096 AI Is About To Improve Email For Accountants

Today we are talking about how AI is about to make the process of writing emails so much easier for accountants. This is a huge topic that I think we need to keep hammering on, keep developing ideas for, to ensure that the people that build awesome tools for us are continuing to invest and just how much of a problem email is right now and how much AI can improve upon it. So doing a deep dive on a whole bunch of stuff that is already there, already in the works, and stuff that we can still get that will make our lives better way easier. Come on in, it's Jason Daly. Okay, real nuts and bolts. What are the ways that AI is going to make emailing better for us? First off, drafting stuff from scratch. Now we've seen this from a few providers. A view of the cloud practice management systems now will let you generate an email from a prompt. This is just basically the same as going to ChatGPT and say write an email about XYZ and that's great, that's a good place to start. But ways we can continue to improve upon this. A huge like component of whether that works for you or not is how true it sounds to your voice. And the reality is if you are emailing out of your practice management system, that thing has access to every email that you've ever written. It's not hard to write a prompt that defines the tone of your voice. And right now all that stuff is generic, it's bland. And to be honest, I can usually tell when something's been written by GPT right now and people are just gonna become aware of that stuff. So what I don't want is to generate an email and my client's going to be like, is that need to send me something that a bot wrote? So I think we can do more to align what these emails look like relative to our normal tone of voice. And so that is user specific prompt instructions. And this is something that practice management systems can absolutely handle on the back end. The very first ChatGPT video I did on the main channel back in February, we created a paragraph descriptor of what our tone of voice is. That's something that tools can do with your email history. No problem. Make that email sound true to me. Go a step further here though. Make it true to me, but also make it true to the tone of voice that I usually use with that client. There are clients that I'm super professed with, there's clients that I'm always cracking jokes with, there are clients that I'm really obscene with. And if I've Got a bunch of email history with that specific client, that kind of custom prompt, that descriptor of how I interact with that client. That's really valuable. And honestly, that can just be another property of what the system retains for that client. So every, every time I go and write an email to Jim, it can call that prompt that it's generated based on past conversations I've had with Jim. And now when I generate an email, there's a much greater likelihood that it's actually going to sound like something I would really say. That's kind of the biggest blocker I see with this right now is they all just feel very sanitized, but absolutely a thing that can just be built into our practice management systems and manage this for us. Now, another interesting spin on being able to generate emails from prompts is finding ways to make this useful when you're away from the keyboard. So, so pull it into the mobile app, voice support, stuff like that, where I can generate actually a pretty solid email when maybe I'm not like sitting in front of the keyboard to pound that whole thing out. Now, ultimately, most of our emails are replies to threads, so let's talk about replying and how we can use AI to squeeze even more out of that. Now, table stakes. Is the thread, like the history of that email thread being taken into account when the reply is written? In order for a reply to really be helpful, it needs to be aware of all of that stuff. Otherwise it's going to generate something that just won't make sense in the context of the conversation. Now, there's absolutely secure ways to do that. Right now, Carbon's doing it. I don't know if anybody else is doing generative replies yet, but we got to start there. When it's suggesting to us a reply that we can use, it needs to take into account the history of that email thread. Now, where this gets leveled up big time is when it can see into all the other organizational data within the system. That is past emails that I've interacted with the client over, that is emails that my colleagues have had with them, meeting transcripts that, the status of projects, the files that we have in the system about them. If we're simply generating emails based upon what's happened just in that little email thread, think of how limiting that is compared to all the other data and information that we have about past interactions with that client. There is way, way more that we can do here that will make these generative replies really valuable. And to be honest, I don't know that it's getting enough attention right now. When Nyle Carter Gray put out on Twitter a couple months ago, if you could build your dream AI bot to do one thing for you, the most common response was manage email better. So the whole like, scavenger hunt that goes into many of the emails that we have to reply to, like, we have not optimized this enough yet. And this is an area where I can super help us. Another interesting thing here, I was reading an article about how it was a more traditional tech company and they have a support team that manages support tickets all day long and they're wanting to train their new folks, maybe they're underperforming folks on the responses of the highest performing people on their team. So the folks who are kind of setting the bar and the model for how they want the rest of the team to kind of communicate through email and for this they like fine tune their own model and all that. But this is where actually generative email can become a competitive advantage. If this is done in a way that is true to the specific tone that you want your firm to have, and it requires some sort of source of truth, and maybe you're going to be that source of truth and you want your team to communicate with the same sort of tone and enthusiasm or whatever it is that you do. The more that we're going to lean into generative email, which I'm telling you, over the, like, next couple years, we're going to transition from 100% of the words going into an email being written by you to a shrinking percentage written by you and a growing percentage written by AI. That's an opportunity to have a really cool influence on what your communications look like, on how to improve upon them, and have a more consistent style of communication across a firm of various different people. Now, maybe this is a good thing, maybe this is actually squashing the uniqueness of the individuals within your firm. I'm not sure what the right balance is here, but exactly how those emails are generated. When you think about the number of emails your entire team sends, it actually becomes really important. And for me to be able to tailor and have more control over those generative suggestions across the course of a whole year, across, like your entire team, it's actually going to have a big impact in aggregate. Now, interestingly here, the better this stuff gets, you start thinking down the road of like, okay, so what if my bot's just talking to my client's bot and like, what is the point of all this stuff once AI is just writing these emails and they're bouncing back and forth. What does that look like? Well, generative email replies right now are like, okay, like probably not just going to send stuff off that it generates without modification. But it will keep getting better and it'll get to the point where oftentimes it is plenty fine. So what is a user interface that's optimized for that? It's probably not this big sort of kind of authoring thing like we have now. It's probably something closer to an approval workflow where maybe it gives you a few options of like the high level concept, how it thinks you're going to reply to this option A, yes, no, whatever. You select which direction you want to go and then it generates a couple options for you and you're like a B. I like B better. Send. Be off. This episode is sponsored in part by Copilot. Modern accounting firms, let me tell you, they run on Copilot. Differentiate your firm with a secure portal that gives your clients a one stop shop to upload files, sign contracts, send messages, access dashboards, all that and more. You know what your clients don't enjoy is hopping all around to all your different tools, logging into this magic, magic, you know what I'm saying? Magic email linking into this other doodad. And every little thing they pop into, it's got a different vibe. They're not sure, did I just get this from my accountant or somebody who doesn't mean me? Well, let me tell you gang, that is where Copilot comes in. Clutch. Because you can put all that stuff in one place. They sign into a single place. That's the only place you're ever gonna ask them to go. And it's a consistent experience. They know they're getting that stuff from you. They're not getting bamboozled by somebody that's posing as you. That's why Copilot was built from the ground up to be not just a portal system, but a platform to build whatever you want on top of it. So you've got a single streamlined client interface that does everything, everything you need to do. Learn more about Copilot, check out the link in the show notes. This episode is sponsored in part by the fine folks at Cloud Accountant Staffing. Do you hire accountants? Bless your little heart. Not the best part of the job, in my opinion. Not something I ever enjoyed. Well, listen, you can build your accounting dream team with talented offshore accountants in the Philippines that work 100% full time for your firm. Their accountants aren't freelancing or contracting for multiple firms. They're all yours. They work exclusively for you and are incentivized to stay with you and your team long term. They're not gonna get swiped Cloud account staffing is 100% dedicated to the accounting industry and founded by a former accounting firm owner that understands your business, knows your pain points. They had to hire some accountants and they said, you know what, we're gonna build our own pipeline in the Philippines, going to pull in some super talented people and then open that up to other firms. Basically that's the story I've been talking about a lot about staffing, building more resilient staffing pipelines for your firms. I had staff in the Philippines like totally red pilled me to like oh geez, like we need to globalize the way that we get our work done. Check these folks out link in the show description Cloud accountantstaffing.com so emailing is probably going to look more like an approval workflow in the near future rather than staring at an empty page. And this means probably giving you buttons to do this stuff rather than selecting text. So even if it suggests an email, like give me a button to shorten it to make it more professional to lengthen it, rather than having to hop in and manually type this stuff out, give me quick and easy ways to modify maybe a specific sentence of what it generated, that sort of thing. Another interesting thought here. Remember Outlook Desktop? I don't even know if it still does this, but remember when like spellcheck was this super cool novel thing and you would go to send an email out and it would go skirt? Hang on, did you misspell this word? Are you sure that you want to send the CM email out and then it like gives you a pop up and you got to go through all the words that you misspelled the entire email and like 12 of them are in your email footer. So you got to go through that every single time until you figure out how to turn it off. Well, how about AI doing this for us when a response is maybe falls outside the boundaries of some sort of parameters we've set. Maybe around salty language, general tone, maybe even more practical stuff like oh, you actually forgot to respond to this one part of their email. So almost like a spell check when it goes out the door, but around maybe more firm level parameters that you have set up. But also just from a helpful standpoint. Hey, have you considered adding this to your email, sir? That might be helpful. It might just be annoying. I tell you, I want my clients to have that when I send them a list of one through seven and then they send me two, six and I would love for them to have that assistant. And then the holy grail here with replies is enriching those generative replies with I'm going to use some words here. Agent fetched external system data. So we've talked about agents a good a bit on this podcast. Agents are basically AI driven users who can actually go out and navigate programs. An example is some of the stuff we've seen from Pixie where they will have an agent go out to a zero file. You know, you get an email from a banker asking for prior year sales and P and L stuff like that. You can actually go out to Xero and fetch the stuff that they request and build it into a suggested email reply. The beauty of agents is because they navigate stuff like a human user, they can hop around and do anything that we can do across systems. So when you get an email and you have to reply, okay, what if that agent then just in the background is actually going out and researching and trying to answer that question for you. So that the generative reply, maybe it just pulled something from a Department of Revenue website or something like that that is like helpful, authoritative stuff and it says, I'm pretty sure this is the answer. You could link to this for a source of authority. But when we have that agent that can go out and do meaningful research, pull stuff from other places and those generative replies are going to get way better. Right? And the goal here isn't to like completely replace your humanity and you injecting into it your own thoughts and opinions and stuff like that. It's just to give you a way better starting point than you have right now, which is just a white screen and a blinking cursor. Now this whole kind of opportunity around agents is also why I still have my money on the accounting space. Specific tools beating out stuff like Microsoft copilot long term. So February Microsoft comes out, look at me copilot everything. Ta da. Look at how cool this is going to be. You can chat with your spreadsheets. I can, you know, talk with this assistant that can see across all of my Microsoft stuff. Super cool, super exciting. Still can't actually get it, but it all sounds super exciting. And on the day that that drops, that might be the best like all in one AI solution that there is. But the long term possibilities and the most compelling that AI will bring us in the next 36 months will be that which AI agents enable, agents that can actually go out and do the work. This is like the Most promising aspect of AI where it is today, folks from OpenAI have even said this. It is the difference between having language models that will like do cool novel things for us, like, you know, summarize our meeting transcripts and being able to chat with transcripts and stuff like that. Those things will look like a party trick relative to agents that can go out and reconcile an accounting file or who I can assign a task to in my practice management system and then that agent goes out and does that recurring task. It is like a digital worker actually doing stuff for you, some of which you will instruct in a very controlled way, many of which will run in the background and you'll never even realize that they're doing this stuff for you. Or maybe they tried to do something for you and they couldn't. These agents are developing super fast. I recently shared on my newsletter what is by far the most compelling browser based agent. And I shared how I was looping through all of my QuickBooks files and having it generate, you know, cache basis your P and LS for each file. And it was already good enough to do that. And it's far from perfect. But this is why my money is still on the software companies in the accounting space. And I've not been in this camp historically. Historically I've gone to the mainstream tools. I put my CAST practice on ClickUp and I can see the argument that the mainstream tools will have the money to invest in really cool novel AI applications. But what they're not going to do is build the domain specific stuff that will be really valuable for us. The way those agents work is going to be very specific to the task that accountants do. And I don't ever see, you know, Microsoft or something like that solving specifically for that. Okay, attachments, let's talk about email attachments. You know what AI ought to be able to knock out for us right now? Hey, can you send me a copy of my prior year tax return? Hey, can you send me, can you send me a copy of the February financial statements? They're sitting in my file system. Listen, robot email reply buddy, you see it right there sitting on the shelf? Go grab it, attach it to an email and say, is this what you want to send to the client? Or generate a reply that says hey Steve, check it out attached. Let us know if you need any, anything else. Just imagine how much of a time saver that could be if you had that for the next 12 months. The number of requests like that that you get or if you are that client rules, portal elitist, the reply that says, steve, go log into your portal. I've told you this a hundred times, I'm not going to go fetch that stuff for you. I look like a golden retriever. Just go, you know how to log into your portal? Go get it. That could also be like the suggested reply, right? But going out to your file system and being able to find related files, attach it to those suggested replies, that's going to be really handy. But it can even go as far as creating new files ad hoc when you need it. And like, a bank letter is a really easy example here. If somebody requests a letter for a banker and you've got 15 examples of very similar letters in your file system, great. Draft up a letter for me that addresses what exactly they need using those other 15 documents for reference. That's absolutely something that AI can do right now. Here's a first draft of that letter and oh my gosh, you just did 90% of the work for me. Thank you, robot. I mean, that's what we do anyways, right? Like I go out and I search the file system. I'm like, ugh, when was the last time I put one of these things together and I save as and I change the client name and all this stuff and like, that's my starting point. AI can totally do that stuff for you. Hey. This episode sponsored in part by Firm360. I know you're just about to ask Firm360 practice management system that gives you a 360 degree view of your firm. All that stuff all in one tool. And if you listen to me talk software much, you know that I'm an advocate of asking my peers, hey, what are you using? What is working and not working? And you know what? There is a whole bunch of reviews online for Firm360. And I'm not talking about the stuff on the Firm360 website. When you know, you know, they just cherry pick the winners. I'm skulking Capterra right now. Come on in. Whole lot of happy people hanging out there leaving reviews for firm 360. So I know I can trust them. Let's talk about Tempe B Certified public accountant. Got a 1 to 10 employee firm quote. Great solution for small firms. We've been using Firm360 for several months now and with each monthly feature update we are loving it even more. We're a very small firm that needed an affordable yet effective CRM solution. And firm360 has filled that need. I've implemented other software programs. This one has been the easiest by far to get up and running. What else do you need to hear? Okay, be more like Tempe. Learn more about firm360 the link in the Show Notes hey, today's episode is sponsored in part by Financial Sense and I am very excited to share. Financial Sense is putting on a virtual conference called Workflow Con and yours truly will be there talking about how to use AI for workflow stuff. Conferences happen in late October. Got actually a pretty like rockstar cast of speakers. Don Brolin, Ryan Lazanis, Kelly Parks, Now Carter Gray, Veronica Wasek, Allison Ball, Terrell Turner, Nicole Davis. Whole bunch of super smart people use the link in the show notes to sign up for free. Yep, I said it for free. They've got all the sessions organized into tracks so you can pick something that makes the most sense for you. You know how I feel about conferences. It is the best way to get out of your kind of group. Think tunnel vision that we all have within our firms. Come check this one out. Hit the link in the show notes for more information. Last we're talking agents again, generating files with agent fetched info. That stuff just gets even better once agents can go out, grab information from third party services, do like a web search. I did a main channel video on ChatGPT plugins. There's a Chat GPT plugin called Web Pilot that I love. It's kind of a replacement for the now defunct built into ChatGPT web browsing the Bing search stuff that they've killed for the time being. Web Pilot is great at this. You can plug any number of URLs in and it will go out and fetch stuff from them and that will inform the response that it gives you. Imagine just having an agent like that that can go out, find the most relevant information online, summarize that, give you a couple links to authoritative stuff in that suggested reply. That's a great head start rather than just building that document from scratch, right? Okay, last analytical stuff. Client Hubs talked about working on sentiment analysis and client communications. Imagine if you got your big old client list. I should clarify sentiment analysis. This is given a bit of writing. How cranky is that person or how happy is that person? We've had AI models to do this stuff for a while, but this is something where if you think about your whole client list and how you view that, it would actually be kind of cool to see like trends in sentiment analysis. Is there anybody that's Super PO'd right now and I might want to know about that stuff? Are they trending upward. Are they trending downward right now? You may not have enough information to make, like, a clear assessment of that. Maybe they only send you one, you know, one email a month or something like that. But when we get to the point where we've got, like, meeting transcripts pulled into our practice management systems, which is obviously a much bigger volume of information, that sentiment analysis could be a really cool, like, valuable thing to track. In addition to trends in that, identifying outliers may be really helpful there too. So if you've got a client that's normally a 50 on the sentiment analysis and there's an email exchange with a staff member that's a 3, like they're totally PO'd or something like that, A big jump like that, that's probably something that you want to know about around analytics. There's probably a bunch to do with. With email. Sometimes I like, worry if too much of that is getting, like, weird big brothery over the staff. Obviously, if I'm running a firm, it's ultimately up to me to ensure that this staff is communicating with clients in a way that is representative of my firm. And we see this in, like, more traditional support environments where people are assessed on, you know, ticket response time and stuff like that. But I feel like there's more sort of probably analytical stuff that AI could pull in specifically around how the team manages email. Since we do it in such a siloed way, the more we can leverage tools to kind of cut across that and spot trends, outliers, who's doing the best? Who can we help? That's probably an area that's ripe for improvement. So what is here in all this? Already drafting emails from prompts. The big thing that we can improve on right now, though, is drafting an email that is more true to my normal tone of voice in general and also my tone of voice with specific clients. Devil's in the details on implementation there of how good it will be. Table stakes now is just generative email. Next is generative email. That is super true to how I email people. Generative replies is here. I know Carmen's doing that. I'm not sure if anybody else is doing that yet. That is the entire three thread of the message is taken into account when generating that email back. What's close? Like, what do we think we'll have in the next 12 months? In the next 12 months, I think we will have companies doing meaningful work on actually incorporating way more than just the email into those suggested replies. It's a bigger technical build. It's Definitely more complex. In most cases it's going to involve what's called vector embeddings or map reduces or summarizations of the information that is in your file system so that it can easily go out and see. Okay, given this email from the client, what are the most relev things and everything that we have to this email and how do I pull it in in a constructive way? But people are already working on this, so meaningful improvements to what replies look like. We'll absolutely have that within the next year, I think if we're thinking about what's out in the future. AI agents in many ways are the final frontier. But man, if you are plugged into AI news, there is so much noise around agents right now and how quickly they're developing. I would call agents probably the biggest long term thing out like one to five years where we get really, really meaningful implementations of agents that can go out and actually do really meaningful tasks. These could be agents in your web browser. They can also be agents on your desktop, clicking around, typing stuff into your tax software, all that stuff. Now obviously the devil's in the details. As soon as that thing can actually put information in and click stuff, you're like, whoa, turbo. Let's, let's make sure we're not making any mistakes here. But this is also why I think accounting firms are actually super well suited to lean into AI because we already have these approval frameworks. We go out and we hire, hire interns that don't know squat and we say, okay, go do this really complex thing and then they submit that project to you and then you tear them absolutely to shreds with all the things that they did wrong. That's just the circle of life in our profession. And so AI seems perfectly suited for that because we've already got the kind of the technical review framework. Do you have any other thoughts on email? What are any other like, particularly painful aspects of email that maybe AI could help us with? I want to keep pushing this conversation because I do think it leads to good ideas, good discussion among the product companies that we on how to keep improving upon this. Because is there anything bigger than email right now? Maybe, but not many. It is really one of the most brutal, grindy aspects of running a firm. So let's keep the conversation going. From the time we started this podcast, we didn't have any generative email. The world's much bigger than this podcast. But we are marching in the right direction. Let's keep fighting the good fight. I'll see you tomorrow.