Burn Tokens, Not Headcount: How Small Teams Should Hear That AI Phrase

**[00:00:00]** There's a phrase from a recent AI talk that I think small business owners should understand. They said, "Burn tokens, not headcount." Now, if you hear that phrase the wrong way, it can sound cold. It can sound like the whole point of AI is to get rid of people, and that is not the useful interpretation, in my mind, for a small business. The useful interpretation is much more practical. Before you add another person to handle repetitive work, ask whether the work itself should be redesigned. It's not people do not matter. On the contrary, it's people matter too much to waste their time acting like glue between systems. That is something that AI can do very well. So, it's completely different concept. In the talk, the speaker claimed that some startup companies were reaching demo day with roughly five times more revenue per employee than companies did 18 months earlier. Now, this is not the typical scenario. We've talked about this before. There's a lot of failed cases, as well. But, I would not build a whole operating plan around one number from a talk. And I do think that direction matters. Small teams are getting more leverage, and AI is changing what a small team can produce. Work that used to require more coordinators, more assistants, more analysts, more writers, more manual checking, and more back office effort can now be redesigned. Some of that work can be automated, and some of it can be drafted by AI and reviewed by a human. Some of it can be monitored by systems. Some of it can be in just eliminated completely, because the process was bad in the first place. And that's where the opportunity is. But, I want to be very clear. It's not let me take this job away from this person, but rather let's see what this person can really do to benefit the company and themselves **[00:02:00]** that the AI can't. There's so many things that AI and computers still cannot do that we need humans for, and that's where we want to have humans. This is not a video about layoffs. It is completely the opposite. If your first thought when you hear AI is efficiency is who can we replace, you're probably starting in the wrong place. The better question is, where are we wasting capable people on low-leverage work? That's much more humane, and it's also more profitable in the end, because a lot of small businesses do not have a people problem. They just have problems with their workflows, and they try to keep solving them with people. And the process gets messy, so then they hire a coordinator or a manager, and then the handoff breaks, so they hire someone above, and then the reporting takes too long, so they assign someone to gather the reports. And then the sales follow-up is inconsistent, so they ask the team to try harder, but then the onboarding is confusing, so they add another person to chase the information, and now you've got a mess of people that is just chasing the information around, when instead of having it all scattered, if everyone just gathered the right workflows and understood how these processes work, you might figure out that you need less people to do that work, but not in the sense of let me fire people, but instead, let me allocate the human resources, which are the most valuable ones, into work that is actually fit for them, where we can actually use what we as humans have as benefits and way more advanced intelligence than what the AI can do. AI is not super intelligent. It's very good at at pattern matching and pattern recognition and solving things and copying things and moving things around, but there are many even logical, simple, common sense things that AI does not have, and it doesn't have experience either. So, you need the humans to give **[00:04:00]** it that experience layer and give it that context that it doesn't have. If you do this properly, then you're not having your team interrupting the person that has the most experience in the company just to get the right answer. That's where you can start leveraging AI to help. Sometimes hiring is definitely going to be the right answer. A growing business, yeah, they need people. And customers need humans. And the teams need leadership, and complex work needs judgment, but sometimes hiring is just an expensive patch on a broken workflow. Same as what we've been talking about over the past few days about AI. You don't want to just put AI in for the sake of it. You first have to fix the workflow. Fix the workflow. And it's the same for humans. So, if you're going to hire people, before hiring them, first fix your workflows. Make sure that everything's there, because otherwise, when you bring someone in, you won't be able to onboard them properly, and then it won't really help. And if you already have all your processes perfect, please, by all means, we need hiring more people. But make sure the processes are aligned. Same goes with AI, and that's where the burn tokens, not headcount, becomes useful. It means use computation before you use another human hour. Use AI to draft before a person writes from scratch. Use automation to route before a person manually forwards. And again, this might be taken the wrong way, and I'm sorry if you take the wrong way, but it's completely the opposite. I just got an email from a client where we've implemented some skills months ago into their flow and he just sent an email saying how it's helped their drafting process become so much streamlined, **[00:06:00]** much better than dedicating 10 hours of work into it. And it doesn't mean they're not writing the stuff themselves, it just means that the AI is helping them or assisting them in getting everything together so that they can write better or so that they can proofread it and then get it out the door. Remember, the AI slop that you see online is AI slop mostly because people are using the least amount of effort possible by just telling the AI, "Hey, write this for me." and then just sending it away without checking. And that's not the point. I mean, everyone could do that, but it's not good. It's just going to be flat. But, if you give the AI context and your brand voice and the way it should handle conversation and information and access to your knowledge base and experience from your most experienced employees, then it can start to draft things that sound like something someone in your company would draft. And you've never seen this in action, it's amazing. I've seen people whom I know how they speak and I've interacted with them and then I've interacted with their AI and it sounds just like them. It's virtually indistinguishable from how they write their text messages and their emails because they spent the time training it and making it better such that it did that. And so, if you can use these tools to streamline your processes so that the people that you actually have make the most out of the time they spend instead of just being mindless drones and copying and pasting information, you're opening the opportunity for them to grow as humans and you're opening the opportunity for them to learn more and to enjoy their job more than just again copying and pasting things all day. They can now focus on real tasks that use their creativity and use other parts of the brain that the AI **[00:08:00]** is not able to use yet, especially when talking about multimodal. That's a whole other ballgame. So, make sure you use the knowledge base before a person a person asks the same question for the 50th time, you use workflow triggers before a person has to remember, use a summary before a manager rereads a 30 message thread, use quality check before a human has to inspect every single detail. I mean, that's a very big one. We see it We're talking about this in the tracking event I talked in in a few last month. It was there's so much data. You might have a bunch of work orders and if you have to go through each of them, it'll take you a long time. But if you just use the AI or whatever, you don't have to call it AI. It just can be a parser. Could be a a logic driven tool that can go through all of them, match items, match numbers, match SKUs, figure out when something is incorrectly numbered, figure out if something has aging more than what it should have, what inventory you should have gotten rid of. It can flag all those things, so then a human reviewer can just save so much time by just looking at a dashboard with everything color-coded so that they could just make an actual decision. So, the human has to make the decision at the end. Don't forget that. That's really important, especially high-level decisions. You still need a human to sign off of them because the human is the one that's liable for them as well. So, it's not all like let's just replace AI with everything. There are instances where you actually have someone that needs to sign off on it. And preparing everything for them to make a better decision is very valuable. Use tokens for the repetitive, reviewable, structured parts of the work. Save human attention for the parts that require trust, taste, empathy, negotiation, judgment, creativity, and accountability. **[00:10:00]** That's the practical way. And I'll try to give you a good example for example, an operations manager at a service business. This person is smart. They already know the clients, and they understand the process. They can solve problems. They've been doing it forever. But if you look at their week or their schedule, they're checking intake forms. They're chasing missing documents. They're updating the CRM. They're creating project tasks. Then they're sending reminders for those tasks. They're answering internal questions from the rest of the team. They're rewriting status updates. They're formatting reports. They're copying the information from an email into a project management tool, and then from there they're exporting and they're passing it on to someone else. And they ask the same three people for the same missing detail every single week. By Friday, they're exhausted. Not because the work was actually difficult. It was because they spent their time just moving around, holding everything together with manual coordination. That person is very valuable, but a large part of their week is just glue work. And that's where AI and automation can actually help. An intake form can trigger a summary automatically. A summary can highlight missing information, like we're talking about. The system can then draft a client kickoff checklist. It can create tasks based on the service type. It can send a polite reminder when some of the documents are missing. It can update a status field automatically based on certain triggers or actions or dates, it can produce a weekly status summary for review, it can flag exceptions for the operations manager instead of making them inspect every case manually. Now, the person becomes way more important because now they are less trapped and they can handle the exceptions. They can improve the onboarding process. I don't know if you've realized this if if if you at **[00:12:00]** your company or someone you know at their company has started using or implementing or deploying AI into the business, what I've seen personally is that you get time back that you can use for things that you would have loved to do before, but you have never gotten the chance to. Onboarding is a great one. How many of you have created a tool or a software or even your own service and you've always wanted to create a wizard for the onboarding or you've always wanted to create a better experience or you've always wanted to do something special for your clients when they come in the door, but you can never get to it cuz you're always busy. Now, these tools start to open up some time and also make it quicker for you to do things with tools like Cloud Code or Codex or Gemini, you can actually tell, "Hey, I need this little wizard that I need people to complete while they're going through that I will explain and take them by the hand." Well, now you can just do it and it's something that previously you would have to hire someone to do and it would take a long time so you would never get to it. We're seeing this transformation with tools we're building with clients, with tools that we're building with clients that it's like oh, all of a sudden I have the opportunity to actually make those fields nicer, easier to read, add accessibility, remove friction because it's so easy to manipulate the software as it stands today. These people can now go and talk to customers, get their feedback. They can go and use their time to train the team. They can use their brain to start noticing patterns. They can do the work that actually needs a human brain. And I That's the goal, not fewer humans doing the same broken work. It's better workflows that let the humans do better work. I hope that makes sense. It's I mean, it's straightforward to me, but there's so much of the narrative making AI look bad, but it's actually good if you're **[00:14:00]** using for the right reasons and in the right way. At least, that's my personal and professional opinion. Think about a sales team. If a company gets leads from the website, referrals, events, and email introductions, every lead has to be reviewed. Someone has to go and check the fit. Someone needs to research the company. Someone needs to write a follow-up. Someone needs to update the CRM. Someone needs to remember to follow up again. And when volume increases, the instinct is to hire just another sales coordinator or another rep. And again, in some cases, that might be the right move. But first, ask what the workflow is doing to the current team. Are they spending the time selling or are they spending the time copying data? Are they having better conversations or are they spending time searching for context? Are they building relationships or are they writing the same first email 20 times? That's where AI can help. A lead comes in, and then immediately, automatically, the AI can summarize that inquiry. It can check the company website. It can compare the lead to your qualification rules. It can draft a first response using approved examples. That's where the examples come in. It can create or update records in your CRM. It can suggest the next step and it can remind the right person if there's no reply. A human still reviews the important messages and the human still handles the real selling. And the human still decides when the deal needs actual judgment, but the repetitive parts are no longer consuming the same amount of time. And that is leverage. I And I don't mean because this gets me sometimes. I Like if you use Google Suite, suddenly you're getting summaries by Gemini about the email. **[00:16:00]** Well, I personally don't don't appreciate that too much because it'll give me a summary of what it understood that was relevant of that email. And sometimes that will that will be okay. Uh if if you receive a lot of messages want to get a brief overview, but many times there's there's context hidden in what the person is saying and just basing yourself off of that then means nobody's reading emails cuz cuz you're just using Gemini off the bot of the box like vanilla version, that is not as helpful in my opinion. I believe you should have your own version of the thing that is based on the context of how your company works because that way it will pull out the things that you have trained it to pull out from the emails and it'll understand what's relevant for you, the person within this particular company. And so it's not just a summarizer and I miss all the context and all the nuances in how the person wrote the thing to mean one thing or another. It can actually help you extract what you are typically looking for in a message. That's really where it's valuable. And that's what I think we should be striving towards, hyper-personalizing these tools that we use so that they give us real value. Now, in customer support is going to be very similar. I mean, you've got documentation, but it's typically very outdated, especially if you're releasing features often. Uh the onboarding process is unclear, so the support people get actually confused customers. The billing rules might not be easy to understand by anyone, so support has to then explain them. And then the internal knowledge base is probably a mess as well. And then when cuz somebody comes along and they say, "Hey, support is overloaded. We can't handle it. We need to hire more people." Then you hire more people, but but again, it's the same loop. Uh if you first go and look at the repeat questions, you might figure out that they're all answering the same thing. So, maybe a voice agent might help there. But first, you have to fix the knowledge base and **[00:18:00]** the workflows and everything. And that might be done in in a few days or even a couple of weeks to have everything ready. And then that will allow you to then scale and grow and actually get value from the people that you're bringing in. If you're looking at how everything that's being repeated and where the tickets are being escalated, you might find a better investment of your time to enhance the knowledge base or enhance the rules so that it helps everyone and not just that particular team member to whom you've now removed some of the load. It will help everyone involved. In fact, it can improve customer care because humans have more time for the customers that actually need human attention. And then there's people like myself. If if I'm calling an office to schedule an appointment, I don't mind if I'm talking to a person or an AI as long as the AI understands me as much as a person. So, if I call a company and the person is not a native speaker of the language that I'm trying to speak in and they don't understand me completely and I have trouble getting my message across, that's worse than an AI that just understands me perfectly and can book my appointment ask me answer my questions, right? So, it's a balance. Are you sending people offshore just because it's cheaper? Then in that case, yeah, I mean, maybe the AI agent will do a better job because it will be in the language and the form and the voice that you want. And it can take a bunch of calls at the same time. So, in some scenarios, yeah, use that. In other scenarios, then it might make sense to use people from abroad. Depending on what your business needs and what your clients are looking for. Sometimes you actually need people from other countries because you need that nuanced language for a particular subset of your customers. And that's great as well. So, there's ups and downs with everything, but as long as you're providing the **[00:20:00]** experience that the customer wants and expects, then it doesn't really matter who is giving them that attention for certain things. Like I said, there are parts where I prefer an automated tool that can get me out the door quicker instead of having to wait for half an hour for someone to answer the message or to have trouble trying to explain it to human that doesn't understand me anyway, right? That's a good scenario. However, if I'm having an issue and it's more than just a troubleshooting thing of have you checked ABC D and E, but rather I have to talk to someone about billing, there was an issue or whatever or then in that case, I want to talk to a person that has empathy. And that's where you have to sort of triage your calls and everything and have those people available because if you only use people, then you will never have enough people to manage everyone's concerns. But if you automate part of it, that frees up the people to actually talk in the scenarios where empathy and other human characteristics are needed. The best AI efficiency work often improves the employee experience and the customer satisfaction at the same time because the employees stop doing the repetitive coordination and customers get faster responses. Then obviously managers and VPs and CEOs get better visibility and owners get more capacity without necessarily increasing payroll. Now, keep in mind this is not across the board. There are AI agents that cost six figures a year. It's not necessarily a cheaper thing. It might be more expensive in some cases, but that's because they're probably doing a volume that's immense and so that versus paying two people or one person a year, but then you're handling the workload of a hundred, then okay, it makes sense for that company, maybe. There are others that are going to be very cheap and you might be doing it because it's cheap. **[00:22:00]** You might be doing it because it's good. The most important thing is that it solves your problem and it doesn't add extra barriers because also you're talking about AI, you add AI into a process and then it just adds more points that you have to then verify. So, if you're doing that and you're seeing that problem, we'll talk about that in another episode, but it's you could be just burying yourself in things that sound nice, but at the end of the day you're creating more work for yourself. So, it's all about clarity, getting everything organized and understand those workflows in your business such that you can simplify them and whether you say AI or workflows or tools or people, it has to make it easier for you to actually do those and do them in a repeatable and scalable way. So, how do you apply this without becoming reckless? Uh one simple rule. Automate drag not responsibility. AI can, like we said already, draft the email, but the human owns the relationship. AI can summarize the ticket, but the human owns the resolution. AI can flag the risk, but the human actually makes the decision. AI can prepare the report, but the manager owns that interpretation. And AI can suggest the next step, but the leader actually owns the outcome. And that's important because small businesses run on trust. And a bad automated message can definitely damage a relationship, especially if you don't have any visibility into it. Like, don't just automate things and let them go and and not see that because you could be alienating your customers that way by not checking things. A poorly designed workflow can confuse the team. So, don't automate blindly. Start with work that is repetitive, visible, and reviewable. That's the sweet spot. Repetitive, it means it happens often enough to matter. **[00:24:00]** Invisible means that you can see whether it worked or not. And of course, reviewable means that a human can check the output before it creates too much risk if it was wrong. Good candidates will include, like, uh, sales follow-up, drafts, uh, intake summaries, support ticket classification, meeting summaries, first draft for proposals, internal knowledge bases, answers, task creation, and status updates. Bad candidates are going to be usually your high-risk decisions that don't have clear rules. Don't start with AI should approve refunds over $10,000. And don't don't get it firing people, right? Maybe not legal judgments as well. Start where everyone agrees that there's a pain, and you can actually see the result, and that if it takes makes it a mistake, you're not digging yourself into such a big hole. And so, that's how you burn tokens safely. So, the the head count side I don't think I can stress this enough. It's not about getting rid of people. It's it's about finding ways to make people use their brains. Do things that are rewarding. And I might get some kickback about that when people watch these and and then they say, "Oh, but not everyone can do whatever creative work or things like that." That's not about It doesn't matter if it's creative or technical or logical or mathematical or musical or artistic. Everyone has a lot to bring to the table. Every single person in the world, no matter what your qualifications are, you have something to bring to the world. So, if you're stuck all day copying and pasting things from one tool to the other, **[00:26:00]** that's not a good use of your mind. That can be automated by a tool, by AI. While you get upgraded to doing something where you're actually moving your brain cells and you're doing creative work. Even if it's not artistic work, you're still doing creative work where the AI is not particularly good at. You're doing things where you can actually make a difference, where your judgment is counted. That's my opinion. If you're just adding people for the sake of adding people, it's the same as if you're just adding AI for the sake of adding AI. Make sure you nail your workflows first so that you're not just hiring into chaos. So, if the workflow is broken before the hire, the new person just becomes part of the problem. They're going to learn the messy process and either you're going to have more things to clean up and more communication barriers or they're going to invent their own shortcuts and then there's going to be another node of confusion within the system. So, just think about that before just hiring and then also think about that before just adding AI to a company. It's a two-pronged approach. Make sure you fix your workflows first internally with a small subset of people, small subset of tasks and processes. Figure out what problem you're actually trying to solve and be specific, not say we're busy. Ask what work is not getting done. What customer experience is suffering? What revenue opportunity is being missed? Which risk is increasing? And is it really a capacity problem or is it a workflow problem? We talk about the workflow problems a lot. A capacity problem means the work is clear, valuable, and already running well, but there's just there's too much of it. And a workflow problem means that the work is unclear and fragmented or very manual or it just depends on memory, which makes it fragile. Those two they need very different solutions. And then third, **[00:28:00]** ask yourself which parts of the role are judgment and which parts are just basically glue. Judgment will include customer conversations, complex decisions, creative problem-solving, leadership, negotiation, and quality review. And glue is just copying and pasting and formatting and chasing and checking, reminding and routing and summarizing and retyping. All that can be automated. So, if it's mostly glue, redesign before you hire more people. And then once you have that, then hire more people if it makes sense. So, it's not a huge thing. It typically takes a week or two to do some of these things, especially if you're just focusing on small tasks. So, think what small automation could reduce the pressure right now. Like a better intake form or a CRM trigger or a an AI summary, recurring report that you do all the time, a searchable knowledge base, and then think, "Okay, if we still hire, how can we we how can we make AI the new Like, how can the AI actually make that new person more effective?" And it's a good way to think about it, I think. Sometimes you should hire and automate. Do both things. Give the person better systems. Give them access to the company brain, so that they can make better decisions. Give them templates, AI-assisted workflows. Let them operate at a higher level from day one. That's how small companies scale without drowning in overhead. That's also where the company brain from the other video matters. If your knowledge is not captured or legible, then the AI cannot reduce much of the load. This is all very important. Everything we've discussed in the past few videos about legibility, about workflows, about knowledge in the company matters, so that you have context and you understand what is really happening in an in the day-to-day, and so that when you bringing people to actually use their minds and help you, they will do good work. Focus on what **[00:30:00]** inputs, outputs, rules, examples, approval points, success measures. Find that out beforehand, before hiring, also before automating, so that you have practical examples that you can use to create a workflow or enhance a workflow within the company. Now, if you're figuring out what you should start with in your company related to all this, there's an AI opportunity map on our website, processdriven.com. It's over at the top. Just click that. It'll ask you a bunch of different questions. Some of them that we've talked about today. It's completely free, and it'll guide you to figure out what you could be doing first that might start moving the needle. And you might be surprised. You're going to see how much you're putting on the table or leaving on the table by not doing a lot of these things. So, go through that questionnaire. You might get a surprise. Figure out what next step is so easy that you can just start today or tomorrow and start getting work done. And maybe you can understand, should I start to hire today or should I first fix and then increase my staff. That's it for today's show, and we'll see you in the next one.