you're looking for a whole bunch of very small incremental improvements. You're not looking for the big home run. So if anyone saw that movie, Moneyball, I love that. And algae, we're not looking for home runs, we're looking for base hits. So you're just looking for 5% improvement in this, 10% reduction in that, right, 2% improvement in whatever else. Those things stacked up will very quickly add up, and they're a lot more realistic to implement. They're a lot easier to implement. They cost a lot less. They generally come with a lot less risk exposure for your firm. So look for a whole bunch of very small base hits and track the performance of your initiatives. So you want to be tracking and you want to be honest about your results. If you're not getting results from stuff, okay, kill it, move on, right? You learned it was worth the effort.
(30:01)
Just stop in there for a sec. Any questions on this? Yeah.
Audience Member 1 (30:07):
Where do you start?
Apoorv Dwivedi (30:09):
So where do you start? Great question. So hold on, let me just jump ahead to and be getting into that, into the firm tech stack. So there's a whole bunch of different potential areas that you could start here. I would go back to what's your firm's objectives? What are your OKRs? What's important for you? Are you focused around cost reduction? Are you focused around labor improvement, productivity improvement stuff? Are you focused around improved customer service delivery? What's important to your firm? Take a look at your tech stack and look at those opportunities. Like I said before, don't look for greenfield opportunities right now. That's not the time to be doing it. Take a look at your existing systems and processes and see is there a way to improve on stuff that we're doing today? I'm just going to back up a little bit on that.
(31:06)
So risks of using generative AI This is like the chat GPT stuff. You've probably heard a bunch of stuff around this already thinking about that. Model input, processing output, right? Chat, GPT and large language models, they're really designed for that output in the language piece. They're not necessarily designed for factual accuracy. There are other AI tools out there that do that, but those large language models, that's not what they're built for. So that's something to keep in mind. There's a whole bunch of discussion right now around copyright issues. There's questions around data ownership. There's questions around bias in the outputs of those systems. So just something to keep in mind. Now, there are a number of tools already available today on the market where you can input your own data. So you can put it on top of your enterprise data stack, you can control what's in there. It doesn't get shared out. It's highly secure. But those things honestly right now are still at the enterprise scale. They're going to be coming down in price and accessibility very quickly. They're not there today, but they will be coming down in terms of the use cases, like I said, quick win projects. And really I would say kind of back to your point, look at what is tasks or activities that are very data driven and repetitive. The ones that are data driven and repetitive are probably the best fit use cases for AI Those are the ones that you can say, okay, we can intelligently automate this or we can support this with automation, and we can probably measure the improvements that we can make. So again, data driven and repetitive, those are probably your best quick win opportunities. I would add onto that, coming back to my point, does it align with your firm's direction? So don't just be implementing stuff just for the sake of trialing it. Make sure that it's alignment with the direction that your business wants to go in.
(33:25)
If you have ever done this before, writing a problem statement actually really helps. So a problem statement is basically just you brainstorm your list of potential ideas. Then can just write out like a quick sentence or two that says, here's the problem. We spend X amount of time on this task, or we do this task so many times for a month and it takes certain number of hours. We think by implementing AI we can reduce this by 5%. That is a really quick way now, I mean it's probably wrong, but it's a line in the sand. It gives you something to start from and it lets you kind of measure expectations. The more you do that over and over, you're going to get better at it. So write out those problem statements that'll help you to compare your different potential use cases that you have.
(34:20)
Again, go through your goals, review the brainstorm idea of use cases. Pick a very small handful that you can implement. There's the back to the tech stack there. So there's a whole bunch of different stuff and it's going to vary for every firm. So there's a whole bunch of stuff that you can do on the admin side internally within your firm. On the production side, there's a whole bunch of things. Marketing, customer service, there's a whole bunch of stuff as well. So for every firm, it's going to be different. I can't sit here and tell you this is probably your best opportunity. Don't know what that's going to be. Now that being said, there are some things that we've seen firms get some real success with. Now, like I said, disclaimer, some firms have had success with this. It's going to be different for every single firm. Analytics that's low hanging fruit. So there are tools out there today. So intelligent automation, I use the spreadsheet example for a reason. You can get AI tools that plug into Excel or Google Sheets, and if you think about all the time that you spend in spreadsheets already today, how much time do you spend putting data into spreadsheets? How much time do you spend setting up spreadsheets, reviewing formulas, all the rest of that. So you can use AI tools to speed up that process. Now with all these ones, I'd say they supplement what a human is doing. They don't replace what a human is doing, and that's very important. So these are all tools to help you speed up your process. Reduce time, reduce costs. It's not replacing a human entirely.
(36:11)
Research tax research was talking about this at lunchtime with a few folks. There are tools out there. Blue Jay is an example. All the top 20 firms are using it. A whole bunch of big tax law firms are using it. Again, very expensive at the enterprise scale level right now, but they will be coming down in terms of accessibility. And all they do is basically sit on top of all the of the tax case law that's out there and allow tax folks to do tax research and speed up that process. Marketing. So marketing content development. Now, the AI tools, the chat GPTs and those types of things of the world, they're not going to replace a real human writer, but they can speed up that process. You can use those tools to develop outlines, develop frameworks. Everything has to be reviewed by a person, but you can use the tools to reduce this time spent on development probably by 60, 70% website, SEO stuff.
(37:21)
Seen a lot of success with this in the last little while. There's a whole bunch of very repetitive data intensive tasks that are associated with SEO or search engine optimization that can be automated. Again, you need a human to review them. Recruiting, applicant tracking systems, whole bunch of them already use AI within them. So you've probably seen these with people upload a resume, they scan through the resume and they give it a score of this person is going to be X percentage fit to what you're looking for. There a whole bunch of that stuff out there already and they work pretty well. For those of you that have used them. They can really kind speed up that process of having to review 200 different resumes, admin stuff. So calendar scheduling, time tracking, expense management stuff, pain in the ass stuff, data intensive, repetitive. You can use the tools to help speed that up. On the marketing side, a whole bunch of use cases there as well. So I mentioned some of that stuff already. Design content development, rich media stuff, whole bunch of stuff that you can do for example with video and audio and other things as well. Again, don't just use AI for the sake of using AI Make sure it aligns to your firm's business objectives and where you want to go. And then take a look at and say, okay, how can we use these tools to help improve that process?
(38:59)
So some questions you want to ask your vendors, right? There's a series of questions there. I won't go through all of them, but basically you know, can start thinking about the vendors that you're using today, the folks that are in your tech stack and how are they using AI? So we've already seen Canopy Carbon. They're already starting to integrate AI and intelligent automation into their tools. You're going to start to see it get embedded in a lot more tools. Like I said before, AI is basically now a plugin. So every single software tool out there that has that model of input processing output is going to be able to plug in AI into their tools and they all are working on it already. But you want to ask the questions. You want to understand what type of AI models are they using? What data are they looking at? What biases do they have within their systems? Understand that if you're going to implement that, what is it going to take? So what's the total cost of ownership of implementing that technology? So not just the software, the tools themselves. What type of skills and expertise do your folks need to have to be able to leverage those tools? There was a memo that was leaked from Google internally that's been making the rounds and they're talking about the threat from not open AI,but open source AI So open source stuff, it's stuff that's in the public realm and everyone gets together online and they work on it and they contribute to it and they improve it and stuff. The challenge with a lot of open source technologies, and there's been a whole bunch of them around for a while, think of Linux and other things like that, is you generally need very technologically sophisticated people to be able to implement them and administer them. So when you're looking at the rest of your applications, take a look at what do we need, what type of skills do we need? What type of people do we need to be able to implement these? And some other questions there as well.
(41:17)
The use cases, again, so this is just good business practices. Go through assess the business case, make sure you've got a clear sponsor and owner, someone that's on the hook for it. Take a look at the benefits that you're aiming for. Take a look at the vendors that are out there, assess the best fit, manage your risk. Don't commit to the long term. Measure your progress, look for those base hits and be open and transparent about your results. Look back at your results and see how did you actually perform and be honest about it. You're going to learn you're going to have failures.
(42:02)
So the question I said, I wasn't going to get into the ethics of this, but are we all going to be replaced by AI? And the short answer to that is if you do things that can be automated, you probably will be to be honest. But AI is a tool and you can either choose to be a tool or you can choose to be a carpenter. So think about that. There's a whole bunch of different software tools that are out there. You just using the carpenter analogy. Carpenters have existed for thousands of years. There's a famous carpenter about 2000 years ago, but the tools have changed over time. And so carpenters nowadays, they still have their role in the function of a carpenter, but the tools that they use have changed over time. So think about that. If you're basically just doing the function of the saw, cutting the wood is something better going to come along and do a better job of cutting the wood probably.
(43:02)
But if you think that if you take the approach that I'm going to be the carpenter, I'm going to figure out what needs to be done, I'm going to apply the wisdom that I have. That's the key thing is at this point, there is nothing in the AI world that can replace human wisdom. And by wisdom I mean the collective knowledge and experience and intuition that you have, the AI models as smart as they are, are still restricted to the data that's fed into them. They're still restricted to the learning, built upon the models that they've been created on. And so yeah, they can number crunch huge amounts of data, but they don't have the ability to have that intuition. They don't have broader wisdom beyond the data that they've been fed. And so think about these tools from that perspective. Don't be the tool, be the carpenter.
(44:07)
Now this was a quote from a gentleman that I met at another conference, Paul Roter, really smart guy. So I've kind of bastardized his quote here. But basically, AI won't replace accountants, but accountants who use AI will replace accountants who don't. So think about that, right? Next steps from here. So next steps from here going away. Think about what you need to do for yourselves and your firms. Just wanted to introduce Hitter Patel here. So he just recently wrote a book on AI for accountants. It's available on Amazon. I'd say pull out your phones, download it right now and Kindle. Start reading it, read it on your flight back. It is a really good starting point. So it'll give you kind of a good basic exposure. Get your head into that space of AI and what it means for accounting firms. Start thinking about your implementation roadmaps. So what do you need to do to activate your teams? What do you need to do to put in place your AI policies and how you're going to approach this and get on it right away. Don't wait for three months from now. Don't wait till next year's plans. You need to get on this ASAP. what's the name of the book?
Audience Member 1 (45:35):
(Inaudible)
Apoorv Dwivedi (45:36):
Yeah. And it's available on kindle and Amazon And there's a number of other books out there. There's a whole bunch of blogs and resources too. We have a spreadsheet of AI resources. Again, send me contact info and I'd be happy to share that out with you. There's a whole bunch of stuff out there. It's a rapidly developing field, but get your head into that space and get going on it right away. If you want a copy of these slides, I think they're available through the app. And if not, yeah, sorry, question
Audience Member 2 (46:24):
How earth all day be available until after the conference. I asked that. Oh, thanks. That's my response.
Apoorv Dwivedi (46:38):
Okay, well maybe there's an AI app for that. We can improve that. But yeah, so okay, if you do want a copy of the slides, send me a note. I'd be happy to share them with you. If you want to take a copy of the slides, customize for yourself, your firm, we can do that as well. You can do that. I don't really matter. So yeah, any other questions? Yeah, go ahead.
Audience Member 3 (47:04):
From about clients of accountants also be knowing about AI?
Apoorv Dwivedi (47:10):
Yeah.
Audience Member 3 (47:11):
So how is that change clients expectations?
Apoorv Dwivedi (47:15):
Yeah, so great question. So the question was, this isn't just something that accountants are working on. Everyone knows about AI and what's going on with AI out there. How's it going to change expectations about AI? So if you're on that pricing model of selling time, most clients are going to get smart pretty quickly to the fact that, okay, if you're using tools to reduce your time spent working on their activities, that should probably be reflected in your pricing. If you're on a value-based pricing model, that question's probably not going to come up as much. So hint, hint, hint, good opportunity to move to value-based pricing. Anything else? Yeah.
Audience member 4 (47:59):
Thank you. And so a lot of what I've observed about AI in accounting and also a lot of what I'm seeing over here confirms that it's mainly being used for internal purposes or for marketing purposes. I was wondering if you have found applications or use cases for client facing activities. I know that that can be a little bit dicey when it comes to ethics and things along those lines, but have you found anything?
Apoorv Dwivedi (48:21):
Yeah, absolutely. So in the marketing customer service field, so you're right, there's a whole bunch of stuff around content development that can be sped up improved with the use of AI tools. There are analytics, for example, that sit on top of salesforce.com that will allow you to slice and dice customer data and generate all sorts of insights. Now, could an analyst do that? Yeah, they could. But when you start automating it, it's when those data intensive repetitive tasks that can be sped up or improved through the use of AI tools. So that's one example. Chatbots, not chat GPT, but chat bots on websites. You've seen a whole bunch of those out there. Personally, I don't think they're very effective. Anyone that's like ever interacted with one of those AI tools on there, they're real pain to deal with. You just want to talk to a real person. They're getting better. But there are ones out there. There were some discussion in previous sessions around customer research, particularly in the areas of things like customer satisfaction. So again, it's using the AI tools to generate additional insights from those types of activities. So it's not necessarily replacing the need for customer research or customer satisfaction, but you can go through large amounts of data and speed up that process. Anything else?
Audience member 5 (50:01):
It's okay. So for years have been going these homes as a, that technology's going to replace us. You better get into advisory because technology's going to replace you and find something to do that technology can't do. And now all of a sudden AI is going to do all the analytics that I've been working so hard to learn how to help clients do in advisory. So now AI's going to do advisory. So now I'm screwed again. So what do we do?
Apoorv Dwivedi (50:33):
Yeah, so a great question, Veronica. I would say for the time being, it's not replacing advisory. So back to my point around human wisdom, there is number crunching that can be replaced by ai, but not the understanding the here's the questions that you need to ask it, right? So I kind of go back to like that. If you think that analogy of the spreadsheet, you need to understand that you're still as a person putting together, here's all the different formulas in those cells in the spreadsheet, here's all the different workbooks I'm connecting. It's not a spreadsheet, it's AI tools. But at this point you still need humans to connect all that together to understand what are the questions to be asking of those tools, how do you leverage those tools? So I think there's actually a bigger opportunity because yes, customers will know about AI, they'll think about it, but it's a chance to really evolve from, I'm just a number cruncher, a bean counter to I understand how to interpret the data to help improve your business for the outcomes that you're looking for. Anything else? Okay, so hopefully that was energizing for you. You have some ideas and yeah, anything else, just feel free to reach out, ask any questions. Thanks.