In this episode of the AAAi Podcast, Bridget McCormack and Zach Abramowitz talk with Farah Gasmi, cofounder and Chief Product Officer of Dioptra, a Y Combinator backed startup that’s trying to make contract work feel less painful. Farah didn’t start in law; she built AI for industries where mistakes aren’t an option. That background taught her a simple rule: if the AI isn’t reliable, it doesn’t matter how fast it is.
At Dioptra, that rule gets applied to contracts. Lawyers using the product say it puts them “in the zone,” like athletes at peak performance, admin noise stripped away so they can focus on lawyering. In a sea of redlining tools, Dioptra wins when the draft that comes back is already usable, not something you have to rebuild from scratch. And the partnership with the AAA takes it one step further: a drafting engine backed by validated ADR clauses lawyers actually trust. The bigger point? ROI in legal AI isn’t just hours saved. It’s whether lawyers feel sharper, more focused, and less drained at the end of the day.
Key Takeaways
- From admin drudgery to vibe lawyering: Most “legal tech” in the past just gave lawyers more buttons to click. AI feels different. One user told Farah: “I feel like an athlete.” That’s not about minutes saved, it’s about lawyers actually enjoying the work again. Instead of slogging through admin, they hit flow. Call it vibe lawyering.
- Not all redlines are created equal: Every tool says it can redline a contract. The question is: do you get something you can actually use? With Dioptra, lawyers say the draft comes back 70–80% of the way there - not 20%. That’s a huge difference. And it lives right inside Word or your CLM. No new platform, no re-training. Just better drafts where you already work.
- Billables vs. business cases: Law firms don’t care if you shave five minutes off a draft. They care if the work looks better and if associates don’t quit from burnout. In-house departments, on the other hand, have to justify every purchase with spreadsheets full of “time saved.” But here’s the truth: neither group sticks with a tool unless the lawyers actually like using it. ROI isn’t hours on a chart; it’s whether people want to come back tomorrow.
- Vibe lawyering with receipts: Speed is nice, but nobody’s betting their reputation on speed alone. Dioptra’s partnership with the AAA adds something AI usually can’t: credibility. The drafting engine pulls from a validated ADR clause library, so lawyers know the language is solid. And because software can be updated weekly now, not yearly, features like this don’t sit in limbo. They roll out, get tested, improve. That mix—trusted content plus rapid iteration—is what makes adoption stick.
Final Thought
For decades, the promise of technology in law was efficiency: shave some time here, cut some costs there. That’s not what’s happening now. The tools that will matter in this new wave aren’t the ones that brag about being the fastest; they’re the ones lawyers actually trust, and the ones that make the work feel better.
What Dioptra shows is that ROI in legal AI is shifting. AI is allowing lawyers to stay in flow, produce sharper work, and walk away from the job a little less drained. Add in trusted partnerships, like the AAA’s clause library, and you get something rare: AI that combines speed with credibility.
This is a signal that legal AI is maturing, that adoption will come not from hype, but from tools that make lawyers feel like athletes in their own practice. And that may be the most important shift of all: the future of AI in law isn’t faster lawyering. It’s better lawyering.
Transcript
Zach Abramowitz: So welcome Farah to the AAAI podcast. We're very excited to hear a little bit more about your product. Why don't you introduce yourself and tell us what made you want to start Dioptra in the first place.
Farah Gasmi: So thank you so much, Zach and Bridget for having me today. So I am one of the founders of Dioptra. It started off as a contract review AI tool, and now it is really the tool that helps legal teams do their day-to-day. So that can be contracts, obviously, but it can be a number of other things like compliance or drafting and so on and so forth. This is something particularly interesting coming into the legal space. I actually started my career in AI. I've been building AI products for the past decade at a number of different industries and companies ranging from healthcare and pharma to financial institutions and insurance. And I was building the AI infrastructure at Spotify most recently. And so we started this company with my co-founders coming at it from kind of the AI perspective. The feel or the need that we were identifying was that legal really needed more reliable AI systems.
When we started this, ChatGPT was coming out. And we had been in this space already for a while because we had already founded our company. But we were doing a lot of consulting work for our AI teams. So think about mission critical AI use cases like autonomous driving, asset protection. We were working with Walmart, Hyundai, Volkswagen, all of those big companies that needed to make their AI systems more reliable. And so that was kind of the DNA of the company, if you will, or the way we had started it. And then there was a lot of demand in the legal spaces and natural fit for LLMs with the legal space. We had had our own experiences negotiating or trying to understand what those contracts were about with our own pain points.
And so at some point we're like, this seems to be a bigger problem here to be solving. We were part of the YCombinator community of founders and realized it wasn't an us problem, it was a bigger problem. And so that's when we decided to focus on the legal and specifically contracts. And then the focus for us was making sure that we were bringing in more reliable AI for the contract space. Fast forward
Bridget McCormack: Makes sense. That the story makes sense if you're specializing in AI for high risk use cases. You know, legal's ripe for that for sure. Can we pause for a minute and have you tell our listeners a bit about Y Combinator? They might not all know what that is because we might have a largely legal audience. What is Y Combinator and how did you hook up with those folks?
Farah Gasmi: Yeah, so Y Combinator is, I want to say it's the most prominent incubator of startups in the world. They have been around for 13 or 14 years. And there's a lot of companies that went through Y Combinator that you would know. So Airbnb is one of them, Coinbase is another one. There are a lot of unicorns that went through Y Combinator. What is special about Y Combinator is that they're obviously investors, but it's also mostly an incubator. And so the acceptance rate in Y Combinator is very low. We joke about it. I think it's harder to get into Y Combinator than it is to get to Harvard. And it is extremely powerful as a network because you have all those founders that help each other that are working together to kind of build the future of what tomorrow will look like. And so the way we got into Y Combinator was a little bit funny because we were not really planning on doing that. One of my co-founders said, guys, the deadline for Y Combinator is actually tonight if we want to apply..
My co-founder is in SF so he's plugged into all of that and we're like okay should we do it? I remember the deadline was at 11 p.m. and at like 10:58pm we clicked submit because we had just done it at the last minute. And they accepted, we had interviews obviously and so on and so forth. I think what played in our favor is the founding team is very technical. So we've all been building AI products for the past decade, each one of us. We had this fairly different set of skillsets. My expertise is more in product management. My other co-founder is really an AI kind of scientist. He was at IBM Watson and my other co-founder is the software engineer, he's done that for a number of years. I think that's what played in our favor.
Zach Abramowitz: So it's interesting that you went to YC and that you also have this sort of consulting background and the company morphing from a pure, consulting play into a tech company because I know that one of the things that Paul Graham actually advises startups in the early stage is to think a lot like consultants, to go boots on the ground and really sort of fall in love with the customer's pain points and then build. Have you found that experience to be true?
Farah Gasmi: Yes, absolutely. I think that was kind of the critical aspect for us. But to be fair, in what we were doing in our space, because we were really helping AI teams build more reliable AI systems. The truth of the matter is that the SaaS model didn't end up working for us. We were stuck in consulting. We started in the consulting space with that mindset of, hey, we'll figure out what the problem is. But we found ourselves stuck into that consulting business model because Accurate AI was perceived as a secret sauce and people don't want to use a SaaS for it. They want to have that skill set internally. So we were finding that every POC we were doing was extremely successful. And the next conversation was, you guys want to join us and work with us and help us build this kind of reliable AI system? No, I'm trying to build this SaaS. So I think we kind of hit a ball at some point in this specific space because of the nature of the problem we were solving. And so that's actually also part of why we decided to pivot. And now, it is really something that they want to build internally that becomes their secret sauce. It's like you say, OpenAI, hey, I'll train your models. What does OpenAI do? You see what I mean? And so that was kind of the perception.
Zach Abramowitz: Meaning companies don't want to outsource reliable AI.
Farah Gasmi: It took us a little bit of time to figure that out before realizing that it was not something where SaaS was going to really work.
Bridget McCormack: How has that impacted sort of your strategy, your roadmap, your business model? is that? Spell it out for us a little bit more. What does it mean?
Farah Gasmi: We were lucky to be part of the Y Combinator Network, right? Because we have a lot of advisors and they have seen our problem with some other company at some point. They've seen so many different companies, so many different journeys, like founder journeys. They're a great resource from that perspective. And so we sat down with them and it was here's what we're seeing, right? And they helped us formalize our thinking. It took us some time to realize what was going on.
When you're on the ground, it's not obvious what problem you're facing, right? And so it took us some kind of noodling to understand what was happening back then and how to change the whole game. And what it took for us is to do a significant pivot, not like small pivots, incremental pivots, but we had to completely change the space. So instead of being at the infrastructure layer, we decided to move upward in the application layer and specifically the legal space. We had our own kind of experiences, like I said, you know, working with lawyers or reviewing contracts that led us into this space. And there was also, if you remember, like two years ago, there was a lot of talk about legal being a huge use case or application for LLMs, right?
Zach Abramowitz: since the day that ChatGPT launched.
Bridget McCormack: Yeah.
Farah Gasmi: Exactly, absolutely. so, and, it was clear to us too, right? And not only that, but there was a lot of talk about hallucinations. That's our bread and butter. We know how to do reliable AI systems. So we thought, and we still continue to think that we add in something to the table with the legal application.
Zach Abramowitz: And this sort of goes to something I heard recently. was interviewing Jonathan Levy from Y Combinator on stage at Legal Innovators.
Farah Gasmi: Here we go.
Zach Abramowitz: Earlier in the summer, and I asked him, said, has Y Combinator gotten more interested in legal tech? I know you've been investing in it for a while, and I know it was an area of interest for Sam Altman, when he was president of Y Combinator, but is it something that YC is paying more attention to? And what he explained to me was that actually we don't look at any vertical, we simply look at top founders. Which obviously must be flattering to you, but in other words, we only look at top superstar founders, and what we're finding right now is that more top founders are drawn to legal, which it sounds like is your story. Have you seen that playing out in the other founders that you're talking to and the type of people that are now trying to solve problems for the legal profession?
Farah Gasmi: I think that is a little bit more true now, than it was when we started two years ago. So just to give you a little bit of a timeline. So two years ago, we had just started and we decided that we were going to look into the legal space. We had contracts as one of the applications, but we were not 100 % set on that. So two years ago is when we started doing that kind of discovery, going on the ground, talking to people, understanding what they needed. you know, getting a sense of where AI was going to be applied. And there was a lot of skepticism obviously back then as well, And so that's when we started, we met the team at Wilson-Sincini and then we started building with them.
And so I would say back then, most of the people that were in the legal tech were lawyers. Lawyers that, because they had got a good understanding of the space, they knew better what AI could do differently. think...
Zach Abramowitz: It was typically a lawyer between like a second and a fifth year or a second and a seventh year that was deciding I think I could make more money, do something more creative, solving some use case that I've been working on here at the firm. And that's where you saw a lot of the founders would agree, yes.
Farah Gasmi: Exactly. Yes. Absolutely. a lot more even in-house. There's a lot of in-house folks as well that were coming from that background. Today, I do think there are a little bit more of those tech founders that are coming into this space.
Bridget McCormack: Let's talk about your product for a minute. YC always says build something people love. And I always think of it like, when lawyers are doing something, almost nobody loves anything about it. But give us the elevator pitch. What makes lawyers' lives better when they use your product? Why does it make their day a little better?
Farah Gasmi: Yeah, you know, it's interesting what you say. It took us some time to kind of understand the thinking and just get that empathy with lawyers' journey and day to day lives and what they like and what they don't like and why is that. And I think there is something fundamentally with AI that is different from the past legal tech, which is that lawyers are seeing that AI can actually, I don't know if it's 10X or 5X, their own skills as lawyers. It makes them better.
And I think that wasn't the case for all the tech that they had seen. A lot of times it felt like more admin tasks that were added on top of their plates. I think the difference with AI is that they're seeing they're getting better, they're better performers in general. And then the other thing that I'm seeing is that the job market is changing for lawyers. People are expecting lawyers to be able to work with AI. And so they're seeing that from their executives, they're seeing that from the jobs they're applying to, they're seeing that everywhere around them, right? And so I think that's kind of the dynamic too. Just wanted to touch on the difference between AI today and other legal tech in the past.
And so to answer your question, AI today is really helping them get better, more thorough in their lawyering tasks faster. And I think that one of the analyses or comparisons I got from one of my users was, I feel like an athlete. Like, you know, when you're in the flow and AI is removing the noise for me and I can just focus on performing at my best. And that really resonated with me because that's really what the low admin type of task is taken care of and they can focus on what really matters.
Zach Abramowitz: So we want to jump into this because this is something that Bridget and I talk about quite a bit, right? I think this is part of the reason that many people don't know how to measure the ROI of the current AI tools because they're trying to measure them like legacy products. And most legacy products that I think of were built first and foremost for the company.
And then in order for the company to get the value, employees were forced to use this. So it actually felt like additional tech that they were required to use. It wasn't built for them. AI on the other hand to me is like Oprah walking on stage and saying today everyone gets a car. what organizations who have bought AI and basic ChatGPT licenses for their company, it's like coming in saying today everyone gets their own personal assistant that's going to 10x them and I just wanted to pick up on one more point that you made It the ROI that people are measuring at firms for these types of products really should be like interviews with the lawyers saying
How do you enjoy working with this product? Does it feel like more of a flow to you? Do you find that the work quality that you're producing is better? I think a lot of people are looking for savings on the bill, and I actually think that what they should be looking for is exactly what you said, better lawyering, and I think that also it's not just the better work product, and Bridget, you know, I've talked about this a lot, is that when you were working with AI, it is like a true, deep work experience, where I think the term like vibe coding or vibe lawyering, is a feeling of like being able to flow and work with multiple dynamic tasks without like breaking concentration. And I think this is so fundamental and yet it won't show up right away on a PNL.
Farah Gasmi: I agree, I agree with you. It's funny that you say vibe lawyering, that was something that we were discussing earlier this morning with my co-founders. That's what I'm seeing my users do, literally. And they're like, I like that, I didn't think about this. Or it's raising a good point, or I can give it feedback and kind of iterate on that. That's exactly the experience, right? I talked to a lot of different teams, the recurring question that I get is ROI. I'm actually, so it's funny, because like you said, Zach, earlier when we were talking, we work with both in-house teams and law firms.
We've realized that it's a different story actually. If you're talking to a law firm versus in-house, it is actually a different story. What resonates with law firms is that experience. How happy are you? I'm a data person. It's extremely hard for me to accept that as a data point, but it is a true data point. That is the only data point that works. How happy are you with using the product? Do you feel that it is saving you time? Do you feel like you're doing a better job as a lawyer? Do you feel like you are being more thorough in your analysis? And that is the only data point that actually has been kind of resonating on the law firm side because everything else, they're not incentivized for efficiency, right? That's not what they're optimizing for. What they're optimizing for is good lawyering. And that's what resonates with them.
Zach Abramowitz: and by extension, talent retention.
Farah Gasmi: Exactly, that's exactly true, absolutely.
Bridget McCormack: Their business models are kind of adverse to each other, right? Like the, you know, the in-house teams are cost centers and the law firms are revenue centers. Which makes it a little bit unusual that you work with both. Often we see founders picking a lane and marketing to that lane, usually in-house teams. Tell us a little bit more about what you're seeing when you're working with the in-house teams. What do they love? What are you measuring there? What makes a difference?
Farah Gasmi: Exactly. Yeah. So on the in-house side, they need to build a business case before they can even start using a product, right? And that is usually numbers. That's how you get in the door. That's how you help them adopt tools. And that usually is referring to efficiency or like time saving and so on and so forth. But that's only the way in. The way to stay is if they like it, right? So it comes back full circle. Do you feel like you are a better lawyer using this tool? Do you feel like it is helping you be more thorough? And if the answer is yes, then that tool sticks.
Zach Abramowitz: It's so interesting, Bridget, we often talk about having the right incentives in place for usage, but incentive sometimes is something that you need to almost explain and paint a picture to the users. And in this case, the picture that you're painting is not one in which you're going to save, you know, 15 minutes on the client's bill or that you're going to be able to mark down certain kinds of work. the real ROI that I think or the incentive that matters to many attorneys is am I doing more thorough work? I think they're much more interested in that by design and the profession, than they are in - could I do this in 25% less time? And that's why again I think the ROI is not going to show up for a while on P&Ls the way that we sort of have been used to. It's like, if I can just automate this sort of legacy task or I can automate this administrative task, like the place where it's really going to show up is, do I work better?
Farah Gasmi: Yes, absolutely. And it is something that it's also something that you see over time, right? It's not something that is going to show up, like you said, something that you can measure immediately. Although, I would argue that there are some use cases where it's easier to measure that. For example, on the in-house side, there's some use cases where our customers' in-house teams don't want to review NDAs. I don't think there's a lot of value in me reviewing NDAs, right? Or low risk agreements payments in general and they want to allow teams to enable the other business stakeholders to be able to do more of that. So that's a clear kind of use case where it's a clear efficiency gain but it doesn't paint the full picture. I think there is more to it.
Bridget McCormack: I'd be interested to hear about how you think about the other businesses in the market also offering contract redlining products. A lot that used to do it before AI who now say they're doing it with AI. It feels like a loud marketplace in a way. And also I talk to lawyers at law firms who just throw their contracts into Copilot or one of the frontier models and get feedback and they think like the frontier models are doing a fine job. Tell us more about what your product does that makes it different.
Farah Gasmi: So, as a founder, I try to not get too distracted by competitors because like you said, it's extremely noisy. The market is extremely noisy, right? And instead what we try to focus on is what are our customers and users saying? And I can tell you what they tell me. The biggest differentiator that they see with the Dioptra, there's two aspects to it. One is the quality of the outputs, the legal outputs. So think about it as a red line or legal analysis or comparison or search or anything, any legal tasks. that the users are conducting the feedback is that it is higher quality, right? It is more usable as is, as opposed to me having to fix or iterate on it to get it at different times. So that's by far the biggest feedback. And that goes back to helping them be better lawyers. The better the quality of the AI outputs, the better incremental performance they get, the more value they get. So I think that's a fundamental aspect.
Zach Abramowitz: Meaning the ROI that you're optimizing for dictates your product roadmap, dictates your iterations, so that you're entirely working on what makes their lives better, what makes them feel better about the work process.
Farah Gasmi: Exactly, exactly. And what are the outputs that are going to get them 70%, 80% there, right? As opposed to like 20% there. So that's really what we are optimizing and focusing on. And I'm happy that that's the biggest feedback we get, right? So that means that we're doing something right.
The second thing is lawyers don't want to have to learn other tools. Right? They want to be able to stay in their tools and that is part of the efficiency gain for them. Right. And so if you were to compare going back to your point, Bridget, about the frontier models, the frontier models work. They're good at summarizing. They're good at flagging risk and so on and so forth. They're not good at working Microsoft Word, being interactive with your tools, whether it's a CLM or DMS, and being able to get you the information from the work you have done in the past. And I think that's the biggest aspect that makes the Dioptra valuable or similar tools is that it is really working in your environment, it has access to your knowledge, what you have done in the past and it can leverage that as opposed to you feeling like you're having to start from scratch.
Zach Abramowitz: I wouldn't have ever asked this before the rise of large language models, but given the consumer adoption of these tools, do you foresee a day in the next few years even where the idea of lawyers always working in Microsoft Word might not actually be the case? Three years ago, I would have thought, yes that makes sense,to be in Word, but I wonder if the new Word will effectively be something like a ChatGPT where, and we'll come back one day and say, well, most lawyers work in their ChatGPTs, you've got to build the product to work with ChatGPT.
Farah Gasmi: Yes, something that my founder, Jonny has taught me is never say never. I keep getting surprised with how people are adopting tech, how they're pushing back on tech and the technology itself evolving over time. So I do think there's going to be opportunities. I don't think we are there yet. I think people today are still having their way of doing things. And if you were to force a different way of doing it like other than Word I'm not going t0 fight that fight because I don't think today is worth doing that. People are too attached to Word that they find it effective for what they need to do.Can it change in the future? I'm sure it will I don't know how yet but I'm sure it will in the future
Bridget McCormack: Can you tell us a little bit more about your approach to rolling out your product? You are, I think, a big proponent of iterating and just getting constant feedback from users. I'm thinking about this all the time with some new products we're working on. And I'd love to hear you talk about it, because I think it's a unique approach and not sort of the way we used to do it with rolling out new products.
Farah Gasmi: Yeah, so, I don't know if I mentioned this. I teach at Columbia. I'm an adjunct professor at Columbia and my class at Columbia is about product management and specifically focusing on AI products. And so the reason I'm saying that is, I tell my students, if there's something you need to keep from this whole session or class is you have to iterate and you have to be able to adapt to new information you collect. And so that's something that I very much believe in and obviously apply in my day to day as a founder and my co-founders apply as well.
And I'm going to take this one step further. I think in the past when we used to develop products, I think there were a lot of iterations, but in a different way. We would do a lot of iterations on prototypes or designs, right? That's how we used to build software or SaaS. And so we would have wireframes and designs and we would iterate on that because building software was expensive, right? And so you would rather iterate on that than iterate on the software once it's been built. And I think that is changing. The cost of building software today has decreased significantly because you have tools like Cursor and other AI development tools. And I think now we are actually going to be iterating more and more with customers with a product that is live for them to try.
And I think the quality of the feedback becomes much higher because not only do they have the designs, but they have the whole interaction. They have the quality of the outputs. They understand what to expect. They have real life experience. Right? And so I think the result is that these lower quality features are being shipped faster, but faster iterations based on feedback received from the customers. and so that's what I tell my customers. When I work with my customers, I always say my goal is to have good partners. More so than ARR, my goal is good partners. I would favor working with real partners that are going to take that feedback and share and tell me what it is that's working well or not working well and giving me other use cases. And you know what? Lawyers are very receptive to that, both in-house and law firms. they are relationship people, right? And they do appreciate those relationships with the legal tech folks. And so we've been blessed from that perspective. It was something that we didn't know, not coming from a legal kind of background. And we have found to be very successful building those relationships with our customers to be able to iterate fast.
Zach Abramowitz: Yeah, and it occurs to me as you say that we're seeing this skyrocketing growth and adoption of AI tools at a pace that we have simply not seen in the past. And as you're saying this, it occurs to me that part of the reason that must be true is that iteration upon software is much less expensive than it used to be. So you can get to product market fit a lot quicker, not just because the initial build was less expensive, but because updating it along the way is going to be drastically less expensive. So we should see rising adoption of those products. Farah, you have a collaboration going with the American Arbitration Association. Tell us a bit about that.
Farah Gasmi: Yeah. So like I said, we love partners and the AAA has been a great partner working with us. And so the way the collaboration works is that the AAA has developed a set of clauses that have been proven and validated in the industry. Dioptera is a word plug in. And so when lawyers are drafting, when they review contracts it's a big use case for us. And one common question that I get is, can the tool help me draft? How does it do it? What information does it leverage when it's drafting? And the fact that we've been able to build this collaboration with the AAA has helped us add another reliable source of information for AI when it's drafting specific clauses. And so now, the lawyers like you said Zach, they're vibe lawyering, but they're vibe lawyering with good quality, reliable sources and the AAA's database of clauses has been extremely helpful from that perspective.
Zach Abramowitz: Farrah Gasmi, co-founder and chief product officer at Dioptra, thank you so much for joining us. This has been a really interesting conversation, learning about the product you've given me a lot to think about in terms of not just how to measure ROI and get it right, but also then how to manifest that in the product design itself. So thank you so much.
Bridget McCormack: Yes. It's been great to work with you, Farah, as a collaborator. And I sort of feel like now as your student, I really appreciate this conversation.
Farah Gasmi: Thank you so much for having me.