Now your product is better, your offering is better.

So maybe you'll get more customers and and that stuff's harder a little bit to evaluate and and this is like the promise of AI think it's like very narrow to just look at it only from a Casa a bank's perspective.

I really don't think people have got their heads around how much things are going to change with this technology.

I went to Bolt and I said just in my regular words in English, I said I want an app where I can cut and paste a chart into the app.

And so now I, I can just copy and paste that code and I get this really cool interactive chart, JS chart.

All your dreams have come true by offloading all that grunt work to the AI.

You can have a bigger team in a sense, or a more powerful team that that can produce more and make your company more competitive.

And welcome to the show.

Hey, thanks for having me on Aiden.

I'm really stoked to be here.

Thanks for for joining.

I know it's the end of the day for you today, but there's no time like the president to talk about AI, so I'm excited to have you on.

So lots of stuff to talk about.

I mean, this is the stuff that you basically live and breathe every day.

Lots of really, really great topics that we're going to get into.

But maybe in your own words, Ed, if you could tell everybody a little bit about yourself and the type of work that you do, so they kind of have a sense for where you're coming from as we're having this conversation.

I, I sort of fell into AI.

It's I'm not a highly technical person.

I just happened to kind of these two worlds collided for me.

So I was formerly head of product at online learning company and that company started to kind of shrink.

I got laid off and I spent about a year doing product consulting, but all anyone wanted to talk about was AI.

Everyone was trying to figure out like, how do we use this?

You know, it obviously looks very powerful, but what can we do with it?

And so I luckily it got attached to a few projects where people were actually pushing it and trying to create production apps.

So, so that's a big deal like production apps versus just building anything, right?

The bar is much, much higher when it comes to building a production app.

Yeah, it's so important because again, you see these shiny demos, it looks really cool.

Or you see a clip on YouTube or somewhere where it's like, yeah, this AI agent can talk just like a human and can do all these things, but you know what happens in production?

Yeah.

So, yeah, exactly.

So like, it's really easy to impress someone with a whiz bang demo, but then where the rubber hits the road is like, is this software reliable, consistent, accurate.

And as a subject matter expert or someone who's trying to do their job, does it help me do my job or does it make mistakes and get in the way and cause me more pain than I have that I already have?

What I saw in my consulting time was essentially like leadership making mistakes, leadership making bad decisions around which projects to go for, or they were handing off the responsibility of doing that work to a tech team that wasn't really aligned with the business.

So I started machine and partners as a way to offer wisdom essentially to management teams to say like, hey, like, let's investigate.

Like what would be a good AI use case for your company?

And maybe there's none, you know, that you have to sort of accept that maybe there there might be nothing worth building for yourself.

But it quickly became clear that we couldn't do this if we weren't actually building something.

So at first it was just like consulting and like, let's just talk about, you know, more product oriented and product strategy oriented.

But it was clear right away that we had to build something because the name of the game is really de risking.

How do you de risk the decision to build software?

And the bar for software building is getting lower.

You know, I think we're all kind of seeing that, but the standards for quality software is still the same.

And so we now what we basically machine and partners is essentially like a little R&D company and we help companies identify use cases for where AI might have an impact on their business.

We do do due diligence on those to make sure these are good use cases that they're actually going to have an ROI.

And then we investigate like how hard would it be to build this, you know, what roadblocks are we going to run into?

What difficulties?

Is this an ML project?

Is this a jet AI project?

You know, what's the best way to go about this?

And at the same time, we up level the existing team of that company.

So if they have engineers, which they usually do, then we're inviting them in and we, you know, we build together.

I think we have a really cool, interesting unique process for doing this.

And essentially that's what we're selling.

We're selling the process to de risk software investment in.

AII think one of the cool things about this is you also get to see what other companies are doing.

So you have this broad purview and I think you know every company's thinking about this, but it's nice to get an outside perspective, right?

Like even if it's a yeah, like what you're doing is great.

You're doing all the right things, don't worry.

What would you say in your in your personal work?

Like how are you working differently because of AI today?

Very broadly, just my mindset has changed.

Almost like every time I have to do anything, the first thing I asked myself is, can I offload this to AI?

Can AI help me with this?

Can AI, you know, be my partner in this?

So that wasn't always the case.

It took me a while to build that muscle or that instinct because I'm 50 years old and I've been doing things the same way for like my whole life.

And so it, it, it takes a little bit of effort to actually, you know, reorient myself towards an AI powered work style.

But so like, you know, it's like you could almost point at any single thing that I work on and I can show you an AI powered or AI augmented, you know, process.

So I, you know, just you asked this question, but off the top of my mind, there's so many things like I do a lot of writing.

So I have various tools that I use from chat, TPT and from Claude to help me write in a more structured, consistent way.

So this is both from the ideation side all the way to the, you know, final polishing side.

So I have, I have some tools for that as far as like back office automation.

We used to have ACRM, just classic, everybody's got ACRM, right?

But the CRM sucks.

It's got, it's got all the you have to work the way you it wants to work.

And we're relatively small.

We don't have you don't that's overkill for our small business.

And really what I wanted was other things.

So we started eating away at the CRM and taking over some of its functions and sort of automating that.

So I'm in the process now of building some automations with tools like N8N and gum loop andmake.com and Relay to build these agentic kind of automations that can that use LLMS in Rocesses along with third party software to that would normally be offloaded to the CRM.

So I would much rather pay these smaller, more discreet companies for doing the exact value that I want than paying the money to the CRM to aggregate it in a way that I don't want.

And then additionally, we just like I'm moving more towards building like what I would call Microsoftware, where if there's a job I need done, I don't look for a tool online anymore.

I the first question I asked myself is can I build this in Bolt or Replit?

Yeah, and that's a crazy thing.

If I can build it myself, that's a powerful notion.

Yeah, you, you were telling me about that.

What was that application that you recently built?

And I think you used bolts and for everybody listening, that's bolts dot new.

Yeah, Bolts dot new is essentially AI don't know.

It's like an accelerator app builder.

It it basically builds I think by default it builds like JavaScript.

I can't remember what the next JS framework with Shaad CN as AUI like it's kind of like these out-of-the-box defaults that are very normal.

You can tell it in in words what you want it to build.

You could say, I want an application that does XY and Z and then it will, it will just sit there and churn and build and you sort of watch it, it'll hit a snag and it'll say, oh, I, I, you know, I had an error.

Do you want me to fix it?

And you go, yeah, fix it.

And then it fixes itself and continues going.

And every once in a while you have to help it along because it it runs into a wall somewhere.

But it's amazing.

I mean, it's amazing what I could do.

So just the other day I had this use case where I write a lot of blog posts.

And one of the things I'll do is I'll go research for the content of the post.

And inevitably in doing that research, all all sort of run into a chart or a graph.

I'm like, oh, that's a really nice chart.

And So what I used to do was I want to recreate that chart rather than just cut and paste it and like copy it.

It won't look good in my blog and you know, as I want it to look good.

So I would go into Google Sheets and I would sort of recreate the chart meticulously by copying and pasting the data points and all this stuff.

And it is such a pain.

Yeah.

And I did a lot.

So it's at some point I've wised up again.

This is like my old mind, my old mindset versus my new mindset.

At some point, I sort of wise up and I said, why am I doing this?

So I would copy and paste the chart into ChatGPT and I would say, hey, ChatGPT, give me acsb of the data that is represented in this chart.

And then churn, churn, churn.

And they would spit out the file.

And then I would open the file in Google Sheets.

And then I would go through the same process of making the colours match my blog and all this stuff, right?

So I did that for a little bit and then I was playing with Bolt and I was like, hey, I think I could actually do better than that.

So I went to Bolt and I said just in my regular words in English, I said, I want an app where I can cut and paste a chart into the app.

And I want you to extract the data that created that chart.

I want you to then build a chart JS that I can embed in my website.

So and I want these this color palette and stuff like that.

So now I mean it works.

It's an, it took 10 minutes for Bolt to create this app with a little bit of prodding for me.

It needed my open AIAPI key and it, you know, it was getting confused around the sort of which model to use and, and stuff like that.

But it was very much very effortless as far as like what it would take to do this normally.

It's interesting, I've been using Bolt as well to build a bunch of things.

The other thing for, you know, for those out there who have kids, I've been using this with one of my kids, my, my 8 year old daughter.

And you know, the first thing I was like, what do you want to build?

I want to build a game.

What kind of game?

Let's build like an ice cream maker game.

And so we built this thing.

She's using the Mac key that where you can speak 'cause she's not great at typing just yet.

And so, yeah, she's asking for features and it's, it's just doing these things and, and it's crazy, you know, just from also building interest in kids as well.

I think this this stuff really helps with that.

Totally.

OK, I found it.

Let me just move it up so it's closer.

I have so many tabs open it.

That's another an old another old habit that I need to figure out how to break.

I'll share it.

I'll show the result.

Does that make sense?

Sure.

OK.

Can you see this?

Yeah.

OK.

So it's not much to look at the chart image converter.

Very.

I did not name that.

The AI named it.

I have to paste a chart.

So give me a second to find a chart that I can paste in here.

I don't know if this will open up in here.

Let's see.

I'm just going to pull some random chart off the Internet.

OK, here's a good one.

So I just click here and I click paste.

You don't see the chart that I posted, but I basically just took a screenshot from the web of a pie chart and you can see that it has the data.

Oh interesting.

So we're we're not seeing the actual chart, but this is like you pasted it and it immediately turned it into just something with data.

Yeah, it it it figured out what the data was.

I don't know if I can share another.

How do I, I'd have to switch tabs to show you the but, but just to keep going.

And then here's the predefined color palette.

It can go dark to light or light to dark, depending on how you want it.

So that's the color palette that I want for my blog because I have a dark blog.

And then here's the chart JS code and it includes the script tag and I can just copy and paste all this code.

So it's got a copy feature.

And you know, that's something that I added.

I added the color, the color palette stuff just by chatting with Bolt.

So, so essentially what it's got here is it's got all the code, the embed code that you need to, to put that into my blog.

So now instead of just having to go, I mean, you saw how fast that was, right?

That it took.

Yeah, it was instant.

Yeah.

And so now I I can just copy and paste that code and I get this really cool interactive chart JS chart instead of just an image that I would have normally posted in there.

Do you have like the the bolt history of like what kinds of prompts you you were sending it and did you have a sense for how long it took you to get this to the state that it is?

Yeah, I think it it probably took 10 or 15 minutes, I mean.

Yeah.

So that's pretty quick.

Yeah, Let me see if I can log in to Bolt.

And OK, so this is Bolt over here.

It's got the code.

So it's quite a bit of code to do all this stuff.

Let's see, you know, so here's the here's the components, it's built chart preview, color palette, etcetera.

And so I could go here if I wanted to update this like you, like you said, hey, it didn't show you when I paste the chart.

I don't know if this is going to work about two.

That's right.

Let's.

See show the chart I pasted on the screen as you as you are recreating and if it's it's you recreating it OK.

Yeah, so this is cool.

So you're chatting in and it just starting to modify the code.

This is great.

Yeah, so it says I'll modify the app component to show both the original pasted image and the recreated chart, like it knows what I'm talking about.

I'll update the necessary files, add this functionality and here it goes.

And you can see it's off to the races writing a bunch of code for this.

I mean, it's like this is the part that takes out.

And so it's quick, it's got better.

So let me just paste the same chart and see what happens.

All right, so there's the original chart.

That's cool.

So it shows both.

Yeah, so I mean that took what I.

I feel like a valued customer or something.

I asked for a feature and like you, built it on the spot during the call.

This is great.

Every software company should do this.

Isn't that crazy?

It, it's amazing to see like no errors, but obviously like for, you know, everybody listening when you start doing this, it, it can get frustrating.

There can be, it can get stuck like you said and but it but you have to be patient and and it's also being able to describe what you want really well and and and clearly too really helps.

I saw a, a conference recently or like a video of a conference of this guy.

I think it was the AI engineering conference or something.

And he was from, I can't remember what company he was from, but he had built like an like an AI agent that would dynamically create code for a, for a mail client.

So he built a mail client that had no code base.

He just said, when I ask here's my AP, here's my Google API key.

And when I asked to see my e-mail, you go show me my e-mail and it dynamically writes code to show his e-mail based only on his Gmail API key so that it would like sort of spit out it would take 3 minutes to show his e-mail.

I mean that's insane, right?

But the, but it's pretty crazy.

But that was like on demand code being written, that's an on demand application being written to show him his mail.

And then it knew like when it showed him his list of mail, he could click on anything in the list and it would go show him the actual mail and he could go back.

So it it knew enough to create this kind of basic application on the fly.

Yeah, that in your day today that that you'll be able to automate away like this.

I did want to also ask you about the, the writing one because I know that you use, I mean, like you said, you use Claude a lot for your writing.

You update your blog a lot.

1 of the common complaints I've heard is that if I use AI to help me with my writing, it sounds like AI and it doesn't sound like me.

And I have to do more work editing it than than writing it myself.

And so I think a lot of people just give up.

And so if you could show us like how you go about it and like what your process is or how it helps you.

Yeah, we, we talked about this.

So I, I went and I looked at the kind of the setup that I have on, on Claude and I thought it's, it would probably be cooler to like recreate it from scratch if people actually want to use a technique like this.

So rather than going in and sort of unpacking what I had already built, because I think like creating it from scratch will show you like how easy it is to, to get something like this up and running.

So oh that.

'D be amazing.

Yeah, I'd be excited.

Let's see it.

So you like Claude for writing versus ChatGPT.

It depends on the week.

Some days, yeah.

Yeah.

It's really funny how those two, you know, have their strengths and weaknesses, but I I'm, I have a hard time articulating exactly what they are.

Like, I think Claude, Claude, certainly for code writing seems really good.

And I think for like it's a certain kind of writing, like the kind of writing I do is like oriented towards business leaders and it's like sort of technical and it's businessy.

Claude seems to have a little bit more warmth and kind of gets my humor maybe I don't know.

Whereas ChatGPT, it can can sometimes be a little like either not talkative enough or sometimes it's overly verbose.

I don't know.

This isn't quite the tone doesn't always nail it for me.

But if I'm trying to learn something I really like learning on ChatGPT, I feel like it's a much better tutor so.

Yeah, it's interesting.

I I I have found the 4.5 model is better at writing.

Oh, whereas like versus 4 point O yeah, I I would agree that Claude's 3.7 model is, is especially like really, really good at writing.

I was going to say that Claude has this thing called project.

So if I go here, you know, there's project.

So if I'm just doing a chat, you know, a chat might happen here like you would expect.

But if you go to projects, then you can create a new project.

So I'm just going to say blog writing assistant.

I'm sorry you have to see me type in real time.

It's all good.

A ghostwriter I'm going to show really quickly.

Yeah, but see it.

The blog so you can get a sense so like here's a deep thought.

So this is like the blog and so.

Hallucinations are practical guide.

I like that title.

So, you know, it's kind of, it's a little bit wonky, but it also has kind of, you know, examples that are more concrete.

And I try and use a lot of analogies and I try to break things down a little bit more.

So like, it's just like my style of writing.

I like to explain things, especially to a business audience that maybe doesn't know this stuff very well.

So I'm going to stop sharing the switch to now, Claude.

So just give you a sense of kind of what the writing looks like.

OK, So what I'm going to do with this project, the project is my blog writing assistant.

So it's something I can reuse over and over again.

So I'm going to add some product project knowledge and I'll upload.

What I'm going to do is I'm going to upload a couple of articles that I've written.

So I have, oops, actually that article on hallucinations.

And there's another article about sort of why developers are still important and you need to have them.

And I'm going to add one more thing.

I'm going to add a style guide.

So I have a blog article style guide and I probably should share that.

I don't know if it'll let me open it.

Let's see.

Did you write the style guide yourself?

Yeah, so here it is.

It's really simple Blog article style guide Audience This is like a header.

The audience for this blog is a non-technical business leaders at various levels and across every industry.

Tone.

The tone for the articles is friendly.

I'm just getting a lot of guidelines for kind of like how I want these articles to be so I don't have to repeat myself.

You reference the the articles you uploaded in the style guide or that's.

Just sits solo yeah, so I'll show you that in a second.

So then the other part of claw, that's really cool.

So that's got set project instructions right here.

So I'm going to click this and I'm just going to paste in what I would put.

So I said you are a professional ghostwriter.

You are inquisitive and helpful.

Write in the style of the two attached articles to mimic my tone and style.

Follow the style guidelines to create an article on the given topic.

Don't start writing until you understand the theme and argument of the article.

Ask as many questions as you need to understand the theme and argument.

Oh, I love that.

So I'm giving it like this.

This is kind of its raison d'etre and those other things are examples of my writing style that it can mimic.

So this helps get away from that problem that you were solving where it's like, oh, it sounds too robotic or whatever.

So let's say I want to write an article about MCP and why it is the beginning of a profound.

Change to Internet infrastructure.

By the way, as you're writing this, you're using this and you know, for your blogs, but you could create something to help you write emails, you could create something to help you write LinkedIn posts.

You could.

Again, it's whatever you want.

As long as you have examples and you can instruct this clearly, you can use it for those things.

So, so I'm saying I want to write an article about MCP and was beginning of product.

So what I do is I'm a little bit different.

Like I don't actually at this point, I don't have it right.

The article I usually go through and say, let's create an outline together.

And that gives me more control over the structure because I'm much more picky about what's in the article.

I want to have like insights and I want to definitely want to get my opinions in there.

So I actually spend quite a bit of time at this phase putting the right words into this box to get out what I want.

And, and I usually start with an outline because the outline is the blank page problem gets me over that hump.

And I can see, oh, OK, I see we're going to be talking about this kind of stuff.

And, and then I can adjust the outline until I get it where I want.

Then I go, OK, now write the first draft, but just for the sake of time and stuff like let's just skip over that and go right to the article writing so it knows the length and it knows all that stuff.

Let's see how well it does here.

I'd see.

Maybe it'll ask me a question.

It probably should, yeah.

Cuz excellent.

I asked it to be inquisitive in those custom instructions to like make sure that it understood the theme in the argument that I was trying to make.

So it's now it's asking like 5 questions is a lot.

But but you know, actually this is, this is what I want.

So what specific aspects of ends?

I don't know if we should do all this, but what specific aspects?

Of yeah, you can basically say like use your best judgement and like go in this case.

Let me give it one thought because I do have a thought here.

I think that's pretty interesting that it chose to do that because you know, MCP is relatively new.

I'm not sure Claude actually has anything and it's trading data about or anything significant about MCP.

So I think it's really interesting that it actually chose to search the web.

That shows kind of some intelligence, right?

So here it goes.

Blah, buddy, blah writing.

I told it I wanted ATLDR at the beginning.

That summarizes the article and I didn't give it much to go on here, right?

So let's see what it says.

The Internet infrastructure revolution, nobody's talking about because I told it I wanted a title that was clever.

So it didn't just do like MCP, you know, and it's spot on.

I mean, that's true, right?

This is cool because this this is the kind of stuff that I would talk about because I like to, I like to bring up these technical things and say, hey, this, these are real things and they have implications and they have ramification on the world that we live in in late 2024.

So I got this obviously from its from its web research.

I could probably enhance this by saying if you do web research like pull data like create some charts that I can use if relevant, right, I could put that in the instructions.

So this is actually giving me some good.

Ideas, yeah, but this is, again, it's a really great start.

You can, you can obviously like pumped a bit more, make sure that the storytelling is good.

What is just like philosophically speaking, because this is a question that, you know, I'm sure people are asking, which is well, like, did you write this article or did AI write this article?

And So what do you say to the crowd that's thinking that?

Oh gosh, that's such AI mean.

It's a good question, but like this article would never have come to pass if I hadn't have done what we just did.

I give myself credit.

Thank you very much.

I think this is a bit of the, you know, the author instigator.

I think of AI as like a sandwich, right?

You, you are the instigator and you are the judge.

And the AI might create the meat in the sandwich.

That's kind of the bulk of it.

But without you instigating and judging whether it's good like it's nothing, it's not going to come up with it on its own.

It's a, it's a bunch of interns.

You can hire a bunch of PhD interns, doesn't mean that the work is going to be really good.

Like you said, you still have to orchestrate it and, and put it together and, and you own the output at the end of the day.

Like if it's, if it's really great, this article will be shared, you know, on the Internet, it'll get a lot of views.

The human judgement still matters.

I, I will be writing this article so, you know, in a couple weeks, if you look at my website and you see that you're going to be like, Oh, I know where that came from.

But, but you know, I, I do take the, I do take that seriously because I find that the AI can, is really good at kind of bulking up, right?

Like, but I need to give the skeleton and I need to do all the finishing touches.

However, I think it's pretty good at the general, you know, struck the the sort of argument and the flow actually better than I am because I tend to meander.

So the writing from the AI is actually much more structured and better.

And then I put my Polish and and I'll do a lot of rearranging post facto to make it like what I want it to be and to be to have my voice.

It just I can just do it so much faster with the tool at at my side.

True.

Thank you for showing this to us.

It's all about the use cases.

One, one thing that might be interesting is, is there anything that's stands out from the orgs that you've worked with, you know, where something was like very transformative or made a big change was a big win?

We have a client who is a, they're a data platform and they scrape government documents for, you know, data and these government doctrine, there's a set schema that they're trying to populate and they're trying to fill out this database.

And then their analysts use that database to then do their jobs to report to their clients and things like that.

The way that they do it or the way they did do it was that they had a team of 70 people in India who meticulously would read through these documents as they were published to pick out, you know, individual data points and stick it into the database.

Once they had built an elaborate form essentially to allow these folks to do it.

And they would triple check it to get the quality out of it.

Obviously, you know, that seems like low hanging fruit for AI, right?

And so it did look like an obvious use case.

And so we said, yeah, let's, let's attack it.

Let's see what we can do there.

We found out very quickly that those documents are very hard to parse.

They're really messy.

They're not clean.

They got the tables, charts, things that are hard to parse.

And so we built a a fairly sophisticated parsing engine that could go through and grab the data out.

And the engine was interesting because we had encapsulated expertise into the engine and the engine was comprised of a bunch of what we call extractors.

And the extractors had a little job, right, That the job is to find one part or one piece of the data or one kind of group of data in this big 150 page document.

And so there was a lot of smarts in that extractor.

And in order to get all the data out of a document, it might take 4050 extractors to get everything you you need.

And you this allowed us to then make those extractors better and better over time and and create a really, really high quality, high accuracy extraction system.

Do they not use humans anymore in this process?

They still use a human.

The the extraction process used to take six hours for each document, and now it takes one hour because we have a human that goes through and validates all the data points that the AI pulls out.

The AI is not perfect.

It's like 95% accuracy.

So it'll go through and find it presents the data to a human in kind of a step by step way.

And then they're able to go, Oh yeah, that's right, that's right, that's right.

Oh, wait a second, this one's wrong.

And then they can fix it.

And that way they maintain the high quality standard that they need.

But they went from six hours to one hour.

That's a huge transformation.

What's interesting though is, and you know, there's thousands of these documents that go through the system.

So you're talking about 6000 hours per per season, per filing season saved significant.

But the thing that is interesting is what what happened was when we did this work, we realized that our extractors could pull much more out of the document than just these data points.

We could pull semantic information.

We could build a knowledge graph of the people that are in the documents.

We could have knowledge that spans multiple documents.

You could look at a a company's filing over time so you can do all this new stuff that wasn't possible with the old system and in the ability to store that information in a new way, then suddenly it has an impact on the analysts who are looking for that information.

So now the value creation, right?

So this is what we're focused on.

The company sort of charges internally $3000 per data poll to get the original set of data that they use in their reports, but they charge their clients like $50,000 for this, for this report.

So the real value creation is on the back in, you know, it's like it's at the end of the process when you're delivering this.

And so you know, going from $3000 to whatever 1/6 of that is, you know, that's not, that's significant times thousands, right?

But if you look at the place where the $50,000 report is created, can we make the generation that report go from one week to one day?

And we think we're working on that now.

And we think we can do that by building kind of like a Python notebook for report building for these analysts, right?

So that's, that is radical transformation if you think about it both from the cost point of view and the efficiency point of view, but also like how are they going to charge for this?

Because right now they charge on the back of the analyst.

The analyst takes a week, it's 40 hours to generate a report.

So if we can do it in eight, then suddenly 4/5 of their billing time evaporates, so now they have to start charging by value or figure out some other way to evaluate their work.

And this is stuff that everybody's thinking about, I mean, accountants, lawyers, everybody's thinking about everything will go to value based.

It's going to have to, I don't see any other way around it.

But it is really interesting.

And you know, the example that you gave just made me remember like this recent realization, which is that it's not only about time saved.

I think, you know, when we're trying to evaluate these things, a lot of times, you know, the thing that comes to mind is efficiency, efficiency, efficiency, time saved, you know, less people, less departments like these sorts of things.

But then there's the additional part which you just talked about, which is while we're at it, we're able to get all this extra data points and sure, maybe you'll be able to create that report faster to create the new type of report that you're able to create.

Maybe it would have taken you like a month to do it.

You wouldn't have done it because the client wouldn't have paid that much of an analyst time.

But now you can give that one month report report like in a week report.

And and so now your product is better, your offering is better.

So maybe you'll get more customers and and that stuff's, you know, harder a little bit to evaluate.

And this is like I think the promise of AII think it's like very narrow to just look at it only from a cost savings perspective.

100 percent, 100%.

I, I don't, I really don't think people have got their heads around how much things are going to change with this technology.

And that's one way.

And another way that people overlook is those analysts, you know, before what they would have to do is go to raw documents, go to spreadsheets, do this sort of arduous, meticulous work to highlight things in cells and all this stuff, right?

It really was not a pleasant job.

And I can't tell you how many times we, when we were doing our product research, we would talk to the analysts and say, you know, show me what you're doing now.

And they're like, oh, yeah.

And we're like, how long does that take?

And they're like, Oh, you know, I could do it in a day, but that would be a really bad day, right?

And I'm like, oh wow, can we make your day better?

And so now the enjoyment of the work goes up, right?

The, the quality goes up, the enjoyment goes up.

So you get like increase in pride, increase in happiness.

And you know, that's not nothing.

And maybe I'm woo woo about it or whatever, but I think that goes, you know, that that leads to retention.

It leads to higher quality work like you said.

And so, yes, you know, the dreamy part of me goes like, oh, yeah, we're going to make everyone's lives, like, better and happier and, you know, higher quality and more pride and stuff.

But then there is also the toll, right?

Because if you can get this work done that much more efficiently than you can let people go, you can do more with less.

And I think business leaders are going to have to wrestle with us around do I let people go or do I make the throughput higher?

Do I do all the Do I let my dreams come true of working like more strongly?

I think AI raises the floor or the bar.

So I just think that what's going to happen is people are going to assume more quality.

I think the throughput will go higher and we'll just expect higher things.

I mean, even from today, when, when, when you interact with like really great applications on the web or, or software, like your quality bar goes up.

And if if you go to a site that loads slow or it's a little like clunky, I mean, you'll, you'll write it off instantly, right?

Whereas maybe like 10 years ago that wouldn't have been the case.

So people are going to want that feeling of like, wow, it looks like you spent a year on this.

How did you do this in a week?

It'll be there.

So thank you for explaining that.

I think, you know, the next question we always ask is how has AI impacted your company?

And I think you started to explain this text extractor thing that you built.

One of the things you were telling me before is that's one of the things that you know recently was able to impact like your ability to do things within the company to do you want to maybe tell that story?

Yeah, it's a, it's a funny story because we have product managers.

They're they're not programmers, right?

They're traditional product managers.

They're and on our side, they're much more like business oriented and kind of like AI potential oriented, right.

But they're not super technical.

And then we have our developers like as, as one does.

So I have this one product manager and he was like, Hey, I'm, I'm tired of trying to explain everything that I want to happen to the developer.

So he started to use these like no code tools to build his own prototypes.

And he was showing the developer and the developers like, how did you do this?

And so he was like, oh, I just, you know, I'm using cursor and I'm putting these instructions.

I'm using a thinking model and he showed his whole technique.

So our developers have been using cursor and AI powered tools, but they just had never thought about using it this way.

So one of our senior developers, John, he's like, oh, I could use this to build more extractors.

So we, you know, the, the way you build an extractor is pretty sets.

We had a few extractors that we had built.

It's, it is time consuming to build the extractor.

And so we thought we were going to have to build like 30 or 40 of these extractors and it was going to take us literally like months, right to get those out the door.

And So what John did, he's like, here's a here's some extractors I built to the, to the the AI coding assistant inside cursor of clawed 33.7 is on it.

And he's like, here's some extractors I built.

Now I want you to build a new extractor for this purpose using that model.

And it spit one out like instantly.

And he was like his you could just see his like eyes light up and he's like, Oh my God.

And then he spent the next day building, I think he built like 20 or 25 extractors in one day like that.

We, we would have thought that would have taken weeks, you know, on our old estimate and was able to then hand off those extractors to other people in the team who are junior developers to start testing them.

And, and like, you know, increasing the quality, climbing the hill on the, on the quality that they're producing.

So that's.

Mind that's huge.

It's mind blowing.

There's two things here.

I know, I know for the audience, there's a lot of leaders in the audience.

I I think there's two things.

One is that by just encouraging AI across the board, you know, interesting cross pollination happens, right?

Like in this case, you wouldn't have thought that because of the the non-technical PM would have inspired like the technical folks to figure out how to use this for software.

You you wouldn't have thought that, but this kind of stuff happens all the time.

So everybody should be using something and and sharing it.

And then, yeah, I mean, it's insane to think that everybody has repetitive sorts of software that they need to write, whether it's documentation or whether it's tasks or, you know, I think about it, or whether it's integrations.

If you build 1 integration and there's like 5 that you could do this kind of stuff.

But it's amazing to see that you actually did that in practice and save that kind of time.

I mean, The thing is like, I think it's interesting.

I I have this conversation with with non-technical business leaders all the time.

They're like, oh great, I get to fire all my developers, right?

AI is going to write all the code and it that's a completely wrong view of this.

Like the truth is that our developer was critical to thinking about how to architect this system.

It's just that building the remaining extractors would have been kind of like a bore.

So the naivete of our product manager in trying something that he didn't know what would work.

And that was going against what our developer who was like, no, he had a bias, a sort of ego bias around like, no, I have to be writing all the code.

I have to be making the decisions, right?

But when he's not dumb.

So when he saw that Chase, our product manager had kind of created the system, he he immediately realized the potential and said, Oh, I don't have to do that boring work.

I can offload it to an AI and I can manage the AI.

So from the leaders perspective, it's like that never would have happened without the great developer that we have that the whole system that would have never been architected.

The AI could have never architected the incredible system that he built, right?

So now I think as a leader, you have to think you've been spending the last five years wishing you had more money to hide, have a bigger development team, right?

You've been wishing that you could get the features out faster.

You've been wishing you could squash more bugs.

You've been wishing you could refactor that code like everyone's been complaining about.

But you can't because you're a resource constrained.

And so, so now all your dreams have come true by offloading all that grunt work to the AI.

You can have like a bigger team in a sense or a more powerful team that that can produce more and make your company more competitive.

And I think the other nice thing is like, as you do this stuff, then, you know, other people will get ideas.

They'll say like, whoa, you did that here, Maybe I can do this there, right?

And it's in the beginning, it can feel like you're, you're pushing, you know, boulder up a hill, but at some point, you know, it's pushing it down the hill And so you can get there too.

So that's a really cool example.

2 examples, both on the text extractor and then obviously, you know, you creating the prototype from from the PMS.

So as we wrap up here and if people want want to get in touch with you, what's the best way to do that?

I mean, following me on LinkedIn is a really good way or connecting with me on LinkedIn.

I'm, I'm, I love connecting with folks and, you know, learning from other people.

So that's a good path.

My website is machine andpartner-s.com.

That's a good way to find me and see what I'm thinking.

Other than that, I think that's probably it.

I, I also teach courses at section school.com, which is like a, a business where I used to work and I teach like how to prompt and advanced prompting techniques and things like that.

So if you're kind of a beginner and you want to learn the basics of AI and stuff that, that could be a good place to go as well.

And we'll obviously include all of those in the show notes.

And, and for listeners, if you sign up for thisnewway.com newsletter, we're going to give you step by step instructions to do the things that Ed showed us today.

So if you didn't take notes that that that's where you can get the content.

And Ed, final question, what are you most excited about over the next 18 months?

Everything is moving so fast.

It's like being I'm old, so I was around in 1996 when the web started emerging and I get these echoes all the time of that time period.

Back then, nobody knew what a website was supposed to be.

Nobody knew what ecommerce was, right?

And then the mobile revolution happened.

Nobody knew what an app was supposed to be like.

So these are amazing, wonderful times to live in because we get to build that stuff.

We get to decide what that world is going to look like, right?

And that I think that comes with like a lot of potential, but a lot of responsibility as well for us to be managing those things.

So I, I really am excited by just the general innovation that's happening across the board, both from the.

The base models, the foundational models, but also I think more interestingly, what everyone is choosing to do with these things.

I'm inspired everyday by people like you and other people who are building companies around this technology to exploit things and create things that we really literally were not possible two years ago.

That's just kind of a general answer, but I think that's what really Stokes me.

Yeah, I love it.

I mean, we're going to build the future.

This is awesome, Ed Thank you so much for coming on the show.

Thanks for having me, it's an honor to be on.

I can't wait to listen to this one and catch up on some of the old ones I haven't heard yet.