Exploring the Future of AI Writing Tools with May Habib, CEO of Writer.AI

In this episode, Phylise Banner interviews May Habib, the CEO of Writer AI. May shares her expertise in natural language processing and machine learning, highlighting the potential of AI to revolutionize the writing industry. May also emphasizes the importance of training and enabling employees to leverage AI, ensuring that no one is left behind in this transformative process. The conversation delves into the future of paid versus free AI tools, the role of curiosity in shaping the right questions, and the need to address bias and diversity in AI-generated content. Despite the rise of AI, both Phylise and May emphasize the enduring value of skilled human writers who can infuse creativity, personal stories, and authentic human interaction into their writing.

ABOUT MAY HABIB
May Habib is co-founder and CEO of Writer, an AI writing assistant for teams. May graduated with high honors from Harvard, is a member of the World Economic Forum, and is a Fellow of the Aspen Global Leadership Network.

The Innovations in Training podcast is sponsored by Symtrain — an AI-training platform that automates time consuming role-play exercises required to prepare your employees for interactions with customers. Symtrain helps teams learn more, faster by creating an immersive CX training experience.

The Innovations in Training podcast is presented by Training magazine.

Transcript

Phylise Banner: Greetings everyone and welcome to the innovations and training podcast. I am your host Phylise Banner. I'm a learning experience design consultant and master chef and training magazines innovations and training test kitchen. In this episode, we're going to explore AI writing tools. With may Habib, CEO of writer AI, I met Mae back in October of 2022, and was immediately impressed with her knowledge, her enthusiasm, and her insight into AI has potential to revolutionize the world of writing. Welcome, May, it's great to have you here.

May Habib: Thank you so much Phylise, I was so heartened when you got so excited in our first meeting, this was kind of before the, you know, major hype around all of this. And so the fact that you and your peers were so excited about it, I thought “woohoo”!

Phylise Banner: Woohoo is right! So, tell us a little bit about yourself may and what you do, and what led you to do what you do now.

May Habib: Yeah, so a little bit about myself … I have been in NLP and machine learning for years now, our first company we started nine years ago in machine translation. And it gave us two incredible insights really being in machine translation that early one was this real understanding of all the content processes that happen function by function inside of a company. So, you think about a company that now needs to enter a new market, everything from top of funnel to bottom of funnel is going into that new language, including training and support and L&D and everything that you need to do to be successful. And so our team had this like, real uncanny ability to like, very quickly grok, a word, a content workflow. And so then when we put that together with our very early look at what transformers could do, because of course, so much of that underlying research was happening for machine translation use cases. And we were at the forefront of using those tools, and approaches, it was obvious what we needed to do. We needed to build a way for companies to more easily make use of transformers and later pre-train transformers to address so many of the complex content challenges that exist inside of the enterprise.

Phylise Banner: Wow. So let's jump right into Writer AI, which is where you are now. And tell us about Writer and its key features, and how it difference differentiates itself from other tools on the market. And, you know, I think three months after I met you, ChatGPT became all the rage right? And, and it was so interesting to see what writer could do versus what ChatGPT could do. And that's when I started going down this rabbit hole to every writing tool that was out there.

May Habib: Sure. So, the the big differences come down to a few pillars. Number one writer is a full stack platform. So, we are both the foundation models at the bottom, generating all of this incredible content and insight. And we are a customizable UI that sits on top of that. So whereas think about you know chat GPT is kind of one window ask response ask response, which is very useful for some things, writer allows companies to actually design the interface for specific use cases, be it a claim response or a client communication or building slides for a presentation you name it, you are able to design the interface for each particular use case that your users are are using. So in the case of L&D, what are we going to ask from our instructional designer as inputs when coming up with a new quiz? Is it going to be reference material that they put in as URLs? Is it going to be the PDF of the instructors guide? Is it going to be notes from the speaker section of a PowerPoint presentation? You name it? That UI is highly customizable. So that's pillar Number one, full stack platform.

Pillar number two is all of the security and privacy. So that whole platform can actually be hosted inside of a customer's own private cloud inside of their own cloud environment. And why that is important is, especially for these high value use cases, we're talking proprietary data, I don't want to be sending all of that to Microsoft, right, and we've got lots of Intel around what Microsoft is actually doing and the fine print here, and it is not pretty. And so by really having that model, or that UI layer, or both inside of somebody's own environment, they are that much more comfortable sending the proprietary data, to train the model and fine tune it and asking, you know, 1000s of their employees to actually input sensitive information or proprietary information in there. So that is, that's pillar number two.

Pillar number three is the massage methodology and enablement around how we actually roll this out and train people on how to use it. And I tell the team all the time, we are very aligned on this, if like, our legacy in this world is that 30% of employees can do the work of the other 70%, we have failed. What we want to do and what our vision is, is to enable every employee to be aI enabled. And so that is not just software and technology that is training that is showcases that is inspiration, that is really making sure that nobody is left behind here. Because the potential for gross inequity at work between those who like really grok this and give themselves leverage and those who don't, that's just not going to be a fun thing to happen for anybody. So those are the three big ways in which we are different, not just from chat GPT. But really just about anybody in this space right now.

Phylise Banner:  Yeah. And you just brought up some points that I want to hit on May. So thanks for throwing these words out here. Because when you talk about this, you know - the ask/response, it's this ask/response model. In some cases, it's free, right, and the tools out there that are free, that are generating text, visuals, music, you know, anything. And then there's the paid version. And in my experience, and I've been trying this out for months, the paid version gives you something better.

And you talk about inequity, and you talk about leaving no one behind. So where are we going with this future of paid versus free? And you know, I think about this from the perspective of learners, I'm doing a lot with universities, how to incorporate AI into teaching and learning. You know, what does that look like? So your thoughts in this space, May?

May Habib: Yeah, totally. So first, let's let's divide up the world into consumer and, and enterprise. And I think, and this is a great thing, that there is going to be a huge coming commoditization of consumer applications. We saw this firsthand, in the localization space that we were in for many years. translate.google.com did not destroy the localization industry, it made it so much bigger, right? Because people could kind of get the light good enough out there, could try things out, and then could go, you know, to town, when something was proven.

I think in consumer, we are soon going to be able to have free multimodal applications where you know, a few words, if you know how to ask the right question and why let's come back to that, right. If you can describe what you want to see whether it's a video or a piece of text or a piece of audio, you're gonna get something pretty good back, and you're going to get that for free. That's consumer.

And so I think some of the quality differences on consumer applications that you're seeing right now, between paid and free consumer, those, those gaps are going to be closed and the big companies are making sure that is like a race to the bottom, which is good for consumers. on the enterprise side. Our bet is that if you look at use cases on this matrix of, you know, high workflow and high requirements, right, so for high consideration products or types of analysis or generations where, from a human workflow perspective, from a regulatory perspective from a compliance perspective, Um, there are going to need to be a set of eyes, whether those eyes are like post processing and LP API's or whether those eyes are expert humans that have to be involved in generating the insight or the content.

And for those use cases, neither the consumer application applications nor the generative AI being built into every platform are going to be sufficient to meet those needs. And that's where we think people need a platform like Writer. Another thing that we really believe is big companies are going to have multiple approaches to all the different generative AI use cases. And we can definitely, and we do help people think through kind of the low workflow, low requirements use cases and those tradeoffs, like, should I get everybody a chat GPT license? Should I let people use Bard? And you know, when copilot finally comes out, like what are going to be the implications to anything I've built home grown and you know, my playground. So there are a lot of things that we're helping people with kind of like in adjacent spaces, but high workflow, high requirement, generative AI use cases, you know, is still very firmly, you know, our territory. 

Phylise Banner: So let's go back to that “ask the right question”. Yeah. What does that mean to ask the right question? And, you know, this is one of the things that I love about working in the L&D and Edu space, is that everything revolves around curiosity. Other people talked about it as engagement. I talked about it as curiosity. Curiosity is what drives learning, right? Exploration, curiosity. So becoming curious, in this space, as I'm interacting with AI, how does that help me shape the right question? Like where are the great questions coming from? And what are we learning? And what do we need to teach AI? No,

May Habib: Let's and let's start very high level, I think, broadly and liberally, trained. Knowledge workers are going to take over at work. And the reason for that is the imagination and ability to bring in to what they're doing kind of multiple concepts is going to have this outsized impact. And let me give you an example. Like, let's say I am a very simple example. Let's say I'm coming up with, you know, 10 headlines, H ones for a bunch of landing pages. If I can direct writer to, you know, use David Ogilvy's principles for capturing a reader's attention, right, I don't even need to say any more than that, right writer is going to do that. But if I don't know who David Ogilvy is, or I actually don't know, I don't have a principled approach to copywriting. You know, I not going to know to ask that question. The person who does not ask that question is going to do a better job. And like, on and on, you know, the people who are widely read, who are I think, come from a liberal arts background are going to do very well here because like, asking the right question has been just part of right there training. I think it's going to be harder for people who are who have succeeded in their careers by being like, very narrowly technically focused, because that is going to be something that I think is just easier to commoditize, actually.

Phylise Banner: So part of my expertise and skills for the next century is asking the right question.

May Habib: Absolutely. Yeah, absolutely. And knowing enough about a space and adjacent spaces and having the creativity to imagine what could come next. I think is, is so necessary,

Phylise Banner: I think of artificial intelligence as being a companion. And it involves having a conversation of some sort of asking for something and being able to control that conversation and relationship. I want to talk about an experience I had just the other day. I again, use many tools in this space. I happen to be using Writer yay. Yes it is my favorite - so I am biased, and that's what I want to talk about - bias here. I asked for Writer to develop a scenario where a conflict arose between two co-workers. And it gave me a wonderful example - a short paragraph. And the names used were Sally and Pete. And that's not good enough for me.

May Habib: Yeah, absolutely. Yeah, absolutely.

Phylise Banner: And I immediately picture two white people. Yeah. And so what do we do in that space? And I, you know, I think about what was called and probably still is called human in the loop in this space. How do we train our intelligent agents to recognize and embrace diversity? Welcoming, belonging? Inclusion?

May Habib: Yeah, those are, those are great questions for us. And it is so subtle, right? Because especially at work, you know, a lot of the low hanging fruit on AI safety and bias we've addressed already, and this is just much more subtle, right? And in so many ways, insidious, I think the technical approaches to this are clear. They're not they're not complicated. It's just like, further down on the list, right? Unless a company says, Hey, Writer, this is really important to me. And like it comes to the top of our list for their for their own implementation and rollout.

So, you know, so much of this question is a, a, like leading the pack type of question, because everybody is still in, like utter chaos land when it comes to generative AI. And most people have not yet really left the sandbox with regards to large scale rollouts. And these are exactly the types of questions I have been waiting for folks to ask a question like this, I have been waiting for folks to ask for an API to actually be able to test our models for their own sets of bias and inclusivity requirements, right, and I'm excited because I know we're going to do great. This is a great example where we wouldn't have done great, but that, you know, we would be addressing, especially if it comes from a customer and in pretty, you know, short course. But most people are not there yet.

Phylise Banner: So, let's talk about writing in general, and the creativity involved in writing. And I love how you talked about the knowledge workers taking over, right? And I have a friend who also says, “You're not going to lose your job to AI, someone who uses AI is going to have your job if you don't jump on board”. So, what is going to be the role of the writer in the future? What does it mean to be a writer? And what does creativity in this space look like?

May Habib: I think the premium for like true human interaction, and writing that reflects real human interaction is, is just going to be so obvious. If it's not already with the naked eye, if you are not a skilled user of AI, the output is glaringly obvious, because it lacks statistics and quotes and personal stories and anecdotes. And that style of writing, I think, is going to be so much more appreciated not just by humans, but by algorithms. And so skilled writers who can turn conversations into insights and insights into breakthroughs and breakthroughs into narratives and all the things that writers are really great at. I don't think that job ever goes away. Just the tools that they use. The maybe rate of composing speeds up, but you can only have so many actual human conversations right in a day.

So, I think we are giving people tools and really that's it. I think a lot of this writing hype is going to die down I hope. And you know, AI will be very powerful. But really just another tool in the writer and storytellers tool belt I love that.

Phylise Banner: And, like, in this L&D space, what I'm seeing happen every day I get a new message about some AI tool being integrated into learning development platform. So now you can use AI in your LMS. And you can create, you know, it can automatically create your, you know, your outline or your structure for your content module, and maybe even analyze your audience and create learner personas and all the things that we do. So yes, it's making my job easier, I will always go back and check. Right, I will always check authenticity, and make sure that that narrative is in alignment with the voice of the organization, the voice of the institution, things like that. So, I'd love to hear from you, what excites you most about the future of AI? And also what you're most afraid of? If anything in this? Yeah.

May Habib: I'm really excited about where the idea of kind of multi-application AI goes. So you know, our ability to help Phylise you know, come up with a scenario and then also send a draft of that scenario, once you approve it to LMS. A, and an email asking somebody, what they think about that approach before, you know, you go in and publish. So I'm excited to speed up my own workflow. I work way too many hours in a day, even when I delegate so much. And I do delegate a shit ton. Oh, we have a very strong team. And yeah, I would love to get just as much done, but get it done in eight or nine hours, instead of, you know, 10, or 11, or 12, or 13 hours. thing that scares me is will people?

How quickly will people understand that sometimes the form of the form of the thing can really impact the function of the thing, i.e., most people are not, you know, first draft drafting inside of an LMS, right, they're starting in a Word doc, or they're starting energy. And when you look at a blank page, or a blank cursor, or even like your own, you know, a template that you had for something, but you are in an authoring environment versus a publishing environment, like was there a first principles approach that you brought to your thinking that is going to go away? If you're actually now looking at a first draft for the first time inside of a publishing environment, right, where kind of like your, your heart rates a little faster, like, you see those buttons right there, like that is a UI meant to push stuff out.

And so, I do wonder, and you know, L&D is a great example, you know, multiply this by like every function that is now going to be merging, authoring and distribution, right, as so many of these platforms kind of get on hop on this train and a lot of them in unthoughtful ways, right? Part of the reason this is like so fast everyone is doing this literally overnight is because it's a single API call writer is an LL NLP company, as much as we are an LLM company, which is why our output is so good, you know, really worries me that companies who have no idea what good content looks like, who aren't NLP nor MLM companies themselves, right, are putting first drafts in front of people who this might be their first L&D job, it might be your one for them, they might have just started at a company. And I do think we're going to see a lot of these platforms roll some of that back, just based on feedback based on what enterprises are seeing and doing. You know, we know from products that our customers use that some of their largest kind of enterprise customers of these products have asked for like an admin switch, I actually just want to be able to turn this off for everybody. 

So, you know, that we will work that out. It doesn't like you know, scare me so much that I would say nobody should be integrating these without experience etc. It sounds so self-serving to say that that's not what I mean. It's really that you know, both we, in our own practice in our own product, really try to help people understand like when you are in authoring mode, These are the types of things you want to be looking for. And when you're looking at AI, assisted content or generated content. And when you are in the publishing platforms, you want to be using AI to edit, right, not draft. Because mixing those two things, I do think impacts the quality.

Phylise Banner: I agree completely. And from an instructional design, learning experience design perspective, in that authoring phase, you really are creating an enormous narrative that has so much flow and taps in to, you know, different knowledge exchanges along the way in this pathway. And, you know, the number one rule of thumb alignment with learning outcomes alignment with learning outcomes, that's all well and good, but it's really leading towards evidence of learning. And where does that evidence lie. And for me, it you need to take that step back to have that holistic look.

And to see what that learner pathway that learner journey is going to be creation, authoring and publishing very, very different things very different places. being asked to author in a publishing platform is like nails on a chalkboard for me. So, I hear you, I get that. Thank you. Thanks for sharing that and stating it so clearly. I’d love to close by asking you for advice to anyone who's new to this, who is curious, who's afraid, who's just dipping their toe in the water? Who is maybe overwhelmed by everything that's out there?

May Habib: Yeah, definitely, I think the first piece of advice is, it's, it's honestly not going to change things as much as the hype, Truly, truly, because the standards are going to go up. And the only people who are going to be able to meet those standards are tough people. So like you, you have a skill set that like, no AI is going to match because it's just you are to multi-threaded into, like what makes for an incredible learning outcome for this to be replaced by AI. It's just the technology is not there, even the technology that is coming. Having said that, to your point of, you know, are the people who learn how to use AI going to be more impactful in their organizations, I would, I would say that is that is going to be true. And I think the best way to make sure you're one of those people is to go into experimental mode as you have. And really get a hands on feel for what the what the output is like with very vanilla prompting what the output is like with more sophisticated prompting, and we do want to do a prompting workshop that is focused on L&D and knowledge work.

So, I'd love to work with you on that Phylise. But once you are once you've integrated AI into kind of farming out for you, right, like little parts of that journey and that narrative and that arc that you've created yourself. I think it all becomes just a lot less, a lot less scary. And I think, you know, that's kind of the crawl walk, I think the run is then, you know, using everything that you have so perfectly, you know, expertly crafted, and turning that and transforming that into, you know, other pieces of collateral and deliverables, right. And especially there are so many people who are individual contributors or consultants in this space, right being able to clone themselves or give themselves leverage with AI, I think is a really powerful next step.

Phylise Banner: Thank you so much May. I appreciate you being here with me today, sharing your insights, your expertise, and for being a champion for innovation in this space.

May Habib: Thank you so much really so lovely to be here.