S7E6 – How is AI Transforming Go To Market for B2B SaaS with Maja Voje

How is AI Transforming Go To Market for B2B SaaS? Inbound go-to-market for SaaS is undergoing a major transformation. What once relied on blog posts, lead magnets, and cold outreach is now powered by artificial intelligence. AI is no longer just a content assistant. It now fuels end-to-end workflows, drives strategy, qualifies leads, and personalizes outreach at scale. SaaS teams are deploying AI agents to track LinkedIn signals, automate follow-ups, and even manage outbound efforts. This evolution is unlocking new levels of speed and scale, but it also brings real risks if automation isn’t carefully managed. In this episode of the Grow Your B2B SaaS Podcast, Maja Voje breaks down how AI is reshaping inbound GTM. She shares what’s working today, where teams should stay hands-on, and how to build AI-assisted systems without losing the human connection that still drives trust in B2B. If you’re building or scaling a SaaS product, this is your playbook for doing it smarter with AI.

The Shift From Tools to Agents to Full AI Workflows

At first, AI tools were mostly assistants, great for writing content or summarizing a meeting. But things have quickly evolved. Today, we’re seeing a shift from using AI as simple tools to building systems powered by AI agents. These agents don’t just generate ideas, they take action. And the most exciting and complex evolution is what’s now being called multi-agent workflows. That means different AI agents working together to complete entire processes. For example, one agent might monitor your LinkedIn posts, another could draft replies, and a third might update your CRM all in one automated loop. This changes how work gets done. But it also raises big questions: Where should humans stay involved? What’s safe to automate? And how do we stay in control? Maja believes the shift is inevitable. Companies across the spectrum from scrappy startups to big SaaS players are already building these workflows. But with this new power comes new responsibility. Just because you can automate something doesn’t mean you should.

Why You Need to Earn the Right to Automate

In the old days, you could just launch something and figure it out later. But AI changes the game. You can’t risk letting an AI agent make decisions inside your systems without knowing exactly what it will do. That’s why Maja pushes a key principle: earn the right to automate. This means starting small. Test your ideas in a controlled environment. Don’t give an AI access to your entire customer list. Instead, run a pilot with a small data set and measure how it performs. Watch closely. Did it behave how you expected? Did it make any risky decisions? Only once you’ve built trust should you scale. In B2B environments, especially with enterprise customers, mistakes can be expensive. One poorly timed message or badly written email could hurt your brand or even lead to legal trouble. That’s why responsible teams run proof-of-concept tests, learn from them, and expand gradually.

Don’t Start With Tools. Start With Your Workflow.

When people first get excited about AI, they often rush to try the newest tools. But Maja recommends flipping that mindset. Don’t start by asking, “Which tool should I use?” Instead, ask, “Where am I wasting time?” and “Which tasks are repetitive or draining, but don’t need my personal touch?” By mapping out your daily tasks, you’ll quickly spot opportunities for automation. Maybe it’s handling LinkedIn DMs. Maybe it’s gathering content ideas. Maybe it’s sending follow-ups or tracking performance. Once you know your workflow, then you can bring in AI to help. Maja shared how she handles her own founder tasks using a combination of standard AI tools and purpose-built agents. For example, she tracks the best-performing posts from influencers she follows, trains AI to write in her voice using past data, and feeds that back into her process. That saves her hours every week and the system keeps getting better. Her advice is simple: if a task isn’t your core skill, it’s often smarter to use a pre-built AI solution. Save your energy for building what gives you a true edge.

Why AI Automation Still Breaks and What to Do About It

While AI workflows look amazing in theory, the reality can be messy. These systems often break under real-world conditions, especially when you’re dealing with large amounts of data or multi-step processes. Maja points out that scraping LinkedIn comments, pulling insights, and trying to automate engagement sounds simple, but the costs add up fast. And many tasks people try to automate actually go against platform rules. That means you could get blocked or banned just for trying to move fast. There’s also the issue of fragility. Even popular no-code platforms sometimes crash or behave unpredictably when real data flows through them. So before you automate anything at scale, test it in a sandbox. Make sure it works under pressure. And don’t plug AI into any sensitive systems unless you fully trust the vendor behind it. Her rule of thumb is this: if you wouldn’t share the data with a close friend, don’t give it to an unknown app online.

Building an AI-Powered Inbound Engine Around LinkedIn

LinkedIn is at the center of many inbound strategies today and AI can help you win there, if used smartly. The platform rewards content related to AI, so founders are piling in. But that’s created a flood of generic, bot-written posts that are easy to spot and easy to ignore. To stand out, Maja suggests a smarter approach. First, create a strong knowledge base by gathering everything your company has written, blog posts, whitepapers, recorded calls, webinars, and use that to train your AI. This helps ensure the output sounds like you and reflects your actual expertise. Then, bring in a bit of external research using AI agents that monitor relevant trends or posts from creators in your niche. Combine that insight with your knowledge base, and you’ll get stronger, more relevant content. But don’t stop there. Always review and refine AI-generated drafts. Add your voice. Add real examples. And when it’s time to publish, do it manually. Then engage in the comments and conversations that follow. Some founders do use AI to generate replies, but only after training those replies carefully. Vague, generic comments like “Great post!” don’t build trust. But specific, thoughtful ones can work, especially if the AI knows your tone and audience.

Inside Maja’s AI Stack and What’s Actually Working

Maja currently runs a handful of AI agents and is testing even more. Her goal is to eventually link them all into one connected system that powers her LinkedIn presence, content strategy, and lead engagement. It’s still a work in progress but that’s the reality for most teams. Some pieces are already saving her serious time. For example, she loads books and whitepapers into ChatGPT to help write high-quality posts. She uses agents to qualify leads based on post engagement. And she uses tools like Mainwish to spot top-performing content trends each week, then feeds that into her design team using visual mood boards in Miro. She’s aiming to build a system with 15 or more agents working together but she’s not there yet. It’s still about testing, learning, and improving.

The One Area Where AI Still Falls Short: Design

While AI can write decent content with the right input, design is still a weak spot. Maja has tested multiple creative tools for graphics, infographics, and visuals, but none have delivered consistent quality, especially when accuracy and brand standards matter. So instead, she focuses on creating better design briefs with AI help, then handing them off to real designers. It’s not perfect, but it’s far better than publishing messy, off-brand visuals.

AI Comes With Real Legal and Ethical Risks

AI makes it easy to connect tools, grab data, and push content fast. But just because it’s easy doesn’t mean it’s safe. Maja warns against blindly connecting tools to sensitive systems like Google Drive or CRMs. Even if an app seems useful, you need to know how it handles your data. Don’t trust every vendor. And definitely don’t scrape or automate actions on platforms like LinkedIn without knowing the rules. In Europe, posting or commenting on someone’s behalf using AI may even be restricted by law. Some tools have already been shut down over compliance issues. Always keep a human in the loop when it comes to anything public-facing.

Why Humans Still Matter Even in an AI World

AI is powerful, but it can’t replace taste, creativity, or judgment. Maja shared how she initially felt excited about AI, but as she got deeper into it, her expectations grew. Now, she sees herself not just as an operator, but as a tastemaker. She works with AI but never settles for the first draft. If you’re leading a team, encourage your people to experiment. Let them test tools. Make it safe to fail. But also help them keep their standards high. Don’t eliminate junior roles completely, AI doesn’t bring the fresh perspective a new hire might. And depending on your team culture, you may want to start by showing working systems before expecting everyone to build their own. Prototyping is getting easier too. Maja uses Miro to sketch out workflows and ideas before handing them to designers or developers. That helps teams move fast even without coding.

What to Expect in the Next Two Years

Looking ahead, Maja sees AI as a permanent shift, not a passing trend. It’s not crypto. It’s not hype. It’s a new capability that’s already reshaping how B2B teams work. So instead of rushing to catch the wave, take time to think. Be intentional. Experiment, but with guardrails. And most importantly, have fun with it. This moment is one of the most creative and exciting times in recent SaaS history.

Key Timecodes

• (0:00) – Boosting AI Content Performance & Automating Founder Workflows
• (0:53) – What Is AI’s Role in SaaS GTM? [With Guest Maja Voje]
• (1:48) – Is Everything Dead? Why AI Agents Are the Future of SaaS Workflows
• (2:55) – Multi-Agentic Workflows Explained: Tools, Agents & Human Oversight
• (4:28) – Why You Must Earn the Right to Automate with AI
• (5:27) – SaaS Automation Gone Wrong: Avoiding Enterprise Pitfalls
• (6:15) – AI Agents: Build or Buy? Key Considerations for GTM Leaders
• (6:38) – Mapping GTM Workflows: LinkedIn, DMs, Offers & Content Ops
• (8:00) – Real-Life AI Marketing Automations You Can Use Today
• (9:43) – How Many AI Agents Do You Really Need for LinkedIn & Lead Gen?
• (11:08) – Iterating AI Models Post-Training: Prompts, Builders & Feedback Loops
• (12:55) – AI Costs, Compliance & Rollouts: From POC to Scalable Deployment
• (15:07) – Data Security in AI: The Case for ‘Least Privilege’ Access
• (16:04) – Rule of Thumb: Don’t Share Data You Wouldn’t Give a Friend
• (16:13) – Sponsor Spotlight: SaaStock Dublin—Investor Matchmaking + Discounts
• (17:22) – Inbound Marketing with AI: LinkedIn Trends & Time-Wasters to Avoid
• (18:54) – External vs Internal Knowledge Bases: Training AI Without Garbage Input
• (20:31) – Why AI Design Often Fails: Creatives, Claude vs ChatGPT & Brand Gaps
• (21:53) – LinkedIn AI Strategy: Commenting, Publishing & Legal Risks in the EU
• (23:30) – AI-Powered Outbound Marketing: ICP Scoring, Lead Research & Social Selling
• (25:52) – Training Your Team on AI: Avoiding Content Quality Pitfalls
• (27:26) – Human-in-the-Loop Design: What to Automate vs Delegate
• (28:43) – The AI-First Founder Mindset: Culture, Talent & Psychological Safety
• (31:20) – AI Implementation Choices: From Prototypes to Governance Guardrails
• (33:29) – PR & Leadership: Why ‘We Replaced 7 People with AI’ Is a Bad Look
• (34:10) – 2-Year AI Roadmap: Think Strategically, Reflect Often, Stay Safe
• (36:20) – Going from 0 to 10K MRR: Learn to Sell, Test Pricing, and Stay Focused
• (38:53) – Bootstrapping with AI: Don’t Waste Model Credits, Focus on ROI
• (39:32) – Scaling to $10M ARR with AI: Ecosystem Marketing & Creator-Led Trust
• (40:47) – Recap: AI Workflows, POCs, LinkedIn Automation & Strategic Thinking
• (42:35) – Connect with Guest Maja Voje on LinkedIn
• (42:58) – Subscribe to the GTM Strategies Newsletter on Substack
• (43:28) – Final CTA: Review the Show, Sponsor, Ask Questions, and Connect

Transcription

[00:00:00.000] – Maja

Algorithmically all the AI-related content performs well. Whenever we are writing about AI, it’s expected to go well. However, what you will see very soon is that you will get a bunch of AI-generated content there. I have been asking myself as a founder a lot, what is the amount of admin work that I am making? How much time does it take me to send out an offer? How much time does it take me to do my DMs on LinkedIn? How How much time does it take me to do the commenting routine so that my posts then perform well on LinkedIn? I have been thinking really hard, what could I potentially automate or even potentially develop an AI agent to perform the task instead of me and gets better and better and better.

[00:00:53.120] – Joran

Today, we will be discussing how AI is changing the go-to-market motion for SaaS companies. We will be focusing mostly on the Inbound channel. My guest today is Maya Voje. She has been a guest before on the show in Season 3 already, where she talked about product market fit and go-to-market. We went into the basics of go-to-market strategies, product market fit, and the challenges founders might occur. Almost two years later, a lot has changed. I had to bring Maya back on the show. Welcome to the show, Maa.

[00:01:19.460] – Maja

Welcome back. Thank you so much for having me. I’m super excited to just explore a little bit what happened in your business, what’s happening in my business, because I think with AI, we are just figuring out together, learning by doing. Let’s exchange a couple of best practices.

[00:01:34.960] – Joran

Yeah, sounds good. I typically ask all the questions. Feel free to ask me, but people are here to listen to you. We talked about two years ago, what has changed? You see so many companies in side of companies, the inside of companies. What has changed, and maybe what’s coming up in the near future.

[00:01:48.920] – Maja

Haven’t you heard? Everything is dead. Outbound is dead, ICP is dead, LinkedIn is dead, everything is dead. I don’t know how companies even exist anymore. Ai has changed a lot. We can safely and surely say that not only AI tools, but also AI agents are here to stay. I have been working almost exclusively, like 90% of my clients have been AI agents, and I’m just not talking about AI-first companies that we can explore later on how are they different to SaaS companies, but also the big ones. Amplitude is launching agents, user pilot is the strategic agent. Everybody seems to be launching agents. It got even more interesting. Now we are transitioning towards multi-agentic workflow. You know what? Work got really interesting for me again because two years ago, I felt that we are still repeating those roadhacking playbooks and even by HubSpot and Dropbox referrer programs. But things got exciting again.

[00:02:50.150] – Joran

That’s fun. For listeners who don’t know what multi-agentic workflows mean, what is it?

[00:02:55.840] – Maja

If you’re okay with ChatGPT, this is the first stage, that you’re just comfortable of using AI tools in your personal or business life. The second stage is that you trust AI to actually perform some actions. This would be like an AI agent that you are using AI intelligence, artificial intelligence, to just perform some tasks. Whenever we are mixing this with automations, like classical automations, just up here, it can be a little bit strange. But these agents are self-learning and can It’s like literally a walk. Not a workflow that you set up once, it’s something that gets better in iteration. Now we are in stage three, which is like us connecting different agents together to the workflow. Still, we have to figure it out what’s the role of human here. Why? Because imagine if you let your AI agents or a cluster of agents into your CRM and said, Okay, agents, go ahead and email all my customers. Get me five new yearly contacts. Come on, let’s go. Technology-wise, as a society, we are just not there yet. We are exploring AI agents and human collaborations. Human is usually still in control. I do think that here, what is super interesting is that it’s redefining our role as workers, as somebody who has been a manager, a founder, or something like that, because now we are happy to deal with human intelligence as well as artificial intelligence.

[00:04:28.340] – Maja

It just makes things a little a little bit different. The underlying premise that I have been seeing everywhere in all AI launches is earn the right to automate. For example, if you have an AI SDR, you won’t say, Oh, now get me my ICP from LinkedIn and go ahead and send those messages. Sometimes you have been in touch with these people before and it would look awkward if you just send them a cold DM. We need to just make sure that first we are deploying stuff on smaller pilots. As we see that they act normally, that we can trust it a little bit, we can expand. This is the right to automate that I think is the underlying premise of everything, what we’re doing with AI at the moment, because people are afraid. I’m afraid, like what will happen if I let the agents out to dance and do whatever.

[00:05:21.220] – Joran

In the end, it goes back to the fundamentals. Make sure everything is working before you go all in on it.

[00:05:27.070] – Maja

Which is different to what we were doing previously with launches, right? Because before, the premise, my very much preferred approach would be to just go out there and then reverse engineer what worked well in practice. But now imagine, if you are dealing with enterprise environments, for example, and you say that you are going to automate their support, and then their support agents are selling their $10,000 deals for $5 or something like that, it didn’t working. We need to make sure that we have this pilot project that we measure and see the result. Then we earn the trust to automate with the client, and we create some leverage to attract more of such clients. That has been just like the usual go-to-market route that I have seen for those agents and platforms.

[00:06:15.540] – Joran

You mentioned a couple of tools at the beginning. How do you see it? Every tool is launching their own agents. Is that typically where people get started? Should they maybe be focusing on tools like N8n where they can build their own agents? How would you recommend people now starting to kick out as well and they want to get started, just use the tools, which are the features already in the tools or go all in on themselves?

[00:06:38.670] – Maja

Okay, I like to start with the workflows, right? I have been asking myself as a founder a lot, What is the amount of admin work that I am making? How much time does it take me to send out an offer? How much time does it take me to do my DMs on LinkedIn? How much time does it take me to do the commenting routine so that my then perform well on LinkedIn? I have been thinking really hard, what could I potentially automate or even potentially develop an AI agent to perform the task instead of me and gets better and better and better, unlike the majority of virtual assistance that I have the pressures to work with, because that was a little bit of a linear process of the improvement. But nevertheless, ideally, I would be able to automate these workflows end-to-end. However, that’s rarely the reality. I would start by mapping out my workflow. For example, if we say now inbound, it starts with me monitoring other creators, what are their best performing posts, then learning from their best practices, training my AI writers to perform the job better and better and better, and feeding it back to what was actually performing well for me so that we can potentially evolve and create better and better and better posts.

[00:08:00.000] – Maja

That would be one part of the workflow that I would definitely like to automate because if I’m just scrolling there on LinkedIn and I say, Oh, Joren posted something interesting, save. Then on Sunday, when I’m doing my content analysis, I would be like, Screenshot. Now The design team, he did something interesting. It performed well. Now, figure out what should we do in order to replicate his success. Yeah, that would be the perfect way how I would start. Then in the second step, I would have to decide where Where would I typically go with just normal automation. For example, I have one, I don’t want to call it agents, but it’s something, Mainwish Replyet, that sends weekly updates of which were the best performing posts from influencers I’m following. Based on that, there is something more intelligent that is actually done in any 10, that creates a Miro mood board from their best performing creatives. Now, I suddenly eliminated two hours Two hours of legwork. Two hours that would take me to just do the screenshots, then organize the screenshots, and then trying to figure it out how I should be using the best practices for design briefing.

[00:09:10.720] – Maja

This is my way how I’m currently tackling this. I buy custom with AI agents and copilots as well. For example, for outbound, I am developing, not that intelligent. For outbound, I just go with pre-trained solutions and try to figure out how to make them work me. Then we have this build versus buy dilemma. For outbound, that’s not the core expertise of my company to build outbound agents.

[00:09:39.620] – Joran

How many agents are you currently using? I don’t know. Can you count them?

[00:09:43.700] – Maja

Five are normal and others are in the testing, approximately seven. We have this prototype of end-to-end workflow for LinkedIn and content creation. I call it AI-assisted workflow. There are combinations using normal AIs. For example, if I build a knowledge base, put all my books and white papers and stuff like that into a project in ChatGPT, and then train it, prompt it to just create good LinkedIn post for For me, that is not an agent. This is like an AI build me using external tools. But once these posts are actually published on LinkedIn, I can be using agents to do lead qualification for me, who from the relevant ICP, but that engaged. I could be using agents to just do a little bit of outbound, a little trip to those people. That is really interesting. Potentially, I would like to bring together approximately 15 to 17 agents, but it’s not working yet. It’s still being developed.

[00:10:48.330] – Joran

I think this is an interesting topic. You’re now using them, you’re iterating them. It’s not working yet, as you mentioned. A lot of people face this, where they might want to start working with agents. Of course, if you buy the pre-built things, it’s going to work a lot quicker. But still, it needs a lot of iteration. It needs a lot of information to actually make it work for you.

[00:11:08.940] – Maja

Then the post-training. Post-training is super important as well. Not only to get it, to get a decent output out of this one, but feeding it back. This performed, this did not perform, this was problematic. From general AI use, I would really recommend to everybody to do something super simple. First think about your workflows, where are you wasting most time, and either start with something which was already purposefully trained or with really simple automations that you can do in 8n make or just other AI agents tools that you are using. I don’t want to be name dropping too many companies at the moment, but there are visual builders.

[00:11:52.840] – Joran

Some are more tactical than others. We, for example, built this huge workflow and it kept breaking all the time. Tell me more about your work, love? In the end, we help SaaS companies set up an affiliate program, right? And one of the biggest challenges they have is to find relevant affiliates. We built an agent, for example, who could find relevant affiliates based on keywords, based on backlinks, based on where they’re being mentioned and there’s no backlink yet. So just finding content which is relevant for them. It processed a lot of data. In the end, we made it go looping, things like that. But at the beginning, it kept breaking all the time. So we ran into all the limitations. I feel like when I look at LinkedIn, everybody’s posting these screenshots of these most complex. I don’t think they work or I am crazy, but I don’t know how you see it. As you mentioned, you’re still iterating it. Things break all the time. You’re now building your end-to-end workflow, which I’m not at, I’m not going to change yours fully. Is it worth spending all this time to at one point have a working end-to-end workflow which might break later and you have to keep it trading?

[00:12:55.540] – Maja

You mentioned something super interesting, not working and costs. Because even for simple jobs such as scraping all your LinkedIn comments, extracting ICP from it. These are costs and you would have to hire external services to do this. Plus, it’s a pretty heavy violation of their LinkedIn terms of service that can get you blocked. There are, of course, risks and liabilities adjusted to it. Whereas I’m a fan of wide working, meaning encoding whatever you like to do in your spare time, I do think that those bills have, first of all, security breaches. They are at their current stage, undeployable to any serious B2B environment. The second thing is just like if it’s not going to work for you as you are having this testing, almost like perfect environment, then it’s very unlikely that you can be even thinking, daydreaming about product lead, because if you let your actual ICP and to play around with this, things can go super bad. In the intermediate time, I am working with a serious team. They are selling AI agents to enterprises, Nasdaq companies. An approach to use in practice is that first they are doing POC, proof of concept, so that they set up these agents to work with the technology set up in the company.

[00:14:18.640] – Maja

Then they have these testing phases, three or four are usually normal. Then it’s being deployed to the work process that they are trying to do. Then once this is normalized, it is getting expanded. This is like gradual adoption pattern that I have been seeing a lot, especially in larger environments. But as I’m testing stuff, there are gazillion tools out there. With some of them, I have serious concerns to connect my drive, my Google account, upload assets, because I don’t trust these people, and I think that’s normal. Learn the right to automate by doing, by peering, like a legitimate company. I think it’s not going away anytime soon in B2B where the majority of big AI agents use cases are.

[00:15:03.560] – Joran

You mentioned something interesting, like don’t just connect all your tools to all.

[00:15:07.200] – Maja

No, don’t be an idiot.

[00:15:08.770] – Joran

Don’t. There’s a lot of risks to it, and it’s super easy to make the connection because it’s often a two-click button, and then you’re giving access to everything.

[00:15:16.630] – Maja

Isn’t it funny? We have to literally talk about this. I should be talking about this to my boomer dad, who’s 60, and he’s having right now the spree of AI-generated songs, and he’s sharing them on Facebook, and he’s sending videos, AI generated like, Good morning, here is a cup of coffee and a dog, and somebody skiing images there. I understand because he was not digital native. But to explain that to people who are working in marketing and sales of B2B companies that get me a little bit concerned.

[00:15:49.820] – Joran

Yeah, and I guess that’s like… How do you say it? They don’t understand the consequences of just connecting something. They want to try new things and tools force you to do certain things. It becomes a bit too easy to connect your entire Google Drive.

[00:16:04.630] – Maja

Maybe we could have a little bit of a mentor shortcut here. If you would not send this to your friend, don’t give it to a random company.

[00:16:13.840] – Joran

This podcast episode is brought to you by Saastock, my personal favorite SaaS event and community. The Europe event is taking place on the 14th and 15th of October in Dublin, and it’s going to be two full days of investor matchmaking, talk from other founders, and meetings, which will be a huge focus this year as they’re planning to have over 15,000 pre-scheduled meetings before the event. As a listener, you’re getting an exclusive 30% discount through the link in the show notes. If your company is already doing above 1 million AR, you might even qualify for a completely free ticket while they last plus up to €400 reimbursement towards your travel expenses. Discounted and free tickets are limited, so feel free to pause now, hit Hit the link, buy your ticket or get your free ticket, and then I will see you in Dublin. As right as it is, we’ll have a booth there as well. So make sure you get your ticket and then come back to this episode. Let’s go back to the beginning because we’re talking about AI and inbound, right? What do you think? It can mean a lot of things. You’re talking about generating content on LinkedIn, but it could also be the LLMs showing results.

[00:17:21.840] – Joran

What do you think?

[00:17:22.850] – Maja

I love it. When we are talking about inbound, we first have to acknowledge that algorithmically, my go-to channel usually is LinkedIn, but I’ve been hearing things from Reddit and X as well that algorithmically all the AI-related content performs well. Whenever we are writing about AI, it’s expected to go well. However, what you will see very soon is that you will get a bunch of AI-generated content there. Why? Because people know that in their LinkedIn routine, for example, they should be first 10 minutes behaving as a human and doing 10 new commenting, for example, summarizing posts from other creators. Then they can have their posts and they should engage with these comments. Then there is the period of nurturing when they were supposed to go and have more conversation as well as keep on engaging with other people. That in a normal founder’s workday would last approximately 3-4 hours. I know people work in the legit businesses, not content creators alone, that are spending 3-4 hours a on LinkedIn, and I don’t think it’s the best use of their time. What is super interesting is to find these placements for the automation. Again, if we are extending it to a workflow that would only produce LinkedIn posts and do assistance in engagement and just post-nurturing to actually generate some smart leads, I would first limit the project.

[00:18:54.820] – Maja

What could be very interesting, especially for high tech companies, companies. These are a couple of builds I’m doing with clients, but nothing that I can really talk about this publicly yet. We have to do this together with a lot of training. Build a single knowledge base with everything you have, like your cool recordings, with your previous blog posts, webinar recordings, yare yare, and get these writers, content creators going on. You could be connecting this to research agents to pull in external intelligence in the system. Here we have external intelligence. Here, this knowledge base is internal intelligence. You could be bringing those data sources together. But for the purpose of just empowering marketing to write legitimate blog and LinkedIn post is a minimal viable deliverable. Everybody is doing this with ChatGPT. Let’s adopt the logic, Garbage in, garbage out. If I ask with a poor prompt, ChatGPT, to write me a blog post, this ain’t working, especially if I’m working in compliance, regulations, industries, and my marketer there is like somebody who just came from college, maybe from a different country that is not a domain expert. The first job to be done would literally be to bring bringing knowledge sources into the single knowledge base, database, however you want to call it, and have good inputs to your AI writers that you would purposely train with prompts and maybe also bringing additional intelligence.

[00:20:31.060] – Maja

Next, it’s about the workflows for just like distribution. Here I have a huge missing part in my process right now. To my best ability, I haven’t been able to do anything beyond recommending design briefing in this workflow. I have tested AI-generated designs, but usually it just breaks. Whereas some of my customers have been testing AI-generated creatives very successfully with their advertiser advertising for my stuff, for infographics, and just other, let’s say, knowledge-heavy materials, I haven’t seen good results myself. Maybe I’m using the wrong type of AIs because there are literally AIs that are trained to create videos, educational videos for pharmaceutical companies based on a similar type of setup that we are talking about right now. Maybe I don’t know how to do this, but to my best ability right now, I haven’t been able to tackle this. Now we have a complete asset. We have text that was generated by AI model of your choice. By now, my ChatGPT is well-trained, but most of my peers use plot for copywriting. It works better for them. I prompted it with just a couple of best practices. How long it should be, how it should be formatting, white space.

[00:21:53.620] – Maja

It’s sourcing from my knowledge base, so I’m not writing something super stupid. Now I have these posts, potentially a LinkedIn post. Then I think about publishing it and doing the LinkedIn routine. Publishing in LinkedIn is very tricky because if you just said it and forget it, can’t post and ghost. This is not the best practice. The majority of people manually do the posting, so the posting set, the post set up, and then this engagement routine. Here is where things at the current stage of this workflow could be very They’re really divided. There are people who love AI-generated comments and they’re using them all the time. Sometimes I am tricked myself that these comments are coming from real person, a technical expert, but These are usually comments that are trained on your language and actually have some basis. These are not comments such as, What a great company. Congratulations. I agree that leadership is important. Again, garbage in, garbage out. Commenting could be potentially automated, but you have to be careful with inputs. Then outbound is automated. I don’t know a single person who would do outbound manually by now. We want to have is another AI-assisted workflow or automation that would literally extract those leads, do the research on these leads, achieve an ICP scoring process, something like that, and at the end, help us shape the message.

[00:23:30.000] – Maja

Now, I would still, as a human, like to be able to edit the message and edit the comment before it is being posted. I also think that right now in Europe, and you’re more familiar with it because I mainly work with the US, that you cannot really just be pushing this AI-generated stuff because this is not legal anymore. I’m not mistaken. Allegedly, you cannot post on behalf of people. This is what I heard. I have seen some commenting tools even shutting down their business for that one. Tapio is still up and running. I don’t know how, but they are. The thing is that from there onwards, I could be connecting this input to the process. My domain knowledge, like assets that I have, towards a tangible business opportunity. Because outbound, I’m not working on my LinkedIn post and on my post because I would be bored and this would be super awesome and inspiring for me. It is sometimes, but it has a business function to it. It’s generating my pipeline. If I’m just there passively waiting for people to say, Hi, you’re all so glad to work together, I would be waiting for a very long time or I wouldn’t even be able to escape my team.

[00:24:43.250] – Maja

I have to make sure that I’m doing what is called social selling. I call it signal-based outbound, which is one of the signals why somebody would buy for you. This is currently what I have been thinking. Now, that has been literally just a workflow for LinkedIn. Potentially, Usually, that could be done for other content marketer jobs. But where I have problems right now is seriously with design, with just making sure that the system is properly trained, because if this would be product-led, and I have seen this, seriously, I told people, at least do a project on ChatGPT, put a couple of your assets there, and then look, I will give you prompts. It’s not a problem. Just start using it. They were too lazy to upload assets. They didn’t know what assets to use, how to edit this stuff. Then they ended up just like, Jeff, the first mistake is to accept the first output it offers. The second mistake is not to do any human editing because there are little things, such as like symbols that it’s using. Some It’s a spacing. If you know any of these tools, you will see from a mile what is AI generated.

[00:25:52.620] – Maja

We need to avoid that.

[00:25:54.160] – Joran

I think the APIs are not there yet to really make the designs you make them, like the infographics. You have a bit more attention to detail than most of the visuals out there, I think. In your case, it’s going to be really hard to create them, I think.

[00:26:07.180] – Maja

We can share a couple of links to the episode. Like a partner who’s doing AI-generated creatives for LinkedIn ads that are out for marketer-generated briefing. We can share how Amos from Swann AI, so it used to be SDR, now it’s a GTM engineer, is doing research and lead qualification with their agents. He has a couple of independent agents going on right there. We can share this article, how I think about this inbound system for now, the set-ups that we are testing with my clients. This is super interesting to explore, but as everything, you have to adjust it to your processes. Whenever we start this conversation, even on done simple steps such as inbound, because everybody’s more or less doing the same type of workflow, you will find these nuances. You need to start by mapping your workflows. You You need to make sure to realize that you still have to do some human resources and onboarding into that. I think that currently we are here to the best of my knowledge.

[00:27:11.640] – Joran

When you map out the workflow, and then when you look at the workflow, how do you determine where does it need human intervention? Where do you need to supervise it? Because you mentioned, for example, LinkedIn, the dashes, the spacing. How do you determine what needs your input?

[00:27:26.520] – Maja

Even when I’m thinking about outsourcing this, what What is something that a VA could do for me? Because at this stage, most AI implementations are still there with junior/VA. We have been a little bit disappointed in marketing with ChatGPT five updates, but I’m thinking in a way, what’s the type of work that I would like to get rid of? Where do I think that as a human, I’m not creating huge added value? This is a bit of a strategy, right? What are your strengths? If you are super good seller, of course, if you are selling something expensive, I would do social selling myself. But if I’m selling some crap, like webinar in this or something like that, that’s the first thing for automation because I’m copy-pasting messages, but come on.

[00:28:10.250] – Joran

Yeah. Maybe a different question. We have a lot of SaaS founders. They have employees, right? They probably want to have their employees using AI. But what advice would you give for them as in… Because we talked about security, of course, like connecting the tools. Actually, for example, be able to upload the things you provided to a client and not even doing that. How do you help them to start leveraging AI and not think about, It’s going to replace my job, but it’s actually going to, as you mentioned, automate the things you hate the most. How do you change the mindset?

[00:28:43.830] – Maja

This one is difficult for me as well because I have been marketer for 15 years and I take certain joy and certain pride into what I’m working on. Of course, there are two stages of my relationship with just the mainstream tools. I will use ChatGPT as an example. First, I was like, Wow, she’s getting to go dead. The teeth is cool. Then I developed a taste for it. Now, I never, ever, ever would accept the output that it is given to me. It would be a prompting war. To get it perfect. I see myself more as a tastmaker. But the thing, just like an attitude-wise, is to start thinking as a founder, what will not change. What is super important are relationship with humans, managing humans, stakeholder stakeholder management, creative thinking to some extent. Some would say strategic thinking. I really present it as what’s the you absolutely hate, what would you like to get rid of, and how we can use AI in order to do that. Then we come to just this opportunity to have them trained. I’m not promoting that you should have an internal consultant or hackathon. Maybe your culture is like builder culture.

[00:29:55.810] – Maja

Assign them tools and time in order to start thinking and make them feel psychologically safe. This is easier said than done. But as business owners, we are doing ourselves a great disfavor. If we are just thinking, I no longer have to hire junior, just everything will be super AI generated because we’re not bringing new talent to the business. They think differently every 10 years. Just like humanwise, I work with people who are 10 years older than me, and then I work with people who are 10 years younger than me. Just us working together, there is something beautiful to it. Everybody brings their own set of experience, their own set of ideas. It’s different to work with humans than AIs, because with AIs, I’m still constrained with my imagination, what can they do? At the moment, they are following our instructions. I really think that these big clips, big steps happens when a human think together. There is another component to it. I would tell it as it is. I would say, This is here to stay. We had to learn as an organization. I would probably compare it with how my mom had to learn how to use a computer, which was super fun.

[00:31:08.720] – Maja

She still sometimes. But yeah, it’s here. These are the tools we are using right now. These are just like everybody’s using these tools. We have to, let’s figure it out how.

[00:31:18.550] – Joran

Nice. When we show it.

[00:31:20.960] – Maja

That will be super interesting to just hear it from you as well. Probably your developers did not resist. Tools are the date.

[00:31:28.100] – Joran

On the business phase, so we ended up, like hiring an agency who could help us set up the workflows. Then we had them create things first before I introduced it back to the team. That’s one thing, right? At least the learning curve is a bit differently that we already have something working and then We can see, Okay, well, how does it work? Does it actually work? We don’t have to go through all that hassle. For example, we did not connect our code to AI or things like that. There’s still limitations there. We’re not just going to go rogue on everything.

[00:32:00.000] – Maja

I have huge reservations towards doing that as well. I don’t know why. Technically, we know why. But also health information, what stage is it made. If SaaS was scary, this is scarier.

[00:32:14.360] – Joran

Yeah, exactly. It’s really nice. For example, I’m non-technical, so nowadays.

[00:32:19.500] – Maja

I didn’t know.

[00:32:20.040] – Joran

No, it may be pretend to be, but-Probably the guests, no. No, but now I can at least build a prototype and show what I want, maybe iterate 20 times Because it keep rain all the time. That’s really nice for me.

[00:32:33.090] – Maja

Now that we are thinking inbuilt AIs in the tools, I’m doing prototyping in Mira, which we both work with. I have been delighted with how it’s creating those lofi prototypes for my designers. I mean, just explore it, but have the courage to just take it for a spin. What I like about what you’re saying is you control an engineer to the architect of this pilot, so you could illustrate directly how they benefit and how it’s connected with their workflows. Because when my dog should get vaccinated, AI agents are fun to build, but practically useless.

[00:33:08.550] – Joran

Yesterday, I organized a SaaS start local in Amsterdam, and we had the CEO of Miro chatting, and he even mentioned during his holiday, he was prototyping and building prototypes with Miro and Lovable as well, where he was basically building things during his holiday to come back and have new ideas to the team. They have 1,700 people. I thought it was pretty crazy to hear.

[00:33:29.530] – Maja

No, Obviously, from the CEO’s perspective, what I don’t like is those people shouting out on LinkedIn like, I fired seven people, and now somebody that I paid 20K. If you ever can do the job of seven people. I just think it’s dumb and short-sided because at the end of the day, as long as humans are transacting with humans, we will find ways to keep ourselves occupied. I don’t think it makes leaders look good.

[00:33:55.950] – Joran

Exactly. We’re going to start wrapping things up. I have two final questions. We spoke took almost two years ago. We didn’t foresee the future back then, but let’s see if we can foresee the future this time. What should founders prepare for the next two years in your opinion?

[00:34:10.110] – Maja

To really, I can raise this AI in your organizations. This is not the next crypto. This is not going away. But to be sober about this, to still keep your taste making, to adjust it to your specific business needs. This is something that I’m putting on top of my agenda for running my company as well. We’re sitting on a big opportunity. Cards are being shifted, shifted to other… So yeah, there are opportunities for new players everywhere, and I would definitely pay an eye on this one. I am like, this is where I’m putting money on the table as well. The second thing that I would like to encourage everybody to do is to think strategically. This will sound so stupid, but at the current stage that we are, everything is changing so fast that it is really hard to find time and bandwidth to just think about your business holistically. Whereas everybody is super pressured on productivity and on the hunt for the ladies and ladies next to, I do think that the beauty in my business happens when I have a lot of time on Sunday and it’s raining outside, where I can be just exploring, thinking freely, reflecting on what was going on, and make decisions at a the heart rate.

[00:35:31.010] – Maja

You are not fresh to react on something, but you can really, really, really think how this is impacting you and your business. The third thing that I would say is that we live in the most exciting times in the 15 years of my career, so have fun doing it. Be hands-on about this. Don’t be afraid. You don’t have to read or listen to three more podcasts before you get something done with AI. It’s learning by doing, having fun with it. As we said, very nicely, being very cautious with what information do you put to the tools that you don’t trust and know.

[00:36:06.680] – Joran

I think that’s something we have to keep mentioning. You already answered these questions in the past, but you might have different answers now. What advice would you give a SaaS founder who’s now just starting out in this era and on their way growing to 10K MR.

[00:36:20.650] – Maja

Learn how to sell. Because the first control prototypes, you will probably have to either acquire from warm outreach or beat introductions from investors, friends, or just if you have the possibility to do outbound in your industry, either on LinkedIn or email, that would be amazing. Usually, this is the fastest and the most secure way how We can just get our foot in the door to do some pilots. The second one is that pricing is getting increasingly interesting. Now everybody is doing this credit system. But then when you think about this from the buyer perspective, I don’t know how much exactly it’s going to cost me to book 50 meetings. With human, I know, with these AI agents burning random credits, if it is working or not, I believe that result-based pricing, outcome-based pricing, as well as just like the predictability of the costs will dominate the conversations in the next year. For now, you will be able to present your implementation as a pilot. You can say there is a price per pilot. This is what’s going to be deployed. This is how they can see if it provides value for it. After they experience the value, you can negotiate a higher deal.

[00:37:38.750] – Maja

Use the pilot logic for your pricing. It’s an easier sell than if you are selling them something which is super far fetch and meet 13 people in the decision-making units to buy. Last but not least, don’t give up too fast because as always, there will be many things that are not working. It’s increasingly hard right now on AI-saturated LinkedIn to have a fair share of organic human attention. Some of the playbooks that have been working decently well before no longer work. You can either play the game, which was the punch of AI, to simplify your life a little bit, or find something that is working. Maybe it’s going to be an old-school channel. Maybe you will have to call customers. Maybe you will have to go and present something in the fair. Maybe you will have to advertise in a freaking newsletter. How is it called? Newspaper. The media, the printed stuff. Just keep your mind open and go where the audience is. This is the best advice that I can give you.

[00:38:41.960] – Joran

Nice. Really gone old-school to stand out in this market. I think the pricing was interesting. The pilot cases allow you to determine the cost, which also allow you to deal with the margins you actually have to sell your product for, which is interesting.

[00:38:53.120] – Maja

That’s really important. I think that we are both very much into this camp. Bootstrapping, just like not operating with funds sufficient enough to really do the subsidiaries for subsidizing the adoption. Don’t burn a bunch of AI model credits for some people that will never pay for the product to give you the wrong signals, what you should be developing. For me, payable client is the ultimate respectable product market.

[00:39:23.520] – Joran

Nice. Let’s assume we pass 10K MRR and we’re going to make a huge step, I know, towards 10 million What advice would you have for SaaS founders here?

[00:39:32.450] – Maja

Think about ecosystem marketing. I have been in close touch with Clay and Haric, and they have both used agencies, partnerships as go-to-market motions as a close level of adoption. It literally not a big learning curve for them, especially if you are doing something which is complicated to implement and companies will not necessarily implement it themselves, such as outbound tools for the majority of people. This is a little bit of an overkill. They already trust agencies. One agency can bring you 12 clients, and this is a super, super, super, super important lever. The other one would be actually a water on your meals. In the age of deep AI fakes and in the age of AI generated everything, people still want to connect with people. Communities, creator, influenceurs, these are still like drastic apps for now. As for the mainstream media and advertising space getting more intense. I just think that we have these shortcuts to trust that is getting increasingly important in B2B. And last but not least, use AI. Think about how AI is influencing each and Every go-to-market motions that you are using.

[00:40:47.520] – Joran

That is nice. That’s the circle. I’m going to make another circle of this podcast. I think we recorded 40 minutes, try to summarize it in 40 seconds, and no AI is involved in this. When we think about AI, I think in workflows, ask yourself, what do I hate the most? Eliminate manual work. Ai agents are changing everything, but earn the right to automate a smaller pilot first, do a POC, trust the automation, then fully automate. There are going to be costs and risks to automating. Don’t just connect all your tools to the automations. Think about garbage in, garbage out. Ideally, build another space, external and internal. Add your domain knowledge, create a GPT, prompt the AI to reuse it, and then think about adding webinars, help centers, blog articles, sales calls, everything you have to be able to have the AI generate good content for you. When we talk about LinkedIn, when automating posts, make sure you also engage with others, as you mentioned. So have that window of commenting. And ideally, do human ads. Don’t simply copy paste because people will notice that it’s actually AI generated. When we talk about the future, think strategically, take time to reflect and make decisions with a steady heart rate.

[00:41:53.960] – Joran

Implement AI. It’s here to stay. So don’t adjust it to the needs you have. Revenue stages, 10K MR, learn how to sell. I go old school, do paid pilot cases and do not give up too fast. 10 million AR, think about ecosystem marketing. Trust is going to be really more important and leverage AI. I wanted to keep for last. Have fun doing things. Don’t be afraid. Learn by doing. Loved.

[00:42:19.640] – Maja

That was like 40 minutes of podcast in two minutes. Awesome. Exactly. Well done.

[00:42:22.720] – Joran

Exactly. Nice, nice, nice, nice. You mentioned a lot of articles. I read the articles before jumping the podcast, so we’ll make sure to link to them. If people want to get in contact with you, what is the best way to do so?

[00:42:35.500] – Maja

Definitely LinkedIn. But if you can please send me some note with connection request because I’m being bombarded with outbound and this is getting out of hand. So please, please, please show me that you are a human and I will immediately and with pleasure connect with you.

[00:42:51.480] – Joran

Nice, nice. We’re going to link towards your newsletter, but do you want to give it a quick plug? Where can they find it?

[00:42:58.110] – Maja

It’s knowledge, sub stack, GTM. You just write GTM strategies, Substack, to whatever search engine you are using and you got this. So we will link it to the episode. This has been super interesting for me on Substack as well, because as I am implementing this stuff, I I’m also writing, but I’m also learning from that are implementing. That has been a fun intellectual experience for me to be able to work, learn, and write.

[00:43:25.740] – Joran

Nice, exactly.

[00:43:26.650] – Maja

That order, very important, please.

[00:43:28.740] – Joran

Learn first and then write so you can actually learn from or learn while writing as well. Cool. We’re going to add a call to the podcast episode on Spotify. If you’re listening on Spotify, let us know what you thought of this episode. As a general, leave us a review if you haven’t, so we can boost the algorithm so we can help more SaaS founders. Thanks again for coming back, Maya.

[00:43:49.890] – Maja

It’s been for fun. Always great to connect with you. And I’m impressed by the consistency and the quality of this spot.

[00:43:56.100] – Joran

Nice. Thanks. Thank you for watching this show of the Grow Your B2B SaaS podcast. You made it till the end, so I think we can assume you like this content. If you did, give us a thumbs up, subscribe to the channel. If you like this content, feel free to reach out if you want to sponsor the show. If you have a specific guest in mind, if you have a specific topic you want us to cover, reach out to me on LinkedIn. More than happy to take a look at it. If you want to know more about Reddit, feel free to reach out as well. But for now, have a great day and good luck growing your B2B SaaS.