From Playtime to Production: How Every Builds AI That Teams Actually Use

Today we're talking with Natalia Cano,

head of consulting at Every, where she

helps companies like private equity

firms, hedge funds, and media

organizations move beyond surface level

AI efforts and embedded the technology

directly into daily workflows. Drawing

on her background across finance, public

innovation, and venture, Natalia brings

a systems level view to AI adoption, one

grounded in how work actually gets done.

Natalia, welcome to Just Curious.

>> Thanks, Sue. Great to be here. What is

Every and what does it mean to be a

multimodal media company that also

consults on AI?

>> Yeah, so EverY uh Every started out

about 5 years ago. Uh and we do three

things. We do we write about AI. We

build AI products and then we share the

best things that we learn across those

two dimensions with other clients uh

through our consulting business. And the

first part of every uh is the media

company. That's how we started out. We

um we write about AI and really about

how people are using it, how companies

are using it and what are the most

practical valuable ways in which people

are

getting value out of AI. From there we

evolved to building AI native companies.

We wanted to really explore where rubber

hits the road and how you could be using

these really powerful tools to build

companies and so we decided to go and

build our own kind of product studio. We

now have five companies that have that

we've incubated in house one of which is

fully spun out and we take all of the

learnings as I mentioned and we share

them with you know variety of clients

that come to us and are ready to kind of

go on their AI adoption journey. So

that's what it means to be at every. It

means to be at the forefront of AI and

to be really excited to help other

companies and people explore how AI can

be helpful to them.

>> I want to follow up on that last point.

Um you said that you are head of

consulting, but every aren't

consultants, they're practitioners. What

does that look like in practice when

you're helping your clients adopt AI?

>> I mean it it means that we don't uh AI

is all we do. It's every day. Everything

we do is testing new models. All of the

internal kind of work and processes that

we have at every we operate in an AI

first way. So for example, a few months

ago, we decided to hire an AI ops

manager and this is someone who just

ensures that we are using AI in the most

compelling and valuable ways internally.

So uh AI is really all we do. When we

say that we're not consultants, we're

practitioners. It means that we are

using these tools at the cutting edge in

much the same way that we we want to see

our clients using those tools. So

consulting is not it's not sort of like

a theoretical exercise for us. We rely

on AI to build and run the companies

that we operate via our consulting stu I

mean our product studio. And so

consulting is just an extension of what

we do every day in running the business

that we rely on.

>> Yeah. when it comes to adoption of AI,

you've said that most companies don't

lack awareness of that need. They lack,

I guess, application. What does that

mean?

>> Yeah, I mean, it really just means that

I think we all could use a little bit

more efficiency and AI has this really

great promise of being helpful in that

particular end and helping us do more or

do better work. And I think where both

individuals and companies struggle is in

really figuring out what exactly that

means to them and their workflows. So,

what we're able to do when we're working

with teams or individuals is we're able

to pinpoint all of the areas where they

could be using AI and then really kind

of hone in on the areas where it would

be most valuable for them and for us to

focus on and to develop tools um and

workflows that they can get real

leverage out of. Right? It doesn't

really matter if you have great

prompting skills or if your projects are

structured in a really beautiful way. if

you're not alleviating pain points that

enable you to do the work that you want

to do, um you're, you know, it's not

really valuable. So, we just want you to

do more of what you care about and to do

it better and faster and for AI to do

that for you.

>> And what does that process look like

when you are helping them identify those

problems that can be solved? Um, where

do you typically start?

>> We started in discovery. So in order for

AI to be helpful to anyone

it that we need a lot of context and AI

needs a lot of context to understand

what it is that you're trying to achieve

and how you define excellence and how

you would go about achieving that

particular thing. So we start by

learning about what your goals are as an

organization right like what is the

north star that you're collectively

working towards. We then partner with

the teams that we're going to be

training and then really ask them how

they go about achieving their collective

goal that ladders up to that broader

goal. Uh and then really understanding

what are kind of like the core workflows

that they rely on to reach that goal and

within those workflows where there are

opportunities for AI to be helpful. And

so the first phase of our work is just

really learning and really deeply

understanding that any work that we will

do is one going to be oriented around a

goal and a priority that we know the

team cares about and it's valuable to

them. uh and two that we are going to

focus on AI skills, adoptions, tooling

that is going to give them direct value

in solving for something that they're

already spending time on and is a pain

point to them because you know this is

something that they're doing regularly

every day, every month, every week and

it's it's a painful sort of it's a pain

point, right? It's it's painful sort of

uh time expenditure.

So we start out all projects by just

kind of capturing that information and

then building essentially an AI road map

of here are all of the opportunities

that we surfaced that for which AI could

be meaningful and relevant. Uh and then

we work with our clients to decide which

of these do you want to prioritize and

pursue. And based on that conversation,

we then build out essentially kind of

like a training plan where we upskill

each of the teams that we're working

with to solve the specific pain points

that we identified in the conversations

that we had and do so via usually uh the

LLM that they're working with or via

kind of like the workflows that they're

solving. Uh,

>> do organizations usually come to you

with a desire to train their business

unit or their functional unit or their

organization or do they come with a

problem that they want to solve that you

help them solve while also training

their organization to use the

technology?

>> That's an interesting question.

I would say most companies

come to us looking to maybe build

something. I would say maybe there's

there's sort of two answers. There's

companies that come to us and and they

say, you know, we we want our team to be

using AI um and so we want to build

something, you know, or uh or or sort of

have some sort of like interface that

solves that solves like a pain point for

them. And then we also have teams that

say, "Hey, we just rolled out, you know,

sort of anthropic or claude or chachi BT

to the entire organization and we have a

sense that people aren't actually using

it really well." And

so those tend to be kind of like the two

broader conversations that we're having.

I would say for the first type of

conversation, what we see is there's

there's this kind of big challenge with

AI that there's so many things that you

can do. you really could build any

product or solve any pain point now um

faster than ever, but there's this huge

risk in

building something that you actually

don't need and doing a lot of that. And

I think we we really try to recommend to

clients um or anyone that we speak to

that before you go about trying to build

a product that you're not sure is really

the pain point that would be most

valuable to address, you really outline

where there is both lowhanging fruit and

high impact AI opportunities and then

seeing how your team uses the tools that

are available to them and where they are

actually getting value out of those

tools.

decide where to go next. So, you know,

that's that's kind of our response to

the first type of conversation that we

have. And then for the second, you know,

where there's like the classic like we

rolled out AI tooling to everyone at the

company. Um, I would say there's still

sort of like a misunderstanding on the

level of depth that you have to go into

to get value out of AI. Um, you know,

we've all had like a magical experience

with AI and I think we've also als all

all had a moment where we've asked it to

do something and it's just totally

sucked. Uh, and the reality about

getting value out of AI is that it takes

quite a bit of effort. It takes quite a

lot of effort actually to create a a

prompt and uh and and and kind of like a

project structure that solves a major

pain point for you. And so in order for

us to do that effectively, um, we're

going to need to work together and get a

lot of context and files and documents

and structure that into a really big

sort of like super prompt that solves a

painoint and we're going to like it's

going to be iterative, right? We're

going to try that a few different ways

and when we find something that works,

then we're going to have something that

the team can really rely on. But there's

this misconception with AI that you can

kind of just roll out TATBT to everyone

on your team and all of a sudden there's

going to be these like massive

efficiencies that are gained when the

reality is that you actually kind of

have to build out the skill of using AI

essentially train the AI agents or the

prompts that you want to use to have

sufficient context and instructions and

information about what you want to get

done and what excellence means for your

team. And then there's an iterative

process in testing those tools and

making sure that they are effective at

accomplishing those goals. And so we we

work with those clients to kind of make

sure that we can go through that process

and that they really understand that by

adopting AI, it's not kind of like a new

tech tool that everyone has access to,

but it's it's essentially kind of a a a

totally transformative change where now

everyone at your organization is a

manager of AI. and there there were sort

of great opportunities but also big

responsibilities that come with that

role. And so we want to make sure that

uh people understand what that role

means and that they're able to be good

managers to the AI tools that they have

access to.

>> And I guess related, you said that

adoption only happens when people see it

solve actual problems. I love that

framing. I hate honestly the dialogue

about like AI adoption because to your

point you just drop like AI on someone's

lap and they don't know what to do with

it or it's not solving a problem not

going to use it and they're certainly

not going to use it effectively. What

makes a good problem for AI to solve?

>> It's pretty simple. It needs to be

repeatable. So it needs to be something

that you well first of all it needs to

be something that you can explain,

right? A good problem that AI can solve

is a problem that you can explain and

solve yourself as if you were explaining

kind of like to an intern, right? Like

if you hired a really brilliant intern

kind of off the street and you needed a

problem solved, this needs to be a

problem that you could explain very well

to that intern and that you would you

would expect an intern to be able to

kind of like solve for you or like a you

know a brilliant chief of staff to solve

for you. So one, can you explain what

the problem is? And can you explain the

solution to that problem with with, you

know, kind of context and information

around what it means to solve that

problem with excellence and with like

the specific steps kind of like

procedurally as you would think about

solving that problem and can you direct

say this chief of staff to the set of

resources that they would need access to

in order to kind of do that task. So

what's a great ch what's a great sort of

um uh you know sort of challenge or or

use case for AI is it something that you

can explain is it something that you can

kind of give context and resources

around is it repeatable right like is it

something for which you can check the

output and you checking the output of

that is still there's an ROI component

right like if you're um if it's going to

be super tedious for you to like check

the output um that you're getting from

AI like let's way like it's like a

thousand rows that you're like uploading

and you're asking it to um you know

create another thousand rows. It's up to

you to always be a good manager to AI

and to check the caliber of the work

that you're getting from AI. So are you

making a good tradeoff in asking it to

accomplish a task that you could

reasonably supervise and get sort of

quality outputs from? So is there kind

of like a quality check? Um, and then I

would say the the last thing is any any

usage of AI that will really give you

sort of leverage or value is something

that you can improve over time and that

um you can kind of like continue and

test and build on. So, is this a

workflow that you could really delegate

and rely on and that you have time to

sort of like step up and continue to

manage? Because uh, you know, again, as

I said, there's this misconception that

once you set up a a maybe a prompt or an

agent that you could just kind of let it

run, but the reality is that you are now

responsible for both the prompt, the

agent, and the outcomes of that. So, are

you prepared to basically now be

checking the work and the output of

that? Um, and is is is the ROI that

you're getting from that sort of

experience worth your time, right? Are

you making a good decision around

picking a problem that you're solving

that you've spent 2 hours sort of like

setting up this prompt or this project

or this structure and it's going to save

you at least 2 hours if not hopefully

significantly more on a regular basis,

right? you don't want to automate or use

AI to solve something that it actually

takes you 5 minutes or 30 minutes to to

do. So I think it's thinking about all

of these different elements, but at the

core it's repeatability, checking the

quality of the outputs, having really

good instructions, and then making sure

that you really have context and

essentially kind of like SOPs, like a

standard operating procedure that you

can hand off to train and test if you

could really delegate a task to AI.

>> Yeah. And how do you find those

problems? And are there questions that

you ask that are great at teasing out

problems that are good for AI to solve?

Yeah, I mean I would say, you know, one

of the easiest one is one of these sort

of me questions is, you know, in these

discovery conversations that we do,

it's, you know, if if you did hire an

intern tomorrow that could help you with

the things that you have on your plate,

what would be the three highest value

tasks that you would kind of think to um

train your intern to kind of accomplish

for you? And so that's that's not a

perfect way to go about it, but I think

you start to get a sense for what are

like the tedious, repetitive, high-v

value tasks that you could get a really

smart person to accomplish if you invest

the time in training them to do that.

There's a very similar sort of

psychology to them doing the same thing

with AI.

>> Yeah. And who do you typically engage

with within one of these client

organizations? Who should be leading

this effort on the client side? So our

perspective is that AI unlike other

technology should really come from the

top. Uh there's a lot of organizations

where there's teams that are using it

there sort of like individuals scattered

across the organization that are getting

a lot of value out of AI kind of like

your your typical bell curve power

users. But in order to get value out of

AI at an organizational level, you need

a coordinated approach and you need

leadership that understands what is the

kind of value that you can be getting

out of these tools. Which means that you

need leadership that is really engaged

and understands just how transformative

these tools are. There's a few examples

of clients that we've we've seen do this

really well. I think notably, you know,

Will, CEO of Walleye, has been an

exceptional leader in um in sort of

under understanding that sort of mental

paradigm that we all experienced at some

point in the last two years and then

really choosing to have the

organizational the organization in this

case, Walleye, uh make a uh a take a

coordinated approach to finding where

there are opportunities to get value out

of AI and making sure that every single

person at the organization is as an

opportunity to see that value, replicate

that value for themselves and then kind

of get scale from that. And when you see

leadership have that light bulb moment

and then generate an environment where

there is a coordinated approach to

getting value out of AI, I think that's

where you see magic happen.

>> Great. And that's a there's a great

interview with um Walley on Spotify and

on YouTube, Dan and AI and I. So, anyone

who's listening to this who wants to

check that out, um, go check it out. Um,

what if the leader is interested in AI

and the organization's interested but

isn't actively using the technology and

doesn't really know what it's capable

of? What do you do then?

>> If an organization hasn't really been

using AI, but they're interested in sort

of

>> someone comes to you and they say, "Hey,

like we really think there's an

opportunity uh to use AI. I haven't

really played around with it yet. Um, I

think it could be like transformative.

Do you

turn them away or can you work with the

leader to get them sort of, you know, up

the

curve or whatever?

>> Yeah, of course. I mean, I I think in

some ways those are the most fun clients

to work with. Absolutely. Uh I there are

so many magic light bulb moments that

happen early on when you are uh you know

just getting familiar with how powerful

these tools are. And if a team has not

started using AI in a coordinated

you know even playful way there are just

so many magic moments early on where you

could start to see the team sort of see

all of the different applications. Uh,

and I would say it's it's especially

cool when a company gets to experience

that together and early on. So 100% I

mean so long as there is an appetite and

an aperture to rethinking workflows and

getting teams to be on these platforms

and then work together to get leverage

out of them. I think the magic is

totally there.

>> Great. I'd love to walk through an

example of how every works. You shared a

case study with me, a private equity

firm that you worked with. Tell me a

little bit about the problem that they

came to you with. Yeah, so this was one

of our very first clients early on. This

is a this is a firm that was interested

in actually doing something much like

what we just talked about, right? they

they did have access to a variety of AI

tools and they were seeing that while

there were sort of disperate approaches

to using the tools, there wasn't sort of

like a coordinated approach that was

making it so that everyone on the team

was getting kind of consistent value.

And so the first thing that we did is of

course just talk to the the different

teams to understand what they were doing

and what their goals were and um you

know kind of learn about the ways in

which they were already using AI and how

they wish they could be using it. Out of

that set of conversations, we

identified, you know, for the, in this

case, it was for the investment team in

particular that they had a pretty

consistent way in which they were using

internal um uh internal notes basically

and internal sort of research to create

like a V1

uh idea or set of questions around like

a new thesis or like a new company that

they were exploring. And then that sort

of that that thesis would go into a memo

that would eventually go to the rest of

the investment team. And so we built out

a bunch of workflows for them. But one

of the favorite one of my favorite ones

that we delivered was basically getting

all of the rich data that they had

internally to help inform a set of

questions that could help an investor

sort of refine their thinking around

their a deal that they were looking at

in a particular sector. So to kind of

like refine their thinking and and

stress test how they were thinking about

a particular company or about like a

particular industry and then using the

rich sort of internal private data that

they had to create like a V1 memo that

they could then take to the rest of

their partners and say hey this is a new

opportunity data that I'm testing here's

all of the relevant data that we have

internally for how we've thought about

this before and why it could make sense

for us to pursue this or explore this.

Um, it made it so that basically the

amount of companies that they could

really think critically on and the kind

of conversations they could have

internally were were richer and more

interesting because we went from being

like a very sort of like lonely

experience of you identify a company

that you're interested in. You have like

so many incredible resources within the

firm that you could be pulling from or

looking at or data that you should be

considering sort of swifting through and

doing that manually to having sort of

like a company that you're looking at in

a thesis being able to do a lot of

preliminary research with an AI workflow

that understands who you are, what you

do, how you think about this kind of

this industry kind of like more broadly

and if it fits into the thesis that your

company has. uh and then to add to that

the rich repository of information that

your firm has collected over the years

that it's existed to really formulate if

there's a there there for the that team

to pursue that opportunity. So that was

one of the that was one of like the very

early on projects that we worked on.

It's been a foundational project for us

understanding specifically how private

equity firms work but also the sort of

fertile ground that there is for more

investment firms to get value out of AI.

uh and you know it's actually something

similar that we're doing now with

another firm but each firm has a

slightly different process for how they

go about the thesis formation say for

example and how their data lives and

where they extract rich and valuable

insights from their internal data and

just helping people think more

critically and find more interesting

insights from the data that they have

access to. I think it's just like a

really fun project to work on and uh and

and value to create for the investors on

the team.

what did you build that on? What was the

sort of tech stack that enabled that AI

workflow?

>> So, you know, for us, we're big

believers that your horizontal LLM is

the most underutilized and

underleveraged tool at your

organization. So, if your team has

already rolled out chat, GBT or Claude

over and over again, we find that most

teams are severely underutilizing it.

And a lot of the workflows that we build

for teams are basically live within chat

GPT with a mix of you know it being GPTs

or shared projects or scheduled tasks.

Um and there every week we see new

things coming out that make that

experience richer and richer. You know

today the browser launched and we're

excited to see

>> how different that that is. But our

experience has been that the easier you

make it for people to feel ownership of

and understand how they can update a

prompt to create new tooling and

resources for them in the context of

TAGBT, the more leverage you can

actually get for individuals and a whole

team. So a lot of our solutions really

live in the LLM that our our clients are

using that the teams are using. oftent

times that's chat GBPT and it it's

basically structured so that it's

incredibly easy to use but also

incredibly effective.

>> You mentioned earlier the sort of

iterative nature of prompt development

or GBT development. What does that look

like in practice? So you you go in, you

ask a bunch of questions, you identify

some opportunities to build solutions to

make workflows more efficient. You come

out with a V1 of your GPT. How how do

you take that V1? What does the process

look like between that V1 of the GPT and

what ultimately they're using at the end

of the project?

>> It varies a lot. Uh I mean early on we

are we are basically creating a prompt

and then asking you know for we call

them AI champions the these are kind of

like our liaison at the firms that we're

working with. We're asking for them to

like run it and test it and then see

where there are sharp edges around that

prompt where it breaks or it stops being

valuable or it's actually not solving

the problem that we thought we were

solving. We get feedback from there. or

we see what context uh might be missing

or what might be missing about the

prompt in order to solve for those

things or to make it richer. We go and

improve it to solve for those things and

then we test it a few times basically to

make sure that the end product is

exactly what we were hoping it would

deliver. So that's what we mean by it

being iterative is we give it a go.

oftentimes, you know, the prompts when

we start out are 85% of the way there,

but it's really that last 15% that makes

a huge difference and it either being

valuable to someone or not being

valuable is if you know when you're when

you're reaching for that tool, it's

solving the specific need that you have.

And so that's what we're trying to solve

for when we're going back and forth and

tweaking it.

>> And then what does the training look

like? you you're working with the AI

champions to identify like opportunities

to refine that GPT. There are many more

people on the team who can get value

from this. You don't just like give it

to them. What is that? What does the

kind of training look like?

>> You know, it so our training is um it's

functional. So, it's team by team. So,

anytime that we're training a team, we

are training say the marketing team or

the compliance team or the legal team.

And so depending on the company, it may

be the case that we actually had a

client recently where they said, "We

don't want everyone to see kind of like

the prompts behind the scenes. It's just

too much. No one has time for that. We

just want kind of like the resource

available to everyone on the team." And

so in that case, we will just work with

a core set of the AI champions. We will

train and empower them to have the the

kind of like all of the resources,

skills, and information that they need

to kind of manage those tools. And then

everyone else will basically kind of

just get trained on using those tools.

But I would say most of the time we're

working with teams that really deeply

want to understand how they themselves

could be getting value out of these

tools. So when we're working say with a

compliance team, we will come in giving

them a prompt or a uh you know say a GBT

for example to to say something very

simple that that again solves a pain

point that that we know that they have

and then we will explain to them how we

built out that prompt and that tool in

order to solve that pain point for them.

Uh we will make sure that they can use

it and know how to use it. So there's a

lot of hands-on time for them to see how

it's structured, see how to use it. But

what we really hope people get out of

these conversations is that they have

the skills to just build more of this

themselves. So we kind of want to make

ourselves obsolete and we want to

continue to help people where they don't

understand how they can be using, you

know, kind of AI. But our goal is for

really people to feel empowered to be

experts in using these tools coming out

of the sessions. Yeah, you mentioned uh

working with teams to solve problems.

What's your take on the rush to stand up

AI labs or centers of excellence versus

like working with line level teams to

solve specific problems?

>> You mean when organizations sort of like

build out like internal AI task forces?

>> Yeah.

>> Yeah. I think I think in the future I

hope more organizations start to think

about having a a a centralized and

coordinated group of people that are

responsible for the AI initiatives at

that company. And I I think the

solutions need to be functional. They do

have to be specific to each team. So uh

so I I do think in the future there are

just going to be more roles where there

are sort there are sort of like I don't

want to think of them as centers of

excellence. That sounds kind of lame,

but I really think of them as kind of

people who are really excited about uh

making sure that when there is someone

somewhere in the company that has built

a tool for themselves that they're

getting real value out of, whoever that

is, that that is being highlighted as an

opportunity for other teams for which

that tool might be relevant. And then

you really multiply the impact of that

tool times the number of other team

members that could potentially be

benefiting from it. And so this is where

I say that there there really does need

to be a coordinated approach to AI.

Oftentimes when we're working with a

client, we will come in and do that to

some extent. But what we want to start

to help shape by nominating AI champions

and having people who are responsible

for the output of their team and who are

sharing that information across teams is

we want to start creating that dynamic

internally where people see

opportunities, share those opportunities

and then everyone benefits from them. So

I I I guess to answer your question in

the future I hope that there are more

coordinated

sort of centralized approaches to how a

company is thinking about AI and

supporting internal AI initiatives and I

think there's plenty of opportunity and

most of the opportunity actually lives

in solving function specific challenges.

>> Uh as we come to a close I'd love to ask

you some questions that solicit advice

for people listening. Uh the first would

be for teams that are just getting

started. What's your like go-to first

move for them to build some momentum?

For teams that are just getting started,

I would say if you start out and think

it before you go and just play with AI,

which we very much want to encourage,

if you think about

what the tedious, painful tasks are for

you, and then you go and have a

conversation with your kind of

enterprise provisioned AI tool and then

ask it how it could help you in

addressing those specific tasks. I think

that would be a good sort of starting

point because as I mentioned, you want

to start solving for things that you

actually need support on. Uh and while

we want to encourage people to just play

on AI, uh we want people to people to

get real value out of it too. So if you

think about what your real needs are and

then use AI to help you think about how

it can help you address those needs, I

would say that that's a fantastic

starting point.

>> Great. And I guess conversely, what's

something you see clients doing that

they should not be doing? I think as I

mentioned earlier, you know, it's it's

so easy to it's easier than ever to

build products uh and also to buy the AI

product duour that sounds like it's

going to solve all of your um you know

all of your data problems or kind of

like what whatever sort of you know

solution says like it's it's going to

you know do rag methodology to surface

the answers that we know whatever

whatever

I think the biggest challenges that

we're seeing companies run into is one

building products they actually don't

need and that don't solve either a

challenge that a customer has or that

the internal team has or two buying a

bunch of AI products or a a set of AI

products that are promising to do

something specific that really a lot of

the time the team is either

underleveraging or they could be using a

uh an enterprise tool an enterprise LLM

tool to support them on. So I would

really start with claude or chat dbt.

Make sure your team really knows how to

get value out of those tools, really

understands what the limits are of those

tools and then if there are additional

tools on top of that or additional

products on top of that to build, you go

from there.

>> Awesome. Last question. Who should reach

out to Natalia and every like what what

is a great client for you?

Yeah, great clients for us are teams

that are usually their enterprise teams.

We work with technology companies,

really big technology companies, private

equity firms, hedge funds, um that are

ready to as an organization make that

pivot into AI adoption and are ready and

prepared to uh do kind of a companywide

approach and build a road map and

execute that roadmap for the entire

team. We love the clients that we work

with. We go deep with a few clients

every year. Uh we we're excited to

support more companies that are ready to

make that transition. Awesome. Natalya,

thank you for coming. I'm just curious.

>> Thanks, too. It's good to see you.