0:00

Today on the A.I. Daily Brief,

0:02

The Rise of the Agent Boss, and

0:04

the Birth of the Frontier Firm. The

0:06

A.I. Daily Brief is a daily podcast

0:08

and video about the most important news

0:11

and discussions in A.I. To join the

0:13

conversation, follow the Discord link in

0:15

our show notes. Hey, hello, friends,

0:17

quick note. Today's main episode ended

0:19

up running very long. It's

0:21

a really interesting exploration of

0:24

the fast coming future, as

0:26

you'll see. And so I decided to just

0:28

let it rip as of the entire episode.

0:30

Tomorrow we will be back with our normal

0:32

headlines as well, but for now, enjoy this

0:34

report about the frontier firm. Welcome back

0:37

to the AI Daily Brief. Well friends, you know

0:39

that we love nothing more here at

0:41

the AI Daily Brief than a big

0:43

prognosticating report, and boy does this year's

0:45

work trend index deliver. Now for

0:47

a little bit of context, you guys

0:49

have probably heard me talk about the

0:52

previous iteration of this report numerous times.

0:54

Right back at the beginning of 2024

0:56

Microsoft and LinkedIn put out their annual

0:59

work trend index, which came out of

1:01

a survey of something like 30,000 or

1:03

31,000 knowledge workers and had some really

1:06

interesting insights. The two big

1:08

things that stood out from 2024 were

1:10

that one, people were using AI, at

1:12

the time they found that 75% of

1:14

global knowledge workers were using generative AI

1:16

and that number had doubled in the

1:18

last six months, and the second

1:20

thing was this sort of secret

1:22

cyborgs insight. We're bringing their own

1:24

tools to work and effectively doing

1:26

it secretly. This year's report, if

1:29

nothing else, shows how dramatically things have

1:31

changed in just a year. This year's

1:33

report really shows how there's been a

1:36

move away from sort of bottoms up

1:38

employee-led AI adoption into thinking about how

1:40

AI is going to change the firm

1:42

from a top-down structural level, which isn't

1:44

to say as we'll see that bottom-up

1:47

adoption doesn't matter, but what bottom-up adoption

1:49

looks like is very, very different. than

1:51

what we thought it was going to

1:53

be last year, where at this point

1:55

employees were mostly just looking to not

1:57

be punished for using chatGBT at work.

2:00

Now, this was all part of a

2:02

larger announcement. In fact, yesterday, Microsoft CEO

2:04

Satie Nadella tweeted, Big Day for Microsoft

2:06

365, Co-Pilot. Co-Pilot has truly become the

2:08

UI for AI, and for me, it's

2:10

the scaffolding for my workday. He then

2:12

pointed to four new features that he's

2:14

been using. The first, unsurprisingly, is agents.

2:16

Satie writes, our new researcher and analyst

2:18

agents have become my go-to 24-7 experts.

2:20

I use them all the time. With

2:22

researcher, the multi-step reasoning aggregates and synthesizes

2:24

information from the web and all enterprise

2:26

data and creates super insightful reports on

2:29

any topic or project. And analysts can

2:31

turn raw data across multiple sources into

2:33

deep insights, forecasts, forecasts, or great visualization.

2:35

Sasha also points out that they're launching

2:37

a new agent store. So sorry Microsoft,

2:39

I like you. I apologize that we're

2:41

going to have to outcompete you with

2:43

our marketplace, but at least the competition

2:45

will be fun. And he also talks

2:47

about co-pilot studio where you can build

2:49

your own agents. Now there's a bunch

2:51

of other features that Satya talks about

2:53

as well. Continuing the really interesting theme

2:55

of companies consolidating around similar naming, Microsoft

2:58

now has its own version of notebooks.

3:00

Satie writes with notebooks I can organize

3:02

all my heterogeneous data for a project,

3:04

whether it's pages, docs, websites, team meetings,

3:06

and Copilot will ground itself from that

3:08

content. And I can turn it all

3:10

into a new modality like an audio

3:12

overview. Now I'm being serious when I

3:14

say that I actually like that companies

3:16

are naming similar products similar products similar

3:18

things. Deep Research is now a category.

3:20

Notebooks is apparently now a category. And

3:22

what makes this powerful is that it

3:24

allows people to think in terms of

3:27

new AI primitives and product categories rather

3:29

than being forced to learn some new

3:31

branding. I wouldn't have expected the branding

3:33

of these products to evolve in this

3:35

way, but I actually think that it's

3:37

quite consumer-friendly. In the short term it

3:39

might be confusing because you don't know

3:41

if you're talking about Google's notebooks LM

3:43

or Microsoft's notebooks, but I think that

3:45

you're already seeing, especially with Deep research.

3:47

What's important about it is that it's

3:49

now a category of behavior. And yes,

3:51

of course, you're going to have your

3:53

preferred tool, but ultimately what matters is

3:56

that you know that that's a category

3:58

of behavior, a category of actions you

4:00

can take. that is roughly consistent from

4:02

platform to platform. There's also a couple

4:04

other cool things. They've expanded their enterprise search,

4:06

which is a huge area of development for

4:09

companies, and they have a new tool to

4:11

turn one type of content into another.

4:13

For example, turning a PowerPoint into

4:15

an explainer video. So these were all

4:17

new features that were released around the report this

4:19

year, which I think makes a lot of sense

4:21

to combine not only a big report, but tools

4:24

that make the trends come to life. And

4:26

especially as we compare this to 2024's work

4:28

trend trend index. It's clear that we're coming

4:30

up on or have actually reached an inflection

4:32

point when it comes to enterprise AI.

4:35

Aparnichenopogata, the chief product officer of

4:37

experiences and devices at Microsoft, did an interview

4:39

with Venture Beat about this and said, were

4:41

around the corner from a big moment in

4:44

the AI world. It started out with

4:46

all of the model advances and everyone's

4:48

been really excited about it and the intelligence

4:50

abundance abundance is available to all

4:52

of the folks, especially at work. She

4:54

also talked about these two new AI agents,

4:56

the researcher and the analyst agent, saying,

4:58

think of them as a really smart researcher

5:01

and a data scientist in your pocket.

5:03

And I think as you'll see this

5:05

is really important, because effectively, this is

5:07

not just thinking about an agent as

5:09

a tool, but thinking about agents truly

5:11

as a colleague or a co-worker. And

5:13

that gets us to the actual work trend

5:15

index, which they call the year the frontier

5:18

firm is born. And right there, even

5:20

in the title. you can tell that

5:22

this is about firm wide structural change,

5:24

not just individual employee productivity. Which is

5:26

not to say that productivity doesn't matter.

5:29

One of the big trends that Microsoft

5:31

saw with their survey this year was

5:33

this capacity gap between 53% of leaders

5:35

saying that productivity needs to increase, but

5:37

80% of the global workforce, including both

5:39

employees and leaders, saying that they lacked

5:41

enough time or energy to do their

5:43

work. Anyone who's dealt with enterprises, particularly

5:46

around AI transformation, has had this experience

5:48

where the constraint on adoption is the

5:50

simple reality that people just don't have

5:52

time to sit down and figure out

5:55

the tools. And these are hands-on tools

5:57

that you have to figure out by

5:59

using them. Interestingly, Microsoft also

6:01

points out that 82% of leaders

6:03

expect to use agents to meet

6:06

the demand for more workforce capacity.

6:08

Again, not to beat a dead drum,

6:10

but in a single year, actually less

6:12

than a year, because the last survey

6:14

came out in May, we've gone from

6:16

75% of knowledge workers using AI and

6:18

80% of them doing it secretly to

6:20

82% of leaders expecting to use agents

6:22

to expand their team's capacity. If that

6:24

doesn't put a fine point on just how

6:27

fast things are changing, I don't know what

6:29

does. And indeed, really the big new

6:31

force throughout this report is the

6:33

emergence of agents, and what agents

6:35

represent as an actual augmentation of

6:37

the workforce, a new set of

6:39

digital workers. Microsoft sees the evolution

6:41

to a quote-unquote frontier firm happening

6:43

in three phases. Phase one, they

6:45

call human with assistant. Every employee

6:48

has an AI assistant that helps

6:50

them work better and faster. Now

6:52

interestingly, if we go back to KPMG's

6:54

recent pulse survey, at this point this

6:56

seems like total table stakes stakes. Remember,

6:59

between Q4 of last year and

7:01

Q1 of this year, KPMG found

7:03

that weekly knowledge assistant usage was

7:05

up from 48 to 61 percent,

7:07

and daily usage of AI productivity

7:09

tools had gone from 22 percent

7:11

to 58 percent, just an absolutely

7:13

huge acceleration. Today's episode is brought

7:15

to you by Vanta. Vanta is a

7:17

trust management platform that helps businesses

7:20

automate security and compliance,

7:22

enabling them to demonstrate

7:24

strong security practices and

7:26

scale. In today's business landscape, businesses can't

7:28

just claim security, they have to prove

7:31

it. Achieving compliance with the framework like

7:33

SOC 2, ISO 2701, HIPAA, GDPR, and

7:35

more, is how businesses can demonstrate strong

7:37

security practices. And we see how much

7:39

this matters every time we connect enterprises

7:42

with agent services providers at

7:44

Superintelligent. Many of these compliance

7:46

frameworks are simply not negotiable

7:48

for enterprises. The problem is that

7:50

navigating security and compliance is time-consuming and

7:52

complicated. It can take months of work

7:55

and use up valuable time and resources.

7:57

Vanta makes it easy and faster by

7:59

automating compliance across 35 -plus frameworks.

8:01

It gets you audit -ready in

8:03

weeks instead of months and saves

8:05

you up to 85 % of

8:07

associated costs. In fact, a

8:09

recent IDC white paper found that

8:11

Vanta customers achieved $535 ,000 per

8:14

year in benefits, and the

8:16

platform pays for itself in just

8:18

three months. The proof is

8:20

in the numbers. More than 10

8:22

,000 global companies trust Vanta, including

8:24

Atlassian, Korra, and more. For

8:26

a limited time, listeners get $1

8:28

,000 off at vanta.com/nlw. That's Vanta.com/nlw

8:30

for $1 ,000 off. Today's

8:33

episode is brought to you by

8:35

Plum. If you build agent work

8:37

flows for clients or colleagues, you

8:39

need to check out Plum. Plum

8:41

is the only AI -native workflow

8:43

builder on the market designed specifically

8:45

for automation consultants, with all the

8:47

features you need to create, deploy,

8:49

manage, and monetize complex automations. Features

8:51

like one -click updates that reach all

8:53

your subscribers, user -level variables for

8:55

personalization, and the ability to protect

8:57

your prompts and workflow IP. Make

8:59

your life easier, your clients happier,

9:01

and your business thrive with Plum.

9:03

Sign up today at useplum.com. That's

9:05

Plum with a B, forward/NLW. Today's

9:09

episode is brought to you by Superintelligent,

9:11

and I am very excited today to

9:13

tell you about our consultant partner program.

9:15

The New Superintelligent is a platform that

9:17

helps enterprises figure out which agents to

9:19

adopt, and then with our marketplace, go

9:21

and find the partners that can help

9:23

them actually build, buy, customize, and deploy

9:25

those agents. At the key of that

9:28

experience is what we call our agent

9:30

readiness audits. We deploy a set of

9:32

voice agents which can interview people across

9:34

your team to uncover where agents are

9:36

going to be most effective in driving

9:38

real business value. From there, we make

9:40

a set of recommendations which can turn

9:42

into RFPs on the marketplace or other

9:44

sort of change management activities that help

9:46

get you ready for the new agent

9:48

-powered economy. We are finding a ton

9:50

of success right now with consultants bringing

9:53

the agent readiness audit to their client

9:55

as a way to help them move

9:57

down the funnel towards agent deployments, with

9:59

the consultant playing the role of helping

10:01

their client hone in on the right opportunities based on what

10:03

we've recommended and helping manage the partner selection process. Basically,

10:05

the audits are dramatically reducing the

10:07

time to discovery for our consulting

10:10

partners, and that's something we're really excited

10:12

to see. If you run a firm and have

10:14

clients who might be a good fit for the

10:16

agent readiness audit, reach out to Agent at B-super

10:18

dot AI with consultant in the title, and we'll

10:20

get right back to you with more on the

10:22

consultant partner program. Again, that's

10:24

agent at B-super dot AI, and put

10:26

the word consultant in the subject line. So,

10:29

back to the frontier firm, phase one

10:31

is human with assistant, phase two is

10:33

human agent teams. Agents join teams

10:35

as digital colleagues, taking on

10:37

specific tasks, human direction. This is

10:39

where a lot of the discourse is right now. Jason

10:42

Clinton, the Chief Information Security Officer

10:44

at Anthropic, recently did an interview

10:46

with Axios, where he said that

10:49

the company thinks that fully AI

10:51

employees are just about a year away at

10:53

this point. Phase three of the frontier firm,

10:55

Microsoft sees as human-led agent

10:57

operated. Human set direction and

10:59

agents execute business processes and workflows

11:02

checking in as needed. This is

11:04

highly resonant from where I sit with

11:06

all of our conversations at Super Intelligent

11:08

and where it seems to me that

11:10

things are going. I mentioned before that

11:12

right now even the advanced firms still

11:15

tend to view agents as one-to-one replacements

11:17

for or augmenters of specific tasks or

11:19

roles or functions. And that makes sense if

11:21

we're in this sort of phase two of human

11:23

agent teams. I've also shared my view. that

11:25

in the future we're not going to hire

11:27

one agent for something, we're going to deploy

11:29

a thousand agents. We're going to have agent

11:32

swarms, we're going to have battle games

11:34

type scenarios, and that looks a lot like

11:36

this idea of a phase three of a human-led

11:38

agent operated firm. But aside

11:40

from just prognostications, what are

11:42

the interesting numbers that Microsoft

11:44

actually found around all of this? First

11:47

of all, they certainly found a lot of

11:49

reasons for and justifications of why employee productivity

11:52

is hampered. Employees are interrupted every

11:54

two minutes by meetings, emails, or

11:56

other types of notifications. The net total

11:58

for the average employee was two... 175 interruptions

12:00

in a day. 60% of meetings

12:02

were ad hoc rather than scheduled.

12:04

Chats outside of the workday are up

12:06

15% year over year. Meetings after 8

12:09

p.m. are also up 16% year over

12:11

year. Around 50% of both leaders and employees

12:13

say their work feels chaotic and

12:16

fragmented. But how are firms actually

12:18

thinking about change? Well, again,

12:20

another big change between these two

12:22

years is that if the 2024

12:25

work trend index show the story

12:27

of bottom-up adoption, These two shows a

12:29

story of much more top-down approaches. The

12:31

Microsoft study found 81% of business

12:33

decision-makers reporting that they want to

12:35

rethink core strategy and operations with

12:38

AI. That's very different than just

12:40

thinking about employee productivity. Microsoft's Cheniprogata

12:43

said, that's a shift between even last year,

12:45

where it was much more bottom-up in

12:47

employee-led. What that tells us is that there

12:49

needs to be much more of a top-down

12:52

AI strategy, but also AI products that you

12:54

roll out in the enterprise, with security compliance,

12:56

all of the guardrails. So what are the

12:58

priorities for these firms and how are

13:00

they thinking about change internally? One

13:02

of the questions was about ranking

13:04

most likely strategies. The percentages reflect the

13:07

share of respondents who ranked the

13:09

answer as a top three most

13:11

likely strategy. Down at the bottom,

13:13

the least popular answer was no

13:15

popular answer was prioritizing AI-specific

13:17

skilling of existing workforce.

13:19

And in some ways that reflects

13:22

continuity with last year. However, one... The

13:24

percentage that prioritized AI-specific skilling of

13:26

existing workforce as a top three

13:28

likely strategy was only 47% even

13:31

though that was the top answer.

13:33

47% is I think quite a bit lower than

13:35

the type of reporting we would have seen last

13:37

year. Meanwhile, the number that are

13:39

willing to admit that they're thinking about

13:41

using AI to reduce headcount being a

13:44

third 33% is up fairly meaningfully from

13:46

where we might have been last year. I tend

13:48

to think that that has more to do with

13:50

macroeconomic instability. and big questions around the

13:52

global work environment than it does around

13:55

just AI capabilities. But whatever the case,

13:57

I do think we're in the midst

13:59

of... a shift of prioritization.

14:01

Now positively, there is clearly a lot

14:04

of interest in using digital

14:06

labor as supplemental. Just behind

14:08

prioritizing AI-specific skilling of

14:10

existing workforce was maintaining headcount

14:12

but using AI as digital

14:14

labor. 45% of respondents had that as

14:16

a top three most likely strategy.

14:18

32% had increasing headcount to support business

14:21

needs, and 40% said that a

14:23

top strategy was prioritizing retention

14:25

with long-term incentives and bonuses.

14:27

And so yes, I do think inevitably more

14:30

companies are thinking about AI as a

14:32

possible headcount reducer, but that's far from

14:34

the only trend. And in fact, I think net

14:36

net, there's really positive indications that

14:38

these leaders who want to become

14:41

frontier firms are thinking about AI

14:43

more as an opportunity technology than

14:45

as an efficiency technology exclusively to

14:47

use a parlance that I've adopted

14:49

before. Now one other totally unsurprising

14:51

statistic is that companies are definitely

14:54

hiring for AI-specific roles. 78% of leaders

14:56

overall are considering hiring for these types

14:58

of roles, and that number hits 95%

15:00

when you ask frontier firms. The roles include

15:02

things like AI trainers, data specialists, security

15:04

specialists, AI agent specialists, ROI

15:06

analysts, as well as AI

15:09

strategist in specific functional areas

15:11

like marketing finance, customer support,

15:13

and consulting. The next big section of

15:15

the report is called Human Agent Teams

15:17

will upend the org chart. And this

15:19

gets into that phase two of human-agent

15:21

teams, where agents join teams as digital

15:23

colleagues taking on specific tasks at human

15:25

direction. A couple interesting observations

15:27

from this. One theme that I think you'll

15:29

hear a lot more about, because it's a very

15:31

crisp way of explaining this, Microsoft argues

15:34

that the traditional org chart may increasingly

15:36

be replaced by a work chart, what

15:38

they call a dynamic outcome-driven model where

15:40

teams form around goals, not functions,

15:42

powered by agents that expand employee

15:44

scope and enable faster more impactful

15:46

ways of working. They compare this to

15:48

movie production, where it's not like you

15:50

have a single org chart, you have

15:52

dynamic teams that are assembled for the

15:54

specific roles in a temporary sort of fashion

15:56

to get the specific jobs done. We

15:58

also started to get... some numbers from

16:00

Microsoft here around agents. 46 % said

16:03

that their companies are using agents to

16:05

fully automate workflows or processes. And

16:07

we also got some information around which

16:09

different areas are seeing the most

16:11

adoption. Not surprisingly, it's areas like marketing,

16:13

customer success, internal communications and data

16:15

science where agentic systems are most breaking

16:17

out. The survey also explored the

16:19

specific reasons that agents in AI are

16:21

being turned to. And while none

16:24

of this is particularly surprising, it's still

16:26

really interesting to see displayed in

16:28

this way. The most frequent response for

16:30

why an employee or a team

16:32

member might turn to AI is 24

16:34

-7 availability. After that are things like

16:36

speed, limitless capacity and the endless

16:38

stream of ideas on demand. And yet

16:40

for all that's interesting about this

16:42

sort of phase two of human -agent

16:45

relationships, it all does kind of feel

16:47

like it's prelude to part three.

16:49

Phase three is where, as Microsoft puts

16:51

it, every employee becomes an agent

16:53

boss. They define an agent boss as

16:55

someone who builds, delegates to and

16:57

manages agents to amplify their impact, working

16:59

smarter, scaling faster and taking control

17:01

of their career. And although it's early,

17:03

there are already indications that this

17:06

is the place we're trending to. 28

17:08

% of managers are considering hiring AI

17:10

workforce managers to lead hybrid teams

17:12

of people and agents. 32 % plan

17:14

to hire AI agent specialists to design,

17:16

develop and optimize them within the

17:18

next 12 to 18 months. In the

17:20

next five years, 41 % of leaders

17:22

said that their teams will be

17:24

training agents. 38 % said they'll be

17:27

redesigning business processes. 42 % said that

17:29

they'll be building multi -agent systems to

17:31

automate complex tasks. Now along with this

17:33

shift to agentic thinking, leaders versus

17:35

employees have started to race ahead. The

17:37

survey introduced seven indicators to identify

17:39

who has a quote agent boss mindset,

17:41

things like familiarity with agents, regular

17:43

AI usage, trusting AI for high stakes

17:45

work, expectations to manage agents, using

17:48

AI as a thought partner and more,

17:50

and leaders were ahead on all

17:52

of those different areas. Microsoft summed up,

17:54

last year employees led the AI

17:56

wave, this year it's flipped. What explains

17:58

the gap? it's because leaders

18:00

are the first to feel the pressure to

18:02

have an AI strategy and the first to

18:04

be held accountable for making it work. They

18:06

see what's coming and know they can't afford to

18:09

wait. This part though is also really important.

18:11

Microsoft continues, managing agents

18:13

also plays to their strengths, delegating,

18:15

guiding, and stepping in when needed. And

18:18

this is really important. In a world of

18:20

agents, everyone is going to be more like a

18:22

manager than they are today. When you have hundreds

18:24

or thousands of agents available on

18:26

demand for any type of function

18:28

you want, you're going to have

18:30

to get good at coordinating them,

18:32

orchestrating their actions, figuring out how

18:34

to plan around their capabilities. This is

18:36

in many ways the biggest shift that

18:39

we're likely to see. And I will also say this.

18:41

This is the biggest reason that the

18:43

current crop of upskilling platforms is

18:45

woefully out of touch with the

18:47

actual needs of modern employees. I

18:49

say this is someone who started

18:51

a couple years ago a platform

18:53

specifically to upskill employees with AI. The

18:56

reason that we have been so aggressive

18:58

about pivoting and changing what that

19:00

company does and responding to the

19:02

actual changes in the marketplace is

19:04

that this is where things are headed. AI

19:06

success will not be. It isn't even

19:09

now. Is your team good at using

19:11

individual co-pilot or assistant tools? Long-term

19:13

success with AI, real transformation, is

19:15

going to be about fundamentally reimagining the

19:18

structure of organizations and empowering

19:20

individuals and teams to manage

19:22

armies. of agents to do

19:24

things that are literally not

19:26

possible right now. That is a very,

19:28

very different challenge than making sure

19:30

that people are good at prompting

19:32

chat gBT. And I think too much of

19:34

our upskilling conversation is stuck in that

19:37

old way of thinking from, you know, the ancient

19:39

days of 12 months ago. In any case,

19:41

as you can probably tell, I think that

19:43

even more than last year, this work

19:45

trend index is hugely instructive for where

19:48

the enterprise is trending and what AI

19:50

and agent adoption is going to look

19:52

like inside companies. I would highly encourage you to

19:54

spend some time with the report. Check it

19:56

out. Think about the implications for your company.

19:59

I'll insert a stand- chill here for super

20:01

intelligent and our agent readiness audits,

20:03

which can help you better understand

20:05

the specific implications for your specific

20:07

company. But whatever you're doing to get

20:09

prepared, the takeaway is that you have to

20:11

be doing something, because the future is increasingly

20:13

here and coming at us a lot faster

20:15

than even it seemed like just a year

20:17

ago. For now that is going to do

20:19

it for today's a ideally brief. Appreciate your

20:21

listening or watching as always, and until next

20:23

time, peace.