Most AI Failures Don't Start With the Code. They Start With the Contract w/ Rory O'Keeffe | S2E15

Full Episode Transcript

This is the full recorded conversation from The Human Protocol Podcast, published by The Human Protocol Group. © 2026 The Human Protocol Group. Please link to and attribute the original episode when referencing this conversation.

THE HUMAN PROTOCOL (00:02.188)
Most AI failures don't start with the code. They start at the time that we're putting the contracts together. If you can't explain where your data goes, you're not deploying AI. You're really outsourcing risk. Welcome to the Human Protocol, the podcast about staying human while building the future. I am your host, Michael Solomon. Today's episode is really a founder survival type of episode because AI deployment is not just a

product decision. It is a legal, financial, and reputational decision as well. And most teams signs contracts like they're buying software. And what they're really buying, it's a risk profile. My guest today is Rory O'Keefe. And we're going to turn this into a practical playbook. What founders miss, what to negotiate, what questions to ask, and how to work with your lawyer without slowing your momentum.

Roaring, thank you for being with us today.

Rory O'Keeffe (01:03.81)
Thank you. It's a pleasure to be here. Thank you.

THE HUMAN PROTOCOL (01:06.44)
Awesome. I'm looking forward to this conversation. And before we get started, I would like to ask you if you can give us a little bit of information, who Rory is, and for my audience to get to know you a little bit better.

Rory O'Keeffe (01:20.034)
Sure. Rory O'Keefe is my name. I'm a founder and CEO of a legal services company called RMOK Legal and also host of Beyond the Fine Print. It's a tech law podcast. My background is very much as a lawyer. I've been a lawyer for longer than I want to admit, but I am Irish, born and raised and very proud to be having born originally in the

Garden of Ireland and the home city of Dublin, it's the capital. But I moved over in 2012 to London, just around the London Olympics. And then I decided it was a good time to bring over the rest, the family and the dog. So we moved over in 2012. So it was quite a shift going from one jurisdiction to another. But throughout all of that period, I've been...

And before I've been a lawyer in criminal law all the way through to arbitration through projects, energy construction, which is your wind farms, your oil and gas, liquefied natural gas, building bridges and gas power plants and gas fields as well for mining off the Irish coast. And then we went straight into the area of technology. And that's where I've sat for a long, time now.

So it's been a really enjoyable experience and setting up my own company was a whirlwind. So for all the founders out there, I now get it. I used to think I knew it, but when you don't live it, you don't really truly understand and appreciate the pressures that my clients are going through. So I appreciate it. And to anybody who I went around saying, I know what you're going through. I apologize. I didn't.

THE HUMAN PROTOCOL (03:08.654)
Awesome. Thank you for going over all that and give us a little bit of a glimpse of who you are. When we were talking about, what type of conversations can we have in the podcast? This one really stood out, right? On what are the founders finding out there, especially in this one with AI and how important it is for companies to pay attention to contracts.

and really all the legal ramifications that come with the implementation of it. And I would like to start with the question, Rory, for you, that when you hear about AI deployment, what do you think founders misunderstand immediately about it?

Rory O'Keeffe (03:58.474)
It's not to be judgmental of anybody who thinks this, but sometimes when we people talk about AI, it's not AI. It's automation. So just automating a workflow or it's robotics, which was the mantra a few years back, pre pandemic robotics was the discussion, which is just, you're creating a script following the key strokes of the individual, filling out a template or filling out just, you know.

boxes on a page which can be easily done in say a mortgage application or you know an immigration form or something that just doesn't require a lot of thinking it's just factual entry data entry pieces and then when we've got to AI because AI has been around for 75 years by the way so we all think it's brand new because of ChatGBT OpenAI launching just over three years ago we're now in a situation where actually no

that element no didn't exist now it does and everybody loves it and everybody wants to use it and everybody's rushing to use it and that's the concern I always have not just as a lawyer but also as a as business owner because you can use something doesn't mean you should you need to understand what the risks are and once your eyes are open to all of those things and it still works within your price point well then yeah go ahead use it

But I think from a starting perspective, I think for many founders or business owners or anybody who's heading teams, they're under pressure to deliver by using these. Of living crisis, we know that. The push for increased efficiencies, your competitors are moving at pace, so you need to move at pace as well. How do I do it? Well, easy way to cut head count after the rush for outsourcing for the last

15, 20 years, um, we've had the now rush to utilize in these AI tools. Before you rush, you really do need to just pause and check to see, this the right tool? Like any other technology you would have utilized and thought through just because the price might very low doesn't mean, I'll give it a go and sure. That's the whole point of being a founder. Take a risk. And if it fails, fails. The problem with this is that if it fails, could cripple your company. So.

Rory O'Keeffe (06:20.62)
just eyes wide open. That's the first thing.

THE HUMAN PROTOCOL (06:24.366)
It's a different type of failure, right? And especially on this one. Especially depending on the scope of this implementation, if you are rolling this out in a big enterprise and it's impacting all your business, yes, as you mentioned, it's like the failure, if it hits, it's going to be a different type of failure that it can really ruin your company. And Rory, what is the most common contract mistake?

you see organizations and people doing out there when buying or AI tools.

Rory O'Keeffe (06:59.522)
So I think there's a number of things. Different people may be hit by different things depending on how the tool is being sold. Some walk in onto the click through terms. So you know when you've got mobile phone devices, cell phone devices, you've got, like for example, Apple does this, not calling at Apple to say that anything wrong or not. Who would I? They're too big. They can take me to court quickly. But no, it's, have good friends in there. So I still need them to be good friends.

But if you take it as an example of those click through terms, you open up the device and whether it's Samsung, Apple or whatever, and they'll say, we've just delivered a new security update or a new patch and your upgrade. And then you have to say, and we've updated our terms and conditions. You must accept them or not. And if you don't, you can't use the device. If you do, you can use the device. It's a big question mark whether that's a legitimate piece for consumers, but leave that aside for another podcast. But the idea for this is...

That's what a lot of the AI tools are driven on, that you won't negotiate their contract terms because it's a long page, maybe put up on their public website. You look through it and say, yeah, it looks like a contract. There's nothing jumping out that's horrific, but that's fine. The things that people miss and sometimes in those contracts is data use. So for AI in the context of machine learning,

it needs the data that's flowing through the tool in order to, for that algorithm to improve and grow. So they need your data, your data equals money, for example, your data has a value to them. And they usually have a clause in there to say, we can use your data, but don't worry, we'll anonymize the data before we start using it. And you go, that's great. means all. Yeah, so if you don't want,

THE HUMAN PROTOCOL (08:44.972)
We'll anonymize.

Rory O'Keeffe (08:50.722)
personal data, PII, if you've got any of those things that even your name, your age, all those things, employees will be using the data, maybe customers might be inputting data, the algorithm will say these people, these demographics, these sorts of individuals will be utilizing this tool, you throw the information in and hey presto, the tool updates and then says, okay, I'll learn, I'll tweak.

I'll revise information. Sounds fine. Except that's personal data, which gets caught in different jurisdictions. In Europe, it's the GDPO, in the UK, there's an equivalent. You've got privacy in, they're updating it in Indiana, Kentucky, Rhode Island. There's the big one in California, and then beyond, right? So you've got multiple jurisdictions having different rules around personal data. In order to not having those regulations hitting you, you anonymize.

Because if it's not personal, it doesn't get caught by all those regulations. Easy. Problem is, what does anonymize means? It means you never can track the information back to that individual by other means. So if you had like a raw image of an individual which you did for anti-media laundering or know your customer type checks, know, as we do when we're setting up new suppliers or vendors.

we will check them all out. We'll say, OK, you hold the money to run those checks, then you destroy it. That means you don't have the original image, which means all you're taking is the output from that analysis. And that analysis might be, this is what a passport looks like in Mexico, and this is another passport that looks like in Belgium. But we've stripped out all the personal information from it, so nobody could ever reverse engineer that information. So it's good. That means anonymized.

THE HUMAN PROTOCOL (10:42.286)
to you.

Rory O'Keeffe (10:44.854)
Now, if you can ever, if there's any way that you can't get to that point, we call it pseudo-anonymized, which means it's still caught by all those regulations. That's what everybody's trying to get to when it comes to anonymize. It's kind of the holy grail for feeding your algorithm with the right data for it to continue to improve and give good service.

THE HUMAN PROTOCOL (11:09.034)
Interesting and...

Rory O'Keeffe (11:10.178)
I know there was a bigger question there, which I've completely forgotten. I'm so sorry.

THE HUMAN PROTOCOL (11:16.046)
And tying it back to the contract, how we can make sure that that's what we need to have there. It's being covered, right?

Rory O'Keeffe (11:20.514)
You

Rory O'Keeffe (11:26.53)
Yeah, so the question is there will always be a clause in there. It could be just a few lines, but there'll always be something there that talks about, we will anonymize it and then we can use it. Or if you give a feedback, we will be able to use this. Or if there's any outputs, we'll have a license to continue to use it even after the contract, depending on what the tool is attempting to do. They might not want it, by the way, but they might because the nature of the algorithm and what they classify as machine learning might demand that kind of data.

THE HUMAN PROTOCOL (11:46.23)
Mm. Mm.

Rory O'Keeffe (11:55.68)
The only way the AI tools, especially the generative AI models work, large language models work is by feeding us a lot of data. And good data is current. So if they've got 6,000 clients on the books and there's 6,000 data sources, then that's very valuable to this provider who's trying to grow and build out their technical roadmap.

THE HUMAN PROTOCOL (12:04.824)
data.

Rory O'Keeffe (12:22.4)
to improve that technology. The only way to do it is to get good data sets. And if you can't get it from your clients or customers, you're going to have to source it elsewhere, either internally yourself, or you're going to have to buy it in and again, value over the data. So that's the first thing I would look for is that piece. There's other pieces which I would look for is around auditability. So if the algorithm goes wrong, how do they explain it? How do they prove it went wrong at what point?

Because there are examples of, I have a story with Gentic AI agents rather, but on this point particularly, there was this story a while ago, big platform, big company, big tech company in the States. They wanted to do the good thing, good corporate responsibility about giving local jobs, because they're in a building, they want to give back to the local community.

So they set up the algorithm to just look at, you know, X mile radius from the building, X kilometer radius, whatever we're talking about, depends which country you're in. And so we'll say mile or kilometer radius of the building, so that any resumes coming through, we will then say, we'll accept those in. And the algorithm just basically turns through that, and that's where they cut them off. You give me your address, physical address, question mark, should you put it on your CV or resume? That's up to you.

THE HUMAN PROTOCOL (13:40.397)
Mmm.

THE HUMAN PROTOCOL (13:44.984)
Mm-hmm.

Rory O'Keeffe (13:51.404)
But the question then comes back, okay, so they did all these interviews. Stellar, great talent, great interviews. The interviewers went back to the HR team and said, quick question, why is everybody coming through of the same background and demographic? And also, they're all white. Are we missing something here? Or is our company only attracting these people? So it might be nothing wrong.

THE HUMAN PROTOCOL (14:20.706)
Like bias, right? Yeah.

Rory O'Keeffe (14:21.154)
just because, you know, it could be just where you were, you know, so if you're, you know, if you're in the middle of Zimbabwe, you know, it's unlikely you'll find, you know, a high proportion of white people, you know, applying. Okay, that's just the way it is. Okay. And hopefully, they'll all change. But the, it's in the reverse in this instance. And they looked through it and they realized what they had done was by restricting the radius around the building.

They restrict because the building's in quite an affluent area. You can read into all this, all you want. I'm not going to go down this path, but as a white middle class male, super privileged and all that, I will let you all dream what you want to dream about this and read into what you want. But the result was, even despite their good efforts, they actually came out with the wrong result. And it was only the algorithm that was fed the information and told her to think.

that led to that result. So again, it comes back to the human, just like in your podcast title, it's the human protocol. The human needs to come back in to say, we fed it the wrong data. And once they expanded the radius and included other regions across the city, around the building, they suddenly got, again, stellar candidates, but from a wide range of diverse backgrounds, ethnicities, heritage, et cetera.

And they got a great mix of candidates and then they chose the best person for the job, whatever they may be. But that was just kind of the why with AI, the problem sometimes you face is that because it's fed so much information, it's sometimes hard to track what was the point in time when it started thinking a different way, which comes to yes, when it started going wrong, it may not be easy to pinpoint.

THE HUMAN PROTOCOL (16:08.792)
Going wrong, yeah.

Rory O'Keeffe (16:14.87)
So question around what you talk about is responsible AI, then there's audible AI, explainable AI, I think is where we've all landed in the mantra and all the conferences. But if they can show how they can explain if it goes wrong, where it goes wrong, and then reverse it back to before it started deviating, that's something in the contracts to check for. Fun.

THE HUMAN PROTOCOL (16:40.106)
And yes, and this is only one use case. Imagine how many we have out there that we haven't heard about. But I'm happy you brought that up, because this is a clear example of how important it is to continue to keep humans in the loop, as I continue to tell everyone. And this is just one of the examples where if we let AI to drive results by itself without us checking them, it's going to

Rory O'Keeffe (16:43.083)
you

Yes.

THE HUMAN PROTOCOL (17:08.983)
it may continue to give us the outputs that were not intended to begin with.

Rory O'Keeffe (17:13.826)
Yeah, but let me give you, there was one thing I wanted to say, because I dangled the idea, but I should follow through, because I know your listeners want to thank me. Agents. So this year is all about agent agents, right? They're throwing it out, or the big tech want to, because they haven't seen the big returns that they thought they would. Okay, that's my view. I'll leave it to all the economists to determine otherwise. But what they have, had a supplier last week talk to me, who's now a client, thankfully. We sat down and explained to me how,

THE HUMAN PROTOCOL (17:24.408)
Mm-hmm.

Rory O'Keeffe (17:43.51)
they had an agent which was set up basically to act as like a scheduling tool. Okay, not pointing fingers at anybody, not gonna name names. Scheduling tool was in, said go and book me a hotel. I need a room for this date, for this location, because I got this big meeting coming up. The tool went off and basically was, it's like as if it's moving around your screen automatically, you know, like when an IT comes in remotely to start fixing whatever you did wrong.

doesn't get fixed by switching it off and switching it back on. then in they come. Problem came around where the tool went off and said, okay, so I need to go through the website and I fill in the information. Website was down, can't do it. Okay, next step is I need to call them. Okay, but they don't have a voice. So they went off and the tool went off and found a, what is a recommended text to voice.

voice cloning tool. They then were able to voice clone the person that they were scheduling this for by getting information from the various conferences, podcasts and things like that, like our voices were suddenly being used. They created the voice, they created the text, then they had to do this by setting up a subscription. So they must have found credit card information in order to do it. They did all this, paid the subscription, created the text,

Christ voice.

THE HUMAN PROTOCOL (19:11.96)
found the credit card information as well.

Rory O'Keeffe (19:14.924)
This is the questions, when you put parameters in place with your own security on your own machines, be warned, What freedom you're going to give to your AR tools. They got the information, then they...

THE HUMAN PROTOCOL (19:28.236)
that's right, because it logged into the person's computer. Okay, yes.

Rory O'Keeffe (19:31.874)
Exactly. All right, so you're with me now. There's many alarm bells ringing from a legal side, okay, but I'm hoping you've got your own practical alarm bells around your own security of information. But the effect was they actually called the hotel and they had the conversation with the booking agent on that side. They completed all the transaction. They had the booking in place.

Rory O'Keeffe (20:03.566)
So there you go.

THE HUMAN PROTOCOL (20:05.558)
Okay, I'll meet you for a moment, but you're back.

Rory O'Keeffe (20:10.334)
One second, one second.

THE HUMAN PROTOCOL (20:15.618)
Yes, you're back.

THE HUMAN PROTOCOL (20:20.3)
It was your screen went black but you're back.

Rory O'Keeffe (20:22.98)
Sorry. Hopefully you can still hear me and everything's great. I get to the good bit and then some cut says, going, no, don't do that to me.

THE HUMAN PROTOCOL (20:27.648)
I... Yes, I can hear you now.

THE HUMAN PROTOCOL (20:34.742)
You just, you just, you just leave me wanting to know the end of the story. And I was like, no, this is not happening right now.

Rory O'Keeffe (20:46.008)
This is cliffhanger that I've just done. Apologies. The result of it was that the outcome was exactly as you asked the agent to do. How they went about to create the result. That was not what you were setting out to do. So when you're thinking through what you're making these amazing prompts and making these amazing agents to streamline how you're doing your services and everything's running so smoothly because you're asleep and

THE HUMAN PROTOCOL (21:00.064)
It's a difference.

Rory O'Keeffe (21:14.917)
The agents are working for you on your behalf and your 24, seven, three and 65 degree, degree, sorry, day service. And yet they're coming up and doing stuff and you're wondering, why am I paying for this and this and this? And what was that? And what? Just as I say, eyes wide open before you start rushing ahead.

THE HUMAN PROTOCOL (21:37.75)
And it's back to the power of prompting. What is it that you want to achieve? And the how, it will remain as important as ever. Because you will dictate how you want things to be achieved. And it's not only about the output. Now we also need to worry about the how it's going to get done if we want to avoid issues.

Rory O'Keeffe (22:00.184)
Yeah, yeah.

THE HUMAN PROTOCOL (22:02.498)
This is just getting more interesting by the day, Rory.

Rory O'Keeffe (22:05.828)
And I am a big fan of AI. Okay. I use AI as a tool. Okay. It doesn't replace me. A lot of people would love to have AI lawyers replace the real lawyers. Okay. But it won't make all those movies very interesting anymore. Those courtroom dramas very interesting if you just have AI talking on our behalf. Yeah, it's not very entertaining anymore. Okay. So just think of suits and all those things.

THE HUMAN PROTOCOL (22:13.164)
Right. Right.

THE HUMAN PROTOCOL (22:28.046)
It will be boring.

THE HUMAN PROTOCOL (22:34.754)
Yeah.

Rory O'Keeffe (22:34.974)
It's not gonna be as interesting. but the reason why I reaffirm, I always tell the legal profession that AI won't take your job. But a lawyer using AI will. So all the law firms, the legal departments are using AI in some shape or form, be it just for communication purposes, or just for managing their schedules, or just for tracking things.

THE HUMAN PROTOCOL (22:49.976)
Mm-hmm.

Rory O'Keeffe (23:03.524)
or just dealing with frequently asked questions as part of a chat bot or something of a of a solution to try to get things done. And I've seen it used really effectively when it comes to people looking for access to justice. So I'm not adverse to the argument lawyers are too expensive. And because it's so expensive, means people who don't have the money to pay for a lawyer just to send out a letter, they are being denied.

a right or a fundamental right to basically fulfill their own rights, their guesswork. What I've found is that some of the AI tools are quite, they're kind of levellers for, for or democratize the justice system to the point where people who don't have, who have accessibility issues, for example, dyslexia, and you don't like writing and reading, so you can get around it. If you've got visual impairment,

you don't you can actually get through things a lot faster. In fact, actually, I know one friend of mine who's 100 % visually impaired and to the point where when they're using AI to describe what's happening on their computer screens, they said it's far more accurate than a human person and actually more enjoyable. Now he hasn't said that to his wife yet, but there you go. And then if you look even further, for those who don't have certain country where it's not their first language.

you know, I'm from Ireland, and just technically the first language, but we use English an awful lot more. But say you went to Germany, say you were in certain parts of the United States, where Spanish is more or German is more. I married a Brazilian. So Portuguese is a big thing in my head. So having an AI tool, says, this is my issue, I'm a landlord. I just don't know how to phrase it in the right way. And I want to use

bossy English or get my grammar wrong or whatever. I'm a bit concerned. You can use the AI tool to do it for you. They can come up with a good first draft of something that gives so much more freedom to people to take some of the justice or some of their control back and kind of jump over some of the reasons why they may have just rolled over and just not entertained any kind of conversation or argument, if I put it that strongly, with whoever.

THE HUMAN PROTOCOL (25:11.457)
Absolutely.

THE HUMAN PROTOCOL (25:28.598)
Yeah, it's especially on translation and there are so many use cases for sure. But I also see as a great benefits. these wearables like AirPods and you know, those meta glasses, they can they can do live translations. And we also have seen, for example, social media network tools like Instagram. Now you are uploading videos and automatically is translating to other languages. So people on the other side, if they don't speak our language, they can

have that translated to their native language based on the languages that they have available. So the use case and the benefits are for sure appreciated there. But you were talking about, for example, law offices. And I wanted to ask you something there, because I know that's your area of expertise. How are you guys in the law offices industry, in the law office and legal office industry?

How are you guys handling data privacy and all the parts coming to security of having the data secured and making sure that does not going out into the internet? Is that you guys are leveraging AI tools that are reassuring that the data stays in your tenant and is not going out? Is that how you guys are approaching that?

Rory O'Keeffe (26:52.534)
Yeah. Yeah, that's so for law firms, especially because if you think about it, and any customer supply chain, you don't want one point of failure and law firms are potential target. We had a case number years ago called the Panama Papers, where there's a law firm out in Panama, which was hacked and a whole ream of very, very expensive and very damaging documentation were released out onto the dark web.

THE HUMAN PROTOCOL (27:20.354)
Remember those.

Rory O'Keeffe (27:21.892)
And that was just a data breach of itself. And that breach only arose because they weren't updating their systems according to the tech providers requirements. said, you need to update. know, everybody gets these, you do you want to update this evening or tomorrow? And they were just putting it on the long finger. Yeah, we'll get to to it. Get to it. was just too busy. And then what happens? Hackers come in, they chance that they find a way in and they presto that company no longer exists. That law firm is gone.

Um, the, the issue that we have for law firms generally is different clients have different confidentiality rules and we've always had it. So before AI came into play, you had, um, and before even GDPR came in 2018, we had all these banking regulations. We had all these structures in place to protect against it. You had what we called on-prem solutions. So we rarely use the cloud for anything because it wasn't reliable. It wasn't secure enough. Now over time.

THE HUMAN PROTOCOL (28:13.677)
Mm-hmm.

Rory O'Keeffe (28:20.918)
Admittedly, the hyperscalers have improved dramatically. There are security measures and clients have become more accustomed because they themselves have run their own checks. They've done the NIST 2 checks, the ISOs, they've done the various levels of standardizations. International standards have been reached across these and actively managed. Not just I've got my standards and then forget all about it.

THE HUMAN PROTOCOL (28:46.958)
Right.

Rory O'Keeffe (28:47.32)
They have to continuously manage their governance framework, improve their systems. Nothing's ever a hundred percent proofed. Like back when I was starting off, you had this big, cause I was obviously junior and and nobody, sat in the room with the server. Okay. So that big looked like a filing cabinet with lights flashing all over the place. And I had to do the backup tapes, by the way, not systems, right? So I've taped. Yeah. Okay. I don't want to age myself too much.

THE HUMAN PROTOCOL (29:11.342)
I remember those daily backups, nightly backups. boy, you're taking me back to.

Rory O'Keeffe (29:18.6)
And the result of it was you just you have to take it out. It was like, oh, we do it every day. No, I do it. And you put it into the fire safe cabinet just in case. it was this is the way it was done. OK, then you move forward and people have big, you know, bigger companies, bigger law firms would have former sophisticated regimes in place. But ultimately, it's on site. When the cloud came in.

as like data center solutions that took a while before people gravitated, but then you had to have your own instance. So you can segregate the data or isolate the data because that's the demand from the client. That if they pull the plug on the relationship, you need to be able to destroy it in line with data retention and other reasons. You have to hold data for a longer term for litigation, potential litigation purposes. It could be six years tax could be 10, 14 years, whatever the reasons are.

You still need to hold it. So the concept of, you know, this big data document warehouses holding all these big boxes of information was becoming less and less because we were going to digital. So that became an interesting shift. think. I remember putting all those boxes together with all these Bible of documents that you had from a litigation case or from a discovery for.

purchasing a company, but you can't do that now. It's all virtual databases in the main. There's still a bit of physical AI came into play to streamline this. It started in the litigation for e-discovery purposes with an acceptance of an accuracy of say like 80 % and suddenly it's an accuracy of 90%. You still have to have the human coming in, double checking to see is it picking up the right information.

We then brought it in for data subject access requests. We brought it in for case management. Now what they're doing with AI tools, which is very interesting, is they're doing predictive analysis. So if you've got a big law firm and you've got all this information around how these litigations went down to Los Angeles, then they can predict if you have a similar case with similar set of information, how likely are you going to get a successful rate? Now that's all factual based.

Rory O'Keeffe (31:38.104)
That's left brain thinking. It's not the right brain thinking. It doesn't mean it's the final answer. It doesn't mean it's the actual answer. So they're utilizing the security measures that they've had to put in place, which are heavily, I was about to say police, but it's regulated by the various regulators, like the bar associations, law societies, regulators of law firms. And legal departments is a different story because they're linked to the company's infrastructure. So they'll have those security measures there.

THE HUMAN PROTOCOL (31:54.285)
Mm-hmm.

Rory O'Keeffe (32:07.694)
When you add AI into it, a lot of law firms are getting their own instance of it rather than a version that links to the web. So they have their own instance of it. Some have created their own AI tools themselves. And, you know, very well, they've commoditized those and selling it to other law firms as well. So it's kind of like you bring it into your system and then only feed it what you have. So if it makes a mistake, it's all within your control. It's not being fed.

something from another legal jurisdiction, you know, because it's like the laws in Korea on an area of law might be vastly different to the laws in, you know, say, South Africa, which would be very different to the laws say in Canada. So you have to, you have to take care and how you train the model. So you want to control how you train the model. And that's what they're doing, especially for confidentiality reasons.

THE HUMAN PROTOCOL (32:37.838)
So stays in your tenant.

THE HUMAN PROTOCOL (33:05.196)
And the training will be with your data only, in this case.

Rory O'Keeffe (33:08.772)
Yeah. Ideally, yes. Not always. Sometimes you can justify, you know, like in contract drafting, you can, you'd be happy to use, you know, clause banks from around the world. If it's like for me, it's all in English. It's not good to me if it's in French. My days of studying French are long behind me. My French teachers be appalled with me for saying that. But anyway, it is what it is. I'm now Portuguese because of the Portuguese family side of my half my family.

now Portuguese, so I have to have to get my head around that. Yes, no, no, absolutely no choice. If you've met a Brazilian, there's no choice. There's no argument allowed.

THE HUMAN PROTOCOL (33:40.738)
You have to, you don't have a choice. Yes. We talk about it.

THE HUMAN PROTOCOL (33:48.686)
That's funny, Rory. And yes, we briefly talk about security obligation and regarding confidentiality and sensitive data. guess the prompt and outputs, it will be as you define what will that be about.

Rory O'Keeffe (34:11.588)
Yeah, the the law schools are teaching some are teaching prompt engineering to the law graduates, which I always find fascinating how they're training lawyers for the future. The generation coming through are so much better off than other generations who are still grappling with improve. And that's their prompt. Like it was maybe a year ago, we would say summarize this.

the key issues of this contract. Now it's going to a very long prompt, which has been developed to ensure that it aligns with the requirements set out by the various legal regulators. have seen, I did point it out to them at the time, when my type of work is that I'm kind of, I'm an external, I'm halfway between external and internal counsel. So we call fractional. I that's the buzzword of the year.

And so but I'll sit in with these companies. And sometimes I've been working with external counsel and I've said to them, I've been chatting to them about the issue. One sent me their advice note during the call. And I was curious, because I hadn't finished speaking. I know I'm Irish, we talk a lot. But seriously, you know, at least pretend you let me finish what I was saying before you gave me your big advice. The advice wasn't wrong, by the way. It just

it's it read differently, because I knew that they had. It's like, you know, those, those quiz shows, here's one I made earlier, you when they're making a new cake or something like that, and the other one just burnt in the oven. Here's what I made earlier. So and then it all looks amazing. But the problem is the the rules and the ethical requirements of a lawyer is for them to apply the critical thinking before an advice note goes out. And I don't think that lawyer

And I did call him out and said that you were doing your job by sending him out advice when we hadn't even finished with the salient facts of the issue, which meant that there was no way for that lawyer to review the note unless he stopped listening to me for a number of minutes, just allowed me to wrap it on and then sent it out. You know, I would say to him, if you're going to be using artificial intelligence to manufacture your advice notes.

THE HUMAN PROTOCOL (36:24.461)
Right.

Rory O'Keeffe (36:35.234)
at least wait 10 minutes after the call. Otherwise, what am I paying you for? And I was obviously I was far more diplomatic and kinder than this because as an Irish nation, we do like to ourselves on being friendly. But sometimes I do switch over to another side of my family, German as well. So I do become quite blunt and matter of fact. it's it is.

unfortunate because we get to the point where lawyers are suddenly thinking, I can do this a lot faster. I can impress on my clients, but you have not given them the solution. An example I had of this was in a negotiation late last year, finishing off a merchant acquisition deal. And we were using AI because the client had a particular tool that allowed them to work through a negotiation strategy to try and get the deal closed faster. Cause you can close it faster. You save more money.

and everybody's happy. Great. It was part AI tool. Yeah, we fed in the information we gave them the data and they came up with the results by saying, yes, you need to buy the assets, not the shares. It's an asset purchase, not a share purchase. Okay, fine. As it progressed in the negotiations, the other side was still looking for a price point, which was much more than we had determined was a fair price. And even using the AI tool, the AI tool was

THE HUMAN PROTOCOL (37:33.558)
And that tool was an AI tool.

Okay.

Rory O'Keeffe (38:01.592)
baffled with this as well. And as it is, it's always very nice and always says, no, you're doing a great job. And thank you very much for inputting off. Thank you. I'm doing great because my completely alien AI tool is telling you I'm doing a great job. But the deal was going to go south fast. Why? What were we missing? Well, we did the okay, we need to get the deal done. We're very close. It's just the price points were way off. Everything else seems to be landing neatly. So let's get in a room. And the phrase is we eyeball each other.

So we remove the technology, we just sit down and start talking like human beings again to try and work out what are we missing? And there's lots of tricks you pull. So it's like, will, can talk for the legals. I'll have the commercially talk for the commercials wheel, but I'll also have somebody junior to me who takes notes. But part of their role is to watch the body language of other people. So do they react or not to the things we're talking? What I picked up with others, it wasn't just me, it wasn't like I was the savior of the deal.

But we noticed that what they were real concerned was staffing. So they're to sell the assets. And the person was selling later informed us very confidentially. I only say it now because you'll never be able to track this back. know, don't use any tools in doing this. But had told us that he was terminally ill. So the founder and the owner, who was 100 % shareholder with, you know, family members, wanted to sell because needed to exit fast.

because their legacy for their family. It was more easier for them to set up a trust and move all the funds, et cetera. Now, that shocked us because we didn't know. The second part was that they said they were concerned about the staff because staff had been around for 25, 30 years in some instances. And if you go back into the jobs market, they knew that they would struggle big time. Their severance package is not going to be great. They were concerned about it. But all they said was

We want to do right by everyone. Now the AI tool didn't pick this up because it's not a fact point that they were drawn. It's not something that I was trained to think about. This is an emotive, an implicit thing, reading between the lines. What were they really telling us? And when we came back, we looked at it we said, okay, although the acid is still what we believe it to be, this amount, X million, we do think that

THE HUMAN PROTOCOL (40:16.386)
Yeah.

Rory O'Keeffe (40:30.786)
we can take on these number of staff members at this price, but under the rules, they must come over on the same terms and conditions they did before with the same pensions, all those things. So we satisfy all those. If you're willing to work with us on a plan to get them across, but also to get them re-skilled into these areas, because we've got duplication. So we want to remove the duplication. So we need to start finding ways.

to do that. Now, when we start having that conversation, they said, OK, well, maybe you need this contract, this commercial deal, introduction to these types of clients, which these staff members have been heavily supporting. And they have this great relationship together. So you could take the assets, pull all these commercial contracts, which in the long term will give you another two to five years of value coming in on top, which will justify the staff members coming through. We went.

Well, I didn't because I'm the lawyer to make those decisions. But the commercial team worked through the numbers and said, yeah, we can make this work. That's why I'm saying left, right side brain. Left side brain was only on the data points. That's all they could do. Right side of the brain is what the lawyers come in, the critical thinking, trying to find the solution, look sideways on the deal. then we work through the figures and maybe bring AI to work through the figures, work out

THE HUMAN PROTOCOL (41:37.07)
Hmm.

THE HUMAN PROTOCOL (41:43.5)
Right. Yeah.

Rory O'Keeffe (41:58.872)
know what the final figure should be. And then we can get to a price point which they can accept. And so can we. And people's jobs were ultimately saved. Which is a nice thing for all. A very odd thing for a lawyer to say, feel warm and fuzzy about something, you know, but because we don't have hearts, but it's the it's the reason why and it's like your podcast, it's a human protocol. It's the human element that has to come through.

THE HUMAN PROTOCOL (42:02.924)
I that.

THE HUMAN PROTOCOL (42:11.617)
Love

Yeah.

Rory O'Keeffe (42:26.07)
into these decisions. So AI is the tool, not the answer. That's my view.

THE HUMAN PROTOCOL (42:29.228)
Yep. Yep. And the left side, right side analogy is perfect for this example, right? Because it makes people really understand this in a different way. That yes, while they can assess the data and make decisions based on the data that you are providing to, there's still a lot more. You're more, 50 or more percent on the table that it's our part as humans to come to assess and review everything. Yeah.

Great example. Rory, I'm going to be honest with you. I was waiting for this conversation to ask you about a topic that is very near and dear to my heart, which is IP ownership. I haven't been able to find a clear explanation or answer to me, and I'm hoping that you can at least clarify to us a little bit more. So in times of AI, using AI tools,

Who owns what? Who owns the outputs? Who owns the prompts? What happens to everything that you're using and that AI is providing a result for? Do we own anything or who owns what here?

Rory O'Keeffe (43:42.648)
Okay, I'm gonna be such a lawyer and say, it depends. Thank you. Well, that's why I'm here. That's why I'm paying the money. If everything was so clear, lawyers wouldn't be needed. The reason, so over the last three years, that question has come up a lot. Broadly speaking, it depends on the terms and conditions.

THE HUMAN PROTOCOL (43:46.158)
That's a very lawyer response, right?

THE HUMAN PROTOCOL (43:55.886)
That's true, that's true.

Rory O'Keeffe (44:09.088)
So in every contract, every technology contract that we use, any services contracts, goods contracts, you'll see there is an intellectual property provision within it. And they'll say, they'll carve it out pre-existing IP. That means what's yours is yours. What's mine is mine. And what happens to that intellectual property during the relationship still rests with the original owners. Unless we agree otherwise. And the unless we agree otherwise is usually our work.

product, a deliverable, something that's created, which the service provider usually has no interest in keeping because it's like a report or it's, it's a piece of software that's been so heavily customized, configured for your systems. has no value to the, to the software provider. So we say, okay, you keep that. That's the, that's part of the value of what you're buying. Okay. It's part of the price and that's usually carved out within it.

There's another piece that usually gets written in which is around. I call it the no lobotomy clause. It's terrible. it's the even though if you've got ideas or knowledge, you've got things in people's heads about how they did things or delivered things. You know, you can't stop people from coming up with the same ideas again. So.

Typically speaking, could take you months, especially in technology context, to be able to reproduce the same thing again. the first advantage that you'd have for that technology still rests with the buyer. But I say no lobotomy. So nobody has to just forget everything. Now the rules around confidentiality will still apply. You can't use it elsewhere. Use anti-competitive, those insider trading, all those things may have a knock on effect within it. But ultimately, that's what your IP clause is.

When you take AI into the solution, this came up with there's CASEL coming through, which we're watching in California. There's cases that open AI is different courts trying to do, but you've got the Stability AI and Getty images. So if you've trained the AI model using copyrighted material, which has image rights in place, and I'm not even going to the deep fakes perspective because that's a

Rory O'Keeffe (46:31.31)
whole litigation piece that's brewing, I think, in many jurisdictions right now. But what we have on the just general image generation, there is a lot of laws coming through which says you need to identify it and tell people it is in a machine readable format, that it has been created using AI. So that's the first thing. So consumer base, we need to note. Sometimes blatantly obvious, you know, somebody's on screen in a video with six fingers, you kind of get

Yeah. My one is you've got a law firm which has really young model like individuals with perfect white teeth. Clearly they're not lawyers. OK, because we're never that rested. So it's it's obvious. But that will change over time because the technology is forever improving. So sometimes it will become harder and harder to distinguish between, you know, a generation or not. But who owns it is usually down to.

THE HUMAN PROTOCOL (47:14.795)
Ha ha.

THE HUMAN PROTOCOL (47:18.829)
Right.

THE HUMAN PROTOCOL (47:28.792)
Yeah.

Rory O'Keeffe (47:29.496)
what you pay for. So if you've got a tool, an AI tool that's generating the images, typically in the terms and conditions to say, we don't want it. So you've bought it, you go off and use it, but you might need to mention that you created using my tool. All right, so that might be just the trade off and the decision kind of comes down to what you agree in your contract. There's a lot going on in this field. was the artist who

used was in the States, was an artist who won a major award for generating a digital piece using AI tools. And the question was, is that original, original image or not? Because if you use AI, well, then where's the originality in it? And they were questioning whether they should have won or not. Ultimately, it was decided, yes, should win, because there was so much that was involved in creating this piece, that I wasn't just

create a apocalyptic image of a world in the year 5000 with three people standing on a bridge. You know, it wasn't as simple as that. There was a lot more went into it, but it comes down to, I always say, terms and conditions.

THE HUMAN PROTOCOL (48:47.726)
It's about the contract at the end of the day. That's why it's so important, right? You have to read it or make sure that someone is reading for you. True.

Rory O'Keeffe (48:50.137)
Mm-hmm.

Rory O'Keeffe (48:55.363)
Sorry, or just get me to read it for you. That's totally fine as well. I know it can be a bit too daunting for people to kind of navigate contracts, we've all done some kind of contract law course in our days.

THE HUMAN PROTOCOL (49:11.662)
Yeah, it's so important. What about indemnization? So is there any clause on these contracts lately that we can sue these AI tools or companies developing these tools if there is any privacy issue, any harm, or this is excluded? How is this being handled lately?

Rory O'Keeffe (49:34.949)
There's a few avenues in this complex one, so I can't give you a full brief answer. Yeah, you look at the terms and conditions and there'll be liability clauses, which you can swap in or out and saying whether you're liable or not. That depends on whether it's a B2B or B2C situation. In a B2C, there'll be local laws that will also impact the use of the tools. There'll be jurisdictional questions as well as to which laws apply.

THE HUMAN PROTOCOL (49:59.15)
Hmm.

Rory O'Keeffe (50:03.255)
If you want to see how those kinds of jurisdictional issues and liability issues play out, look at the big tech players, look at their standard terms and conditions that are available online, freely available publicly to people. You'll see like software developer liability points. Sometimes they restrict to regions. Sometimes it's restricted to country, depending on size and where the service being provided from and where the consumers who are enjoying it also.

The other part with this one is on privacy. You've got in the background a shift in some jurisdictions around class actions. So there's a growing number of kind of like these big litigator funded actions where it's if there's a volume of data that has been breached and exposed to the dark web or just generally you end up in a situation where they say, well, there might be 1000 individuals data subjects impacted.

could they combine together because theoretically, they may not even have the money or interest to run an action, and they wouldn't know where but if you've got these big funders for these litigation actions to pull all these plaintiffs or yes plaintiffs, then you end up in a situation where that actually becomes more of a threat to companies than it had been in the past.

piece around regulator actions. So if I look at European and UK context, I can't speak for the other regulators, there's a it's like, it's an enforcement year for them. It's surprising because last year, there was 1.2 billion euros worth of fines imposed in the European Union alone. But what they're saying is that they're looking to say where we're not talking about taking a softly softly approach with

THE HUMAN PROTOCOL (51:50.19)
Wow.

Rory O'Keeffe (51:57.699)
enforcement, we're now actually taking out the megaphone and really squeezing things. Because artificial intelligence has upped the game on the sophistication around security attacks. Fishing those emails where you could kind of spot before you click that something's a bit off on it. Not so easy.

THE HUMAN PROTOCOL (52:20.726)
Mm. Mm. Right.

Rory O'Keeffe (52:22.917)
Think of voice cloning of the CEO or if you're a media boss, they could send you an image, could send you a video. You think you might be on a team's call with the individual, but it's not actually them. So there's that part, I'm not scaring everybody, listen, there are security measures in place as well. it requires like founders, especially a smaller companies need to be live to all of these risks and run the checks, the double checks. That's why we have multifactor authentication.

THE HUMAN PROTOCOL (52:33.324)
Not, yeah.

THE HUMAN PROTOCOL (52:38.38)
That's reality.

Rory O'Keeffe (52:52.293)
It's why we have all these extra controls in place to prevent people from themselves, right? Yeah, yeah, yeah. And honestly, the regulators have been like the ones I've spoken to most with the UK and Ireland. They themselves admit that they know there's going to be mistakes and people human error creeps in and they know humans are typically the reason why there's a breach. But it is

THE HUMAN PROTOCOL (52:57.858)
Better be safe than sorry later, right?

Rory O'Keeffe (53:19.129)
They do understand, you need to show your work. And maybe this year is show your evidence. Show me the evidence, instead of show me the money. But show me the evidence that you have put all the necessary commercially reasonable controls in place to safeguard the machines from doing the things you don't want to do. In the first year of chat, GBT, the Italian regulator suspended chat GBT for about six months because there was an inadvertent disclosure of personal data for one account.

to somebody who didn't want or need to see that information, who was unrelated to that original personal individual and got to see it. because of that unintentional, yes, but clearly the tool itself hadn't been updated to protect against somebody's personal data to appear to somebody completely different and random. know, I think of it, if you're not concerned about your personal data, think of it another way. Maybe it was one of your employees.

You know what? That you are a developer of the next major car for the Formula One racing and you got a new engine or a new formula for the tires. And then you just put it into one of the LLMs just to say, write the scene a little bit better so I sound like a an intelligent human being. And unfortunately, you weren't intelligent because you put it into a tool which was not an internal instance of it. It was linked to the web.

And for some, and once it's linked to the outside world for that entity, that software provider or the LLM provider, you end up in a situation where they have that information. So you've disclosed it. So what do you do now that that could be used to train the model for the next person who asked the question, how do I improve this formula for tires and formula one? And they'll say, we've got a great idea here.

Here's a suggested formula. I'm not saying that that's going to be as easy as that, because every time you ask them a question, gives you a different answer. But if you have trade secrets, the secret sauce of your company, you have to lock that down very, very much from utilizing anything that's open to anything outside of your digital or ecosystem.

THE HUMAN PROTOCOL (55:43.734)
Awesome, thanks for going over all that. And we have been through different phases and talking different topics today on what it's part of a future relationship with any AI vendor or any vendor. But one piece that I would like to talk to next with you as we are approaching at the end of our conversation is termination and exit, which I think everybody.

Rory O'Keeffe (56:08.271)
Hmm.

THE HUMAN PROTOCOL (56:09.996)
You know, when you're starting a relationship, you are not thinking about the end. But I think, especially in businesses, it is important that you know what's going to happen at the end of it if you get there. and concerns that I have that I would like to get your thoughts on it, Rory, is what happened with your data? Does it get deleted? How everything that has been an output from this relationship, your workflows and integrations,

How is this being handled today from your point of view and your experience?

Rory O'Keeffe (56:47.469)
So there's a few angles to this, which although yes, everybody walks into a new business relationship with a big handshake and a smile. Lawyers start from the reverse. OK, I know we're all going to be doom and gloom, but that's what we're being paid to do. If everything goes wrong, what's the worst thing goes wrong and how much is it going to cost us?

then how do we prevent that from happening or how do we ring fence the extent of that potential risk to the company or individual. Exit is very important. Traditional tech contracts would have an exit plan or an implementation or transition period, which is either paid or not over X amount of time, which sets out exactly what everybody's supposed to do in that instance. That might be appropriate. In a SaaS model, some of them might just say, well,

because all we're doing is we're providing you a conduit and we give you outputs, you already have the outputs, that's fine. And all we're doing is dealing with the data you fed in. So how do we extract that data? So it could be most confidential as a cause is we'll have you either return or destroy within X amount of time. Okay. There's usually a caveat to that, which says we will return to destroy except if we're required under law to retain it for longer.

or there's a litigation where we need to hold on to it. Okay. Then, but you only hold it for that purpose and nothing more. So it might be a dispute because you might be terminating because you've had a falling out. So they're the examples we normally have. And we say, okay, that's an easy one in the SaaS model. It's all gone. It's easy. AI is slightly different. I would say that if you're throwing in anything that you don't want to be retained, used, or held by anybody anywhere else, even by this software provider, the SaaS, this AI provider rather.

you will have to get your own instance of it. And that when you switch off the lights, it's all gone. So you don't get to use it. Which brings me back to that. Yeah, yeah. So you need to, you can't, I want to trust people, but I don't. So you need to, yeah, yeah, yeah, yeah. Yeah, yeah. So like, except when it comes to my kids, I don't trust them at all. So that's fine.

THE HUMAN PROTOCOL (58:47.4)
Basically, you are responsible for it.

THE HUMAN PROTOCOL (58:59.16)
That's why you're a lawyer. Part, part,

Rory O'Keeffe (59:09.305)
You know, but like I do trust people, but, but it doesn't require you to eyes wide open. I keep saying, it's like when they put in the clause around, we'll run on with the data and that we continue using it. I need to know how, and I might, I'm the lawyer, so I might not get the technicalities of the processes that are going to implement. So get somebody in your team to do that. Get one of your geniuses out there in security, just to sense check it or get a third party.

THE HUMAN PROTOCOL (59:22.958)
Yeah, me too. Yeah.

Rory O'Keeffe (59:37.081)
to swoop in to do a quick sense check to say, yeah, yeah, this works. And we can confirm that everything is done because if your regulator or your founder or your shareholders start saying we need to audit, you know, your tech and we need to audit because there was a breach. We don't know where the breach arose. And you cannot show how this tool damaged or didn't damage anything. Then you don't have the paper trail that you need to show us. You need to have

timestamps, need to have logins, you need to have audits in place, as we would have for general data purposes, you need to have the same in this instance. So when you exit, yes, you need to think about the course, the time frame, the mechanics, the replacement supplier, using it yourself. The other part is tech debt. That's the last thing I would say on this, which is the concept of you're locked in. So you need to ensure that whatever you put in, you can easily take out.

and that you can easily move like we do with banks systems as well. So you need to have the ability to move to a new provider seamlessly because you don't want to be in a situation where they say, oh, well, you're going to have to give me three other people to help you with the transition. Plus, you need to pay the day rate for X number of months. And they drag their heels just to squeeze out the maximum amount of money out of you before you part ways. That's not what you want.

You want a clean in and you want to clean out as well. So you should be looking for them to show how they'll make it as easy as possible for you to move. And yes, because remember, you're going to have to learn a new piece of technology. You'll probably have to do a train the trainer with people. You probably have to change management takes time. Convince people that the new technology is amazing. It's one of the dashboard is even shinier than the last one. Whatever you're going to have to do as part of the change management program.

moving to piece technology. That's going to cost you enough. So why don't you make it very simple on the exit piece?

THE HUMAN PROTOCOL (01:01:42.257)
Yeah, they begin, right?

Cool stuff, man. Thanks for going over that. And last but not least, this is personal one in a way. It's a piece of advice that you will tell a younger Rory, a younger version of yourself seeing him today. So what would be a piece of advice to him, especially in this new future that we are getting into with AI?

Rory O'Keeffe (01:02:12.709)
I love your question and hate your question at the same time. That's why I think with the context, I would I would probably say, first, I remind him change is good. So change is a constant. You've always known it, but change is good. The next thing I would tell him is so AI in the

THE HUMAN PROTOCOL (01:02:16.462)
That's what I'm here for.

Rory O'Keeffe (01:02:41.281)
new sense, the generative AI piece is coming through. It's not going to mean the end of the legal profession. Even though I think it's a lot of people out there would love to see the end of the legal profession, but it's not going to see the legal profession. But what's important is to is not to never rush. Just take your time.

And with business, I would just say as another thing I would say to them, because I've made, you know, some obvious errors and certain choices of suppliers, not in the AI context, but just generally on stuff and saying, so what is my answer? As every founder out there, you're going to make choices. You have to make a decision either way. Procrastination is the death of your company. So just so long as you're moving forward, you're making progress. I have this

thing in the past, I had, it was a few years ago, I had set myself as one of those, you when you've got personal targets you want to put in place for the year, professional targets in the year. And then I always asked, what do you want to do personally? said, you know, I really want to get back to reading more. And I read so much as a lawyer, but not the stuff that, you know, you used to read. Like I was a big fan of Terry Pratchett books, fictional books and John Grisham, even though I criticized every single legal courtroom drama.

that was going, but I still enjoyed it. Michael Crichton, who wrote the Jurassic Park series, loved all that stuff. But I don't read them because I just didn't give myself time. know, family, kids, all these things interrupt. And I said, I need to do it. And they said, but I never have time. And then was a friend of mine who said, yeah, but what are you trying to do? I I need to read a chapter, you know, a week at least. he said, do you? You commute if people still commute. on the rare occasions we're not working from home or remotely.

When you can mute you say, why don't you just set your target of 10 pages? Because I want 10 pages won't take me long. I could skim through that quickly and say 10 pages. I said, well, why? Because if you're doing 10 pages every day, and maybe you skip a few days, do five days, that's 50 pages in a week, more or less. And a book is what, 300 pages? How many books are you gonna get done then in a month? Oh, yeah. And don't, don't, don't.

Rory O'Keeffe (01:04:58.277)
Commit yourself to that, just saying, that's what I'm gonna try. If I get there, I get there, okay? Same as New Year's resolutions. We always go too far, too fast, go too big. If you're making that, by the end of the year, you'll have a few books finished and compare that to the last year. And you're sitting there going, yeah, okay. It's like, makes sense.

THE HUMAN PROTOCOL (01:05:18.847)
Hahaha

Rory O'Keeffe (01:05:20.985)
But the mantra is like it's micro steps. If as a founder, you're making those steps and you're going forward against your business plan, you're making even the smallest steps forward is still progress. And you should be happy you're making progress rather than looking at all the things that you want to achieve. And you haven't because there's so many big ideas. It doesn't matter. You're making progress. Focus on those things. Life will be a lot sweeter.

THE HUMAN PROTOCOL (01:05:42.826)
Awesome. Awesome. Thank you for going and telling us that. And Rory, really, thank you for being with us today. This has been very practical and I appreciate you a lot.

Rory O'Keeffe (01:05:55.439)
Pleasure, pleasure. Thank you for having me on the podcast. Thank you.

THE HUMAN PROTOCOL (01:05:58.718)
Awesome. And if you are a founder about to sign your next contract, send this episode to your co-founder, your obsolete, or even your lawyer. Follow the Human Protocol and join us every week, how we explore how to remain human while building the future. Thank you all. Talk soon. Take care.