Andy Karuza: Good morning everybody, and good evening over there in London. Today I have Tony here from Victor, which is victor.com. Without further ado, we would love if you could introduce yourself and what you do.

Antoni Olendzki: Yeah, actually it is no longer getvictor.com; we bought the domain last week, so it is victor.com. We also announced our fundraising today: a $75 million Series A led by Accel and a bunch of really helpful and amazing angels—people like Lenny from Lenny's Podcast, the founders of Slack (we got Cal and Stewart), the founders of Deel, Framer, etc. So yeah, we are very happy and very fortunate. It is victor.com, the AI coworker. I can tell you more about what we do and why we built Victor.

Andy Karuza: Yeah, please do. That is exciting! There are always exciting updates from you guys every day. It seems like things are moving fast over there.

Antoni Olendzki: Yeah. So Victor is an AI coworker you hire; it is not software that you configure. You install Victor in Slack or Microsoft Teams and you onboard Victor as if you would onboard a new teammate. You give it access to your Gmail, your CRM, your Stripe, Linear—whatever you use—and then it runs your operations end-to-end. It can do inbox triage, customer research, pipeline updates, weekly reports, and manage whole Slack channels. Essentially, we are bringing the promise of AI to normies. Victor is plug-and-play. If you have heard about OpenClaw or any of these personal agents, we built Victor as the first multiplayer agent that lives where you work, which is Slack or Microsoft Teams.

Andy Karuza: Yeah, it is very exciting. Today we are talking about potentially getting business owners to a 10-hour work week. Before, it was just like a book you would sell people to try to get them into entrepreneurship, but now it is probably a reality with tools like Victor. I guess what does this actually mean, having a tool like Victor and the ability to automate a lot of tasks almost as if you have a digital employee?

Antoni Olendzki: Honestly, I would say 10 is the headline, not the number I defend. What is actually happening is you stop showing up to the mechanical layer of business, like weekly reports, meeting prep, CRM hygiene, or follow-ups. Those hours do not disappear; they get reassigned to your coworker, Victor. Then you can do the judgment-heavy stuff, like pricing, hiring, positioning, and the three hard customer calls of the week. It is kind of the same working day, but a completely different mix. That is why Victor is the best—the best hires do not need to be told what to do. You just add Victor, and once you start onboarding Victor, Victor starts picking up stuff for you because Victor is proactive. Victor spins up its own computer for your workspace and starts learning your workflows, the whole context of your workspace, and who is who in the company. So, not a 10-hour work week exactly, but all the annoying and repetitive stuff disappears. Over time, you might sometimes wonder, "Am I the intern, or is the AI the intern?" because what it can do is becoming so impressive.

Andy Karuza: Yep, yeah, we have noticed that ourselves too. We have used Victor a lot and we love how it works. Victor is very proactive, so he actually does reach out. He is kind of listening for opportunities for him to step in and help. Nice. So I guess, what is the biggest misconception founders have about automation and AI agents today? I feel like a lot of small business owners still have not jumped on the bandwagon yet.

Antoni Olendzki: I would say it is the "set it up once and walk away" mindset, just because most founders or people still approach agents like they would approach Zapier, but with more IQ. You wire up the triggers and then you are done. Real adoption looks like hiring an operator—like a junior operator—where month one is a bit messy but you are correcting outputs, teaching the edge cases, and building trust. Then, month three is the compounding. Really think about how to give as much context as possible. Over time, teams who treat agents as an employee win; teams who treat agents as a feature will churn from Victor in two weeks.

Andy Karuza: So yeah, I guess the goal is really not to—well, the goal is to work less, right? Does this allow you to operate at a completely different level? Where do founders and operators waste most of their time that can be realistically eliminated by an agentic tool?

Antoni Olendzki: Let us say the top three spots are status reporting loops, like the Monday state of the business where you are pulling all the numbers from Stripe and building a dashboard. That takes a human three hours to pull all the CSVs together and make it look nice. That would be one. Inbox and pipeline hygiene is another—triaging 80 emails and updating 40 HubSpot records by hand. Meeting prep is the third. Anything that has some sort of logic that you could teach and just explain in words to an AI is a waste of time to do manually. Once you have done it yourself three times, I would say explain it, write it down, and give it to an agent. The agent will most likely do it. You check it once or twice, but do not waste time doing repetitive stuff.

Andy Karuza: Yeah. Are there any unique edge cases that you have seen people use Victor for? You know, maybe not like some of the headline use cases like you just mentioned, but what type of edge cases have you seen that were pretty interesting?

Antoni Olendzki: Yeah, so one interesting one: Victor takes care of all the parking tickets for one of our users.

Andy Karuza: I could always use that!

Antoni Olendzki: Yeah. The cool thing about Victor is there are no limits. As I said, if you can describe it in words and there is a tool that does part of the job, you just connect Victor and Victor is going to do it for you. It is that simple. We have a plumbing company, and Victor does the whole commission reconciliation. There is also a company—it is wild—a trash collection company in California where Victor is connected to the tracking system on all the trucks. Victor navigates the whole fleet and tells the drivers where to go and what to do.

Andy Karuza: Geez. So I guess, why do most teams fail to automate? What are they getting wrong when they try to implement some of these agentic tools out there, whereas maybe Victor might be a little bit easier for them to integrate?

Antoni Olendzki: I would say they skip onboarding. They plug in the agent the same way they would plug in a SaaS tool—account created and done—and then they think they can just play around with it. No, an agent without context is like a new hire with no desk and no SOPs (Standard Operating Procedures). That is one failure mode. The second failure mode is they optimize for cost instead of quality in month one. I would say the smart model that gets it right 95% of the time is 10 times more valuable than the cheap one that gets it right 70% of the time. Wrong outputs destroy trust faster than the price tag, so really do not optimize for cost instead of quality. Victor is available on most of the large language models, but it still works best if you go for Claude 3.5 Opus.

Andy Karuza: Yep, and I guess what kind of integrations do you guys have today? You mentioned Microsoft Teams and Slack. Do you plan any other types of integrations in the future? You have 3,000 other integrations that work with Victor too, right? But what are the primary working channels that Victor is going to be working on in the next year or two?

Antoni Olendzki: Yeah, so Victor is going to stay in Slack and in Teams for the near future, because that is where most teams live when it comes to work. Then, Victor integrates with over 3,000 tools. If Victor is not integrated with a given tool, Victor can build its own integration via MCP (Model Context Protocol). There are requests for Victor to be available on WhatsApp as well. It is big, especially in Asia, where people run their businesses on WhatsApp. We are going to roll it out at some point, but for now, it is very much focused on SMB and enterprise, which is Slack and Microsoft Teams.

Andy Karuza: Where are you seeing the most value? Are enterprise companies signing up more, or do you find a lot of SMBs are coming on board?

Antoni Olendzki: Right now, it is pure product-led growth on our side. We recently had a salesperson start to just reach out to enterprise customers, but I would say it is mostly SMBs that need an additional pair of hands. They need help with the operations stuff that no one could handle. The interesting part is it is not just the people in San Francisco or New York; I mean, they are also using us, but it is businesses in Iowa or Texas. There is a landscaping business in Iowa with 40 employees, and there is this lady, Christa, who runs it. She just needed an additional pair of hands, saw one of our Instagram ads, and started using Victor. Now she is going crazy with it. She has never touched the terminal and she does not even know what a CLI (Command Line Interface) is; she just needed something to talk to and get stuff done for her. That is why the idea of a coworker that lives in Slack is so amazing, because we do not see any limit. We treat Victor—our mission is to build the AI for everyone else. Everyone else who cannot build their own agent or set up a Mac Mini with OpenClaw. So, yeah.

Andy Karuza: Yeah, it is almost a no-brainer for any startup these days too, just to be able to work with these agentic tools to maybe cut some of the costs down that it might require from hiring employees, to really get their product to market and start scaling right away. Can you walk us through a real example of an AI agent replacing hours of manual work? You mentioned the trash company, but are there any other use cases that you want to cover?

Antoni Olendzki: Yeah, so for a single player, definitely a big one. Victor has permissions, so you can add Victor to different channels, but you can also talk with Victor in your DMs. You can connect Victor to your inbox and you can tell Victor if it is only you who has access to it or the whole team. If it is your personal stuff, Victor can triage your emails and reply to them. Victor can also manage all your tickets, whether in Linear or Jira. You can connect Victor to your GitHub so Victor can build workflows around the admin-heavy stuff in your day-to-day job. You can also use Victor for a more multiplayer setup where Victor orchestrates and becomes the product manager of an engineering team. Victor is told literally, "Hey, this is the product sprint for the next two weeks." Victor fields tickets to engineers, reviews the PRs (Pull Requests) for each engineer, pings people when he finds issues, and gives a status report to leadership. The idea with Victor is running a company of 5 like you would run a company of 300 people, basically giving the power for teams to just run lean.

Andy Karuza: What do you think that is going to do for the job market in the future? It sounds like a lot of teams are just going to be able to hire all these agents and get a lot more work done than what their hiring needs would require.

Antoni Olendzki: I think it just creates more opportunity for people to do more. I have never met an engineer, or frankly anyone, who is completely out of ideas. It is always a question of time versus capacity. I am actually at a conference right now—it is an Anthropic conference—and there are engineers from all sorts of companies like Personio, Spotify, and all the big banks. No one is firing right now; they actually need more people because you can just do way more with AI. When Victor takes stuff off your plate like creating reports, you do not need a data science person to run dashboards for you anymore, so that person could probably be used for conversion work on your website instead.

Andy Karuza: Yeah. Any fundamental shifts you might see in the market going forward? Maybe the way people approach business or how they approach marketing, things of that nature?

Antoni Olendzki: People will probably stop—well, that is a good question, how people would... I don't know, I am literally amused every day by the things that are being posted. We have a Victor community with over 500 customers there, and we have a show-and-tell channel where people post examples of what they are doing with Victor and how their businesses are changing. I am just astounded every day, so I literally have no idea how this is going to change business. But I am more on the optimistic side; this is going to impact us in a good way and it is going to shift the way the job market looks. I do not know if you have heard about this concept from the book by the sociologist called Bullshit Jobs. You probably know someone who has been just sitting at a desk job and pretending they are typing emails all day; that is probably going to be gone because you cannot pretend that you are doing something that is totally not adding value when an agent could do that. The question is, once this person is out of this misery, what are they going to do? But I definitely think there is an element of adapt or, as you see on Twitter, "ngmi" (not going to make it). You should probably get on it, be interested, and come from a place of hope versus fear when it comes to using these tools. Learn them and try to coexist versus blocking yourself completely.

Andy Karuza: Yeah, I am generally optimistic as well. Every technological shift like this always has an initial fear factor in the early stages, and then everybody figures it out after that. Thankfully, a lot of those corporate jobs where people are not actually doing any work will actually get filtered out, and they will bring in people that actually want to use the AI tools and get the work done. Yeah. So how do AI agents handle messy, unstructured tasks compared to clean, rules-based workflows? I feel like one of the biggest problems with people is that they are terrible at communicating. The best people at working with AI always are great communicators; they know how to ask for what they want. How would you approach that if you are working with an AI agent and you have unstructured, messy data? Is there a better way to work with it to allow it to complete the task efficiently?

Antoni Olendzki: Number one: break it down. Do not try to one-shot it. If there are pieces of the unstructured data or something that you want the agent to complete, you can just tell Victor, "Hey, let us enter plan mode." First, plan out exactly what you want the agent to do, explain it clearly, and tell it to ask follow-up questions. Once you are aligned with the agent on what needs to happen, then tell it, "Okay, now go." The main complaint from people who are not very careful or thoughtful about how they deal with agents is, "Oh, I just got some slop and this doesn't work." Of course, if you had a new person come in—say someone from a different department at your company comes over, like someone from legal when I am working in finance—and I say, "Hey, I need to build out a dashboard to forecast Q4 revenue, go," of course they will be like, "What? Give me context." The phrase "context engineering" is so important. The cool thing about agents like Victor, and agents in general as they become better at this, is they are good at interpreting natural language. In the very beginning, you had to be very careful, break it down into sub-points, and be very deliberate about what you want. Now, you can talk to it just as if you would voice note a friend. The more you talk and the more details you give—it does not have to be structured—the agent is going to know what to do.

Andy Karuza: Yep, and in many cases, it will also ask you for feedback and iterate on your question. Victor is very collaborative in that way.

Antoni Olendzki: Yep, exactly.

Andy Karuza: That is good. So I guess, what does a modern, AI-first operations stack look like for a startup today? Obviously, a lot of it could start with having an automated employee like Victor. What other recommendations would you have there for a company?

Antoni Olendzki: I would say keep it thin. Not 14 different point tools, but just one agent layer that sits across everything. You keep your systems of record there—Slack, Gmail, HubSpot, Linear, whatever—and then the agent lives as a coworker on top. The shape that is winning for us is Slack as the UI and the agent handling work across integrations, with a couple of dashboards for humans to watch. The AI-first stack is just less stack overall.

Andy Karuza: Yep, and it is really about the integrations like you said, which is why you guys focus on integrating directly with Slack and Teams where people already do their work today, and then you have all those extra integrations that people can work with outside of that so you have that single layer, right?

Antoni Olendzki: Yeah, and I mean, "rip and replace" is a fantasy because of your team. The move is an agent that just joins these tools, not another app that they have to open. Think of it as a new hire that uses your existing stack; it is not a consultant who wants to migrate you. Victor won't tell you, "Hey, suddenly you need to stop paying for all of this stuff and I'm going to do everything for you." No. If the question is how to adopt this, it is not about changing everything first. You need to see it as an actual coworker. That is why this whole narrative of "are we going to be replaced by AI" is somewhat true to an extent, because you are adding AI on top of your tool stack to orchestrate and do stuff for you.

Andy Karuza: Yep. Do you see any certain types of software tools that might go away or lose relevance with some of the capabilities that agentic tools have?

Antoni Olendzki: I do not want to be quoted on this, but I think that is the question of the day in the whole industry. Bigger picture, we see a lot of companies building certain types of software in-house because it is fairly easy to do now. Of course, sometimes you want to stay with a vendor, but there are companies that are building their own internal Slack—though the question there is, what is the point? But if there is a business case and money to be saved, we will see way more businesses building their own internal CRM, for example. For the most sophisticated stuff, it is also a question of opportunity cost as a business. Do you want to dedicate your engineering resources and spend money on tokens to build out and use your own CRM, or do you want to just go with a vendor who has a support team that can do most of the stuff that you need them to do and just outsource it?

Andy Karuza: Right. I always feel like some of those calendar tools maybe could go away. With AI tools, you can easily connect with your calendar and then figure out good times to book and everything, so some of those simple SaaS tools might be able to be replaced by the capabilities of agentic tools. Is there any risk with over-automation? I could see a startup founder easily being like, "Victor is the best thing in the world," and they just use it for everything. Is there any sort of risk at all with over-automation, especially for early-stage companies?

Antoni Olendzki: Yeah, definitely. I would say silent errors at scale. One bad call is fine, but 400 bad calls without noticing is very messy. The second one, actually more importantly, is skill atrophy on the team. If nobody on your team has done a customer research call in six months because the agent does it, nobody has the context to know when the agent is going wrong. It is easy to get lost in it, especially with Victor, because Victor can do so much for you and your team. He can jump into threads and reply, and sometimes you might just say, "Okay, Victor's got it, I don't need to double-check or read through this." But you probably should. Over-automation is basically outsourcing without a QA layer. The fix is the same as with humans: clear guardrails, regular review, and keeping humans in the loop on high-stakes calls.

Andy Karuza: Yep, I like that. How do you ensure accuracy and reliability when AI is making operational decisions?

Antoni Olendzki: Three things. First, you only automate where you can spot-check. If you can't tell whether the output is right, you shouldn't be running it unattended. Two, you keep humans in the approval loop for anything that leaves the building, like emails, invoices, and customer-facing posts—that is for sure. And three, you measure. You track output quality the same way you track a human hire: accuracy rate, review time, and trajectory over weeks. The agents that improve are the ones with a feedback loop, and the ones that don't, you probably don't use.

Andy Karuza: So yeah, I guess if a founder fully embraces AI agents over the next 12 months, what does their role look like going forward? What are they no longer doing on the day-to-day? What are some of the first tasks that you are seeing your customers cut out of their day-to-day roles?

Antoni Olendzki: You stop owning the operations layer completely. No more "I'll update the CRM tonight," no more running your own weekly reports, no more drafting your own follow-ups. But you would still own, as I said, hiring, customer conversations that matter, roadmap calls, sales, and relationships. You become the director, not the operator. For most founders, that is the role they always wanted but they couldn't afford. You stop doing the busy work, you stop being overwhelmed and sitting up all night trying to make sure that something is reconciled and connected properly.

Andy Karuza: Yeah, that can be outsourced. That and spreadsheet analysis. It is everyone's dream to get rid of all the operational tasks; people just want to get back to building, exactly. So I guess, what is next for Victor? Obviously, a lot of exciting updates that I just learned about today myself, too. You guys are moving fast over there, seems like some new news is happening every day. What do you guys have coming down the pipeline?

Antoni Olendzki: Two directions. Number one: because we are building the AI for everyone else, we are building deeper into verticals. We launched Victor for media buying, so Victor can run your Google and Meta ads. Victor has this predefined skillset for that, built by the head of growth at Whisper Flow. Next, we are launching Victor for customer success, Victor for real estate, and Victor for construction companies. We are giving specialized skills to Victor so he becomes truly plug-and-play with zero onboarding on your side. Victor is added to your workspace, you choose the skills that apply to your business, and Victor becomes the operator for that given vertical.

Second, we also want Victor to be built not only by us. For lack of a better word, we are building a Victor army. The vision is having 16-year-olds who can run to their corner shop or their local, boring business that could use an agent like Victor, onboard them to Victor, and make money off of that. We want everyone to be a Victor ambassador. Hopefully every business in the world—but definitely in America—will have no excuse to not have an AI coworker like Victor on board doing the stuff that, frankly, humans shouldn't be wasting time doing.

Andy Karuza: Yeah, that is... Oh, go ahead, sorry.

Antoni Olendzki: I mean, then there is a lot of cool product stuff that we are rolling out, like fine-tuning and making Victor faster, cheaper, and more efficient. But yeah, the vertical launches make Victor plug-and-play as a more domain-shaped agent for your choice. You can also create your own skills; we are launching a skills marketplace where if you want to make money with Victor and share your knowledge and skillset, you can submit a skill and start selling it to people. If you are a specialist in landscaping—we have someone actually who is already selling Victor skills for landscaping businesses—you can buy a bundle. So yeah, that is also something interesting on the roadmap.

Andy Karuza: It is really exciting, you guys are working on a lot of great things over there. People can pick up a Victor today now at victor.com, right?

Antoni Olendzki: Yeah, victor.com. $100 in credits is included, it's free to start, and no credit card is required. You just add Victor to your Slack and literally talk to Victor like you would talk to a coworker. You add Victor and you can say, "Hey Victor, I'm doing XYZ in my day-to-day, what can you help me with?" Victor is going to tell you step-by-step what he can help with and what he can't. You can also just integrate Victor with all the tools that you are using and say, "Hey Victor, look at my setup, am I missing something? Am I doing something wrong?" We have a customer testimonial story where this guy was setting up a new real estate development. He had this whole plan, along with the legal and financial paperwork, and the first thing he did was dump it onto Victor and say, "Victor, is there something off? Could I be doing something better?" Victor found $4 million in savings for him.

Don't be scared, that's the thing. People are kind of afraid like, "Oh, what can I do with this?" It is literally like talking to Victor like you would talk to a human being. Give as much context, like we discussed before, and let Victor do his thing. We built Victor in a way that we also are very conscious of the approvals part we discussed earlier. If something is a bit more—not necessarily irreversible, but dangerous or high-stakes—Victor is going to ask you for approval. Once you say, "Victor, do this," Victor is not going to just go and one-shot it. Victor has guardrails to make sure. If you want to make a change in your CRM—if you want Victor to suddenly change all the emails of your customers in your CRM for whatever reason—Victor won't just say, "Hey, I'm cool with it, I'm going to do it." It is going to ask, "Are you sure? Do you want me to do this?" And once you give approval, then Victor is going to go ahead.

Interviewer: Well, Victor sounds like the perfect employee and it sounds like the first hire that every company should have. Tony, thank you so much for coming on today. Once again, that is victor.com, you get $100 in credits right away so no money out of your pocket to give it a go. Definitely a super valuable tool; we use it at NachoNacho and I have recommended it to a couple of other companies too that love using it as well. Definitely check it out today at victor.com, and I will see you soon, Tony.

Antoni Olendzki: See you, thank you. Thank you.