Form factors for your new AI coworkers — Craig Wattrus, Flatfile

00:00:00.000 | I think we've all noticed tools like vZero getting pretty good at generative UI and creating

00:00:21.240 | good-looking things as well as Claude code being able to let us run things more complicated locally

00:00:28.920 | and build on those things so I think the the thing that comes out of this is designers product people

00:00:36.180 | and engineers all building together and I'm really excited about that because I've never loved the

00:00:39.960 | the divides and between these things so this really lets us get rid of in my mind get rid of

00:00:47.280 | mock-ups get rid of the click-through prototypes and all the hand-wringing about whether the thing

00:00:54.420 | that we're building is worth the engineering effort so as we as we go into this it's time for us to

00:01:00.900 | jump in and feel the material that we're working with and see what emerges so I'll give you a super

00:01:06.300 | quick overview of flat files AI stack this is not an official diagram but it's how I see it and we

00:01:13.740 | migrate data big if you need to move a lot of data between systems frequently you use our developer

00:01:21.420 | platform and since we're a developer platform LMs are good at writing code makes it the perfect place

00:01:26.100 | for a lot of AI at the bottom here we have our customers flat file applications that they deploy to

00:01:33.060 | our infrastructure then there's this like real-time context which is the data and the validation outcome

00:01:39.360 | so what are the errors and warnings and things that are in that that data that dirty data and then our

00:01:44.820 | AI agents the tools they have and the jobs that they can run and then what gets shown to users so I see it

00:01:51.960 | as four buckets here there's more there's invisible so it's kind of like the ghost in the machine almost

00:01:57.840 | called called that ghost ambient so it's kind of happening in the space but you're not directly

00:02:03.840 | working with it in line so it's actually in your work in your workflow and then conversational the ones

00:02:09.960 | that were I guess all arguing about I think that's what I learned being here at this conference here's

00:02:16.560 | an example of invisible so when you start if you sign up for a flat file we go in the background we take

00:02:22.740 | your email address you find the company you work for we look it up and in the background the AI agents

00:02:27.960 | are writing a flat file application so they're writing code and essentially sending you up a demo

00:02:33.060 | that is perfect for your use case so if you come in you're from an HR company you're gonna get an HR

00:02:38.340 | demo and while that's running you don't need to know that AI is working on it so that I'd say is like

00:02:44.340 | it's working in the background here's something working more ambiently it's a very initial take on

00:02:49.800 | this but you can see there's something an agent analyzing the data in the background this is a tool

00:02:55.740 | actually I lead this team for AI transformation and you can see the little sparkles pop up on the columns

00:03:03.180 | when it finds opportunities to fix it so that's ambient this is in line so you're busy working in the data

00:03:09.900 | and the AI is able you're able to use the AI I'm directly in line here to fix the data these agents are

00:03:17.800 | writing code that then gets run on this data set so you could have a million rows and 50 columns or

00:03:23.740 | whatever you want and that code will run really fast which is pretty cool and then finally the

00:03:30.340 | conversational ones we're all used to so this is build mode it's the no code low code agentic system

00:03:37.180 | that writes flat file apps now so before you would probably have to have had a engineer at the company

00:03:43.120 | building these applications now it can all be built up so that's pretty cool

00:03:49.060 | um and that's that's kind of the the general surfaces I think about um I listened to um Amanda

00:03:56.060 | Askell um from Anthropic talking to Lex Friedman about um building um Claude's character and in that moment I

00:04:02.940 | realized I'd been doing something a little silly I'd been giving engineers feedback on our agents like oh it

00:04:08.940 | shouldn't start saying this and it shouldn't use these words and and why should it do this and I realized I was I was doing

00:04:14.880 | like I would do design copy right I was I was I'm in my my my normal instinct and when I heard her talk I realized I needed to go from

00:04:23.820 | controlling um to being a character coach and and and actually building out um the the nature that I wanted

00:04:30.820 | so this is a v0 I hope have you most of you used v0 from Vercel before um yeah um so this is a v0 I built one of my early ones and it was um I

00:04:42.860 | called it a chat tuner it doesn't look like much but that wasn't the focus um but I could essentially um put our

00:04:48.540 | orchestrators so the system prompt for our ai orchestrator for build mode um in here and then I can modify it I can say what is it

00:04:55.980 | like if I tell Claude to be more friendly versus more balanced versus more concise what what does more cautious mean to

00:05:04.020 | this model um and the point of me showing this is just to say like the design of the final thing is always a

00:05:12.020 | tempting thing to design to um but now we can actually go and build tools to help us to design that um and this

00:05:20.180 | brings me to like uh I have like three themes um the first theme which is feeling the material I'm a

00:05:25.740 | woodworker so you'll have to forgive the analogies to physical material but if you're going to uh design

00:05:31.660 | something with a physical material you have to feel it right you have to what are the properties of it

00:05:36.700 | and you need to understand it and so I feel like before with design we were kind of looking at everything

00:05:43.260 | through like layers right mock-ups and prototypes and and kind of trying to see what was going to work

00:05:48.940 | and what wasn't what we need to do now is go feel the material feel feel how these models work

00:05:55.500 | my new north star is like creating an environment for these llms to shine right what's what's this form

00:06:03.260 | factor that can help them nail their assignment stay aligned and grow as the models get better

00:06:09.420 | right that's that's my new goal um we're basically anything we do with an llm I feel like we're putting

00:06:15.180 | it in a box um and that's you also hear people say that llms are like interns like oh it's an intern with

00:06:22.540 | a phd and so I try think now if you're putting an intern with a phd in a box like it better be a good

00:06:28.860 | box um and so we need to put effort in um this was a conversation we were having about what tools does this

00:06:36.700 | uh co-worker this new form factor this new model like what tools do we give it when it shows up for work

00:06:42.140 | um and I got fixated on this idea of cursors I was like oh what happens if we just had a mouse or a

00:06:47.660 | trackpad I'm a trackpad person um so that's probably controversial but uh essentially what happens if we

00:06:53.740 | gave the AI um those tools and so I I created this v0 and they moved it into cursor um and I was like well

00:07:01.900 | I work in design tools a lot so I don't migrate a lot of data so this is the best place for me to feel this

00:07:07.500 | right to feel this material so I created a canvas um and I could give it orders and be like hey and

00:07:13.580 | honestly I was I was very enthusiastic about this um for like a few seconds it felt like I was touching

00:07:19.820 | the AGI a little bit but I also very quickly started feeling like I was putting a Formula One driver in a

00:07:25.740 | Prius it just it felt like I was constraining it and controlling it um it could only move one thing at a

00:07:31.900 | time um but so so learning from that um was uh something like this was also um a v0 um that I

00:07:41.420 | used Claude code on eventually and this is a new uh product that we're working on which brings like the

00:07:46.780 | all the stuff we've learned about uh migrating data to consumers to let them work on their data

00:07:53.020 | um but you can see the AI is is operating in the space um and it's it's got presence and so it's

00:07:59.820 | it's able to read multiple files while writing into another one um it's not like me who can only focus

00:08:05.740 | on one thing at a time even though I think I can focus on more um it's not true and so this is us moving

00:08:12.860 | from determinism to inference and figuring out what this material feels like and so um that's feeling

00:08:21.660 | the material right like working with the model getting it into your space understanding how it feels

00:08:26.940 | to work alongside of what's it capable of and then the form factors that we're putting on them actually

00:08:32.300 | actually you can now go build it and play with it um and feel it um the next material analogy I have

00:08:38.860 | which is finding the grain once you've got the characteristics of the material you understand it

00:08:43.660 | usually the piece of material that you're building with and you're creating with might have its own

00:08:48.620 | characteristics and so as we're creating these form factors uh finding the grain is about feeling it

00:08:53.980 | out where is it smooth and rough um where is it weak where is it strong um and we'll have to remain

00:09:00.540 | humble here because um things are going to change and are changing so quickly that whatever we

00:09:06.140 | rebuild is going to most likely need to be rebuilt this was an example of that build mode agent I asked

00:09:12.540 | it to do one thing which was enable the auto map plugin so it's just automatically maps data from

00:09:18.140 | the source data to the target data and I get a wall of text and it's not bad because this went and I

00:09:25.500 | saved me probably a week of work um I didn't have to have a product manager write a prd send it to an engineer

00:09:30.460 | get the in the road map get the engineer to write it qa this was all just done right all that code was

00:09:36.380 | written but the noise gets in the way and so this was a v0 of of kind of rethinking the tool ux what could

00:09:44.460 | it be like and so um the way I thought about this was if you're designing for a if you're if you're going

00:09:50.860 | to a co-worker and you're going to do something complicated for them and you want to communicate you

00:09:54.940 | think okay I'm going to choose my words carefully I'm going to communicate visually I'm going to stop

00:10:00.220 | and check whether whether it's right and so I wanted this to feel similar and so you can see here split

00:10:06.300 | personal details it's visually telling you what's doing saying hey is this right then it's saying I'm

00:10:11.820 | aligned I took a snapshot you can roll back I'm holding you accountable you approved this and then

00:10:17.820 | telling you what you can do next we also wanted to it to be able to express itself so if something went

00:10:22.780 | wrong and I'm shaking its head and a little bit of frustration which is probably what the user is

00:10:27.180 | feeling too when something goes wrong and then finally it can back off when it gets something wrong

00:10:35.260 | and sort of say okay I'm handing control back over to you and that's a lot more intro that feels a lot

00:10:41.580 | better and it felt like we had found the grain and found the right place to put this this material

00:10:48.780 | with this and so what's really cool about this one is that as we're implementing it we've realized that

00:10:54.220 | it can you can fit in other places so not just in conversational flow it can fit in line and this

00:11:02.700 | is going to be in our kind of like inline transform functionality really soon

00:11:09.420 | so I think like as we as we find a new technology and work with it we run the risk of just automating

00:11:15.980 | the tedious things and I was so excited about those previous two talks because there's kind of like

00:11:20.220 | some emergence in there right like something interesting that we wouldn't be able to do before

00:11:24.220 | and I'm most excited about those things like what what emerges from from playing we stopped playing for a

00:11:32.540 | few years um when we were kind of got the internet and we were like really excited and css3 came out

00:11:37.900 | and then like html5 we were playing a lot um now I feel like we're all playing again and so that's really

00:11:43.900 | exciting for me um this is an example of me playing um I created this v0 and I I we've been in search of this

00:11:53.340 | characteristic of an agent that is that feels forward leaning and what I mean by that is it's an agent that's

00:11:59.740 | curious and it's excitable um but it likes getting done um and it's very focused um so not going crazy

00:12:07.100 | right like we've all seen the lms kind of go too far when you give it a task and that doesn't feel good

00:12:12.540 | um so here I dropped a json file and a csv file um and the agent decided um you know what would be good to

00:12:20.060 | do is combine those two things because the data look pretty similar um and so here we can see it's it's

00:12:26.540 | combined the the file the two files into one um that's a good thing that it did I didn't have to

00:12:32.220 | ask it to do that um it picked up on it um and then after that it wrote a report so it told us what it

00:12:39.420 | was doing said hey I found some duplicates this is probably what you need to do next and so it built up

00:12:44.700 | context um and I was actually just trying to play with Claude 4 here and feel the material and kind of see how

00:12:50.940 | it would be but I realized um I'd kind of come across this nature that we were after um it made some

00:12:58.860 | suggestions and generated slide deck which I I asked it for um so within just dropping two files um it's

00:13:05.660 | it's done something emergent um and now we're baking this into our our new product called obvious which

00:13:13.820 | is coming soon another one was we had this idea of giving our agents a knowledge base so all the

00:13:20.220 | customer calls we'd had with them were all recorded and transcribed like most of ours are um and we had

00:13:25.580 | documentation from the customer and so we put it in into knowledge base and then when we analyzed all of

00:13:31.500 | this customer data we surfaced up um suggestions based off that I was fully expecting better

00:13:36.780 | suggestions I was fully expecting more suggestions got those but then here um the the agent decided

00:13:44.620 | I can't fix this but I know how to fix it and so I'm going to tell you how to fix it and so it suggests

00:13:49.740 | here that the user actually goes to HR and gets them to generate the missing employee IDs and what emerged

00:13:56.140 | here was something I wasn't expecting maybe you look at this and say that makes a lot of obvious sense but

00:14:01.180 | to me I wasn't expecting it to be able to help the human to go and do the job um where it couldn't

00:14:06.460 | um so that was really exciting I don't think I would be able to get to that without um playing and and

00:14:12.380 | being curious um and then the last thing I want to talk a little bit about is eyes on the future and we

00:14:19.420 | all have our eyes on the future because how can you not there's always something new now um with models um

00:14:26.620 | so I I like to think about it as like what's your pelican on a bicycle um and one of my pelican on a

00:14:33.580 | bicycles is order complete I'm super excited about it's probably a bad idea um actually to use an LLM for

00:14:38.940 | this but I'm like I want to make an order complete that is backed by an LLM and so this one has a hundred

00:14:44.940 | suggestions for fixing some data um and it's kind of like a bake-off um between these two things I'm yet to

00:14:51.260 | find um a model that is both very fast and very good at this problem um but this is a a benchmark

00:14:59.420 | or something that I've created just for myself to be able to feel the materials um that we're getting

00:15:04.620 | and so I think about that for my design practice now like what are the things I care about and can I

00:15:09.740 | like design into the future and start to think about the form factors I want and then build an

00:15:15.100 | application that can actually test that so yeah that's all I have for you today I I'm very excited to

00:15:21.820 | see all the new form factors um that we build um with our new tools thank you

Form factors for your new AI coworkers — Craig Wattrus, Flatfile

Chapters