The implications of AI on a Center of Excellence

Summary

The video script delves into the rapid rise of AI, especially ChatGPT, and its potential to revolutionize various industries. It emphasizes the importance of prompt engineering, data quality, and architectural considerations for organizations to harness AI's full potential. The speaker uses analogies to illustrate the early stages of AI development and warns against underestimating its future capabilities. The script also highlights the need for enhanced skills, such as prompt engineering, data governance, and business analysis, to leverage AI effectively. It encourages establishing centers of excellence to foster innovation, collaboration, and best practices within organizations, enabling them to identify AI's sweet spots and maximize its benefits.

Takeaways

• 😀 GPT (Generative Pre-trained Transformer) models like ChatGPT have brought AI capabilities to the forefront, generating human-like text based on prompts.
• 🚗 We are still in the early stages of AI capabilities, like driving a Formula One car compared to a regular car.
• 🧾 Prompts need to be more explicit and detailed to get better results from GPT models. Prompt engineering is a new skill to develop.
• 🧩 GPT models excel at solving puzzles and reasoning based on provided information, rather than using general knowledge.
• 👨‍💻 AI highlights the importance of skills we should already have, like good business analysis, data quality, and documentation standards.
• 🕵️‍♀️ AI forces us to be more disciplined in areas like governance, change control, and metadata management due to added risks.
• 👩‍🏫 AI creates a need for new skills like prompt engineering, but also emphasizes improving existing skills like data governance and architecture.
• 🚀 AI can be a catalyst for getting buy-in from executives to fund projects and improve areas we've neglected in the past.
• 🏆 Centers of Excellence (CoEs) should act as innovation hubs, bringing together people passionate about AI and exploring its use cases.
• 🎯 Organizations should focus on finding the sweet spots where AI can provide significant ROI, easy adoption, and consistent results.

Q & A

What is the main topic discussed in the transcript?

The main topic discussed is the impact of AI and specifically GPT (Generative Pre-trained Transformer) on businesses, and the skills and processes organizations need to adopt to leverage AI effectively.

Why did ChatGPT gain such rapid adoption with 100 million users in 2 months?

ChatGPT gained rapid adoption because it was easy to use (just type in a prompt) and the results were staggeringly good, allowing users to generate text like song lyrics and content that seemed beyond current capabilities.

What analogy is used to describe the current stage of GPT development?

The analogy of a car is used. ChatGPT is like a high-performance Porsche GT3 on a track, which gives a glimpse of the capabilities of a Formula 1 car, but doesn't represent the full potential of AI, which is still in its very early stages like a flip phone compared to an iPhone.

What are some key skills organizations need to develop to leverage GPT effectively?

Organizations need to develop skills in prompt engineering (crafting effective prompts), data quality, business analysis, following industry standards for documentation, understanding system architecture, data governance, and change management.

Why is prompt engineering important for GPT?

Prompt engineering is important because longer, more explicit prompts are needed to get better results from GPT. Prompts can be thought of as code that needs to be monitored, refined, and managed for dependencies, just like traditional code.

What are some use cases where GPT excels?

GPT is good at solving puzzles and generating text when given all the puzzle pieces (context and data). It performs well in areas like generating code, interpreting legal documents, and writing lyrics because it has ingested large amounts of relevant data.

What is the role of a Center of Excellence (CoE) in the context of AI adoption?

A CoE acts as an innovation hub, bringing together passionate individuals and teams to collaborate on AI initiatives, find use cases with strong ROI and ease of adoption, and ensure consistent and reliable results.

How does AI highlight the importance of existing best practices?

AI punishes mediocrity and exposes weaknesses in areas like data quality, business analysis, documentation, and architecture. It forces organizations to adopt disciplines they may have neglected in the past due to the increased risks and potential impact of AI on business operations.

What are some key pillars or skills needed for a strong AI Center of Excellence?

Key pillars or skills include vision, leadership, governance, change control, methodology, standards, metadata management, architecture, security, change management, project management, tooling, and innovation.

How can AI be used as a catalyst for organizational change?

AI can be used as a catalyst for organizational change by highlighting the need for improvements in areas like documentation, architecture, and data governance. It provides an opportunity to get buy-in and funding from executives to address these issues, which were previously overlooked or underfunded.

Outlines

00:00

🚀 Explosive Growth and Potential of AI

The speaker talks about the rapid growth of AI, particularly GPT, which gained 100 million users in just two months. GPT's ability to generate realistic text outputs, such as lyrics for a country song, has brought AI to the forefront. The speaker uses the analogy of driving a supercar (GPT) versus a Formula 1 car (the full potential of AI) to illustrate that we've only scratched the surface of AI's capabilities. The current perception of AI is based on experiences with ChatGPT, which is like a flip phone compared to the iPhone of the future.

05:00

🛠️ Prompt Engineering and AI as a Tool

The speaker highlights the importance of prompt engineering, which involves crafting effective prompts to get the most out of AI models like GPT. Prompts need to be longer and more explicit, with detailed context to get better results. AI excels at solving puzzles when provided with the necessary pieces of information. The speaker emphasizes that AI should be used as a tool, not a replacement for human expertise. AI can generate convincing but potentially incorrect answers, known as "hallucinations." Proper delegation and validation of AI outputs are crucial. The speaker predicts the need for new skills like prompt engineering, rather than entirely new job roles.

10:01

🎯 Importance of AI Highlighting Existing Best Practices

The speaker discusses how AI is highlighting the importance of existing best practices that organizations may have overlooked or undervalued. AI's performance is heavily dependent on the quality of data, business analysis, documentation, and architecture. AI's ability to read and process this information forces organizations to improve their standards and follow industry best practices, such as using the Universal Process Notation for process mapping. The speaker suggests using AI as a catalyst to get executive buy-in and funding for improvements in these areas.

Keywords

💡GPT (Generative Pre-trained Transformer)

GPT refers to a series of AI language models developed by OpenAI, with capabilities ranging from text generation to language understanding. In the video's context, GPT's significance lies in its ability to quickly generate coherent, contextually relevant text based on prompts, illustrating a leap in AI's accessibility and potential applications. The speaker highlights GPT's rapid user adoption and its intuitive interface, allowing users to generate creative content, such as song lyrics, with ease. This underscores GPT's role in democratizing AI use and sparking widespread interest in its capabilities.

💡ChatGPT

ChatGPT is a specific application of the GPT model focused on generating human-like text in a conversational format. The script emphasizes ChatGPT's role in bringing GPT's potential to a broad audience, demonstrating how natural language processing can facilitate interactions that feel conversational and intuitive. Through examples like prompt engineering and solving puzzles, ChatGPT is portrayed as a tool that can perform complex tasks, from generating text based on detailed instructions to reasoning through problems given a set of conditions.

💡Prompt Engineering

Prompt engineering is the practice of crafting inputs (prompts) designed to elicit specific outputs from AI models like GPT. The video script explores this concept extensively, suggesting that effective prompts can significantly enhance the utility and accuracy of GPT's outputs. It involves not only the formulation of questions or commands but also the strategic structuring of these inputs to achieve desired results. Examples include generating business documentation or performing specific tasks like writing termination letters, highlighting the importance of detailed, clear prompts in leveraging AI capabilities.

💡Car Analogy

The car analogy is used to contextualize the current state of AI and GPT technology relative to its potential. Comparing non-users to driving a Honda Accord, GPT users to driving a Porsche GT3, and the yet-unrealized potential of AI to the experience of a Formula One driver illustrates the vast spectrum of AI's capabilities and the stages of its accessibility and performance. This analogy serves to emphasize that while current applications like ChatGPT offer a glimpse into AI's potential, the future possibilities are as different and more advanced as a Formula One car is to a standard vehicle.

💡Use Cases

Use cases refer to the specific scenarios or problems to which AI and GPT can be applied effectively. The video script discusses exploring and identifying successful applications of GPT, emphasizing the importance of understanding where AI can provide the most value. This involves recognizing tasks that benefit from AI's reasoning and generative capabilities, such as document generation or puzzle solving, and integrating AI into workflows in a way that enhances productivity and problem-solving.

💡AI Skills

AI skills encompass the knowledge and abilities required to effectively use AI technologies like GPT. The video underscores the necessity of developing these skills across organizations, not as specialized roles but as competencies integral to various job functions. This includes the ability to craft effective prompts, validate AI-generated outputs, and understand the applications and limitations of AI tools. The script advocates for AI literacy as essential for leveraging AI technologies to their fullest potential.

💡Hallucination

In the context of AI, 'hallucination' refers to instances where a model generates incorrect or nonsensical information as if it were factual. The video script touches on the importance of being aware of this tendency in AI outputs, underscoring the need for human oversight in validating and contextualizing AI-generated content. This concept illustrates a critical challenge in using AI: ensuring reliability and accuracy in its outputs.

💡Metadata

Metadata refers to data that provides information about other data, which is critical for organizing, managing, and understanding data within systems. The script highlights the role of metadata in enhancing AI applications, such as by informing AI-generated user stories or test scripts. Effective management and documentation of metadata are portrayed as essential for leveraging AI in tasks like process automation, illustrating the intersection between AI capabilities and data management practices.

💡Center of Excellence (CoE)

A Center of Excellence refers to a team or entity within an organization dedicated to promoting excellence and innovation in a specific area, such as AI or Salesforce usage. The script discusses the CoE in the context of guiding, standardizing, and innovating AI practices within organizations. It is portrayed as a hub for collaboration, knowledge sharing, and governance, ensuring that AI technologies are used effectively and ethically across different departments and projects.

💡Innovation

Innovation, within the script, is tied closely to the exploration and implementation of AI in solving new or existing problems in ways that significantly improve efficiency, effectiveness, or capabilities. The video emphasizes the role of innovation in identifying and capitalizing on AI's potential use cases, advocating for a proactive and open approach to experimenting with AI technologies. This includes fostering an environment where employees are encouraged to share ideas and experiences with AI, promoting a culture of continuous improvement and adaptation.

Highlights

GPT reached 100 million users in two months due to its ease of use and impressive results in generating content like song lyrics.

ChatGPT is compared to driving a Porsche GT3 to illustrate the advanced capabilities beyond traditional tools, but the complexity of AI like a Formula One car's controls highlights the early stage we are at with GPT.

The potential of GPT is often underestimated when only seen through the limited experience with chat interfaces.

Prompt engineering emerges as a crucial skill for effectively using GPT, requiring detailed and explicit instructions to generate valuable outputs.

GPT excels at solving puzzles given all the necessary pieces, showing its strength in reasoning and generating text based on specific inputs.

Applications embedding AI with well-defined prompts can significantly speed up tasks like creating user stories or test scripts, turning hours of manual work into minutes.

The importance of not abdicating responsibility to AI is emphasized, as users need to be able to validate GPT's outputs to avoid hallucinations and inaccuracies.

AI's current stage is akin to the flip phone era, suggesting that we are only at the beginning of its capabilities and impacts.

Effective use of AI requires better prompts, which is a new form of interaction that needs to be learned and optimized over time.

AI punishes mediocrity by quickly exposing poor data or analysis, emphasizing the need for high-quality inputs.

Documentation quality and the architecture of systems become increasingly important as AI relies on these elements to generate accurate outputs.

AI-driven changes in roles are about acquiring new skills rather than creating entirely new job titles, such as prompt engineers.

The analogy of AI as Google on steroids highlights its potential to vastly outperform existing tools when provided with well-crafted queries.

The concept of 'prompt drift' suggests that the same prompt can yield different results over time, necessitating ongoing monitoring and adjustment.

The Center of Excellence (COE) plays a crucial role in guiding AI adoption, emphasizing innovation, collaboration, and the strategic implementation of AI across teams.

Transcripts

ai's been around what 15 16 years I

think what's happened in the last year

now uh GPT has suddenly and open Ai and

chat GPT has suddenly brought everything

to the Forefront GPT it got to 100

million users in two months and why was

that well first of all I think a it was

easy you went and you type something but

B the results were staggering you could

type in and go give me the lyrics for a

country song about this and it would

actually go and generate that and GPT is

just very good at working out what the

next best word is um so I spend quite a

lot of time on this on the confidence

circuit at the moment and the first

thing is that people think of GPT the

potential of GPT Through The Eyes of

what they can do with chat GPT over the

last like six nine 12 months um

and the analogy I like to use is the car

analogy there so everyone out here can

probably Drive there's a there's a

hondre cord fantastic car um and that's

what it's like not using GPT at all um

some of us have been lucky enough to

drive like that Porsche GT3 on a on a

track and you're staggered by the levels

of grip the acceleration you just can't

believe actually how fast a car will go

around a corner and you think okay I've

now I now understand what the world of a

Formula One driver is like and that's

sort of where we all are we think we

chat GPT we've seen what the future

looks like and then you you walk off the

pits and you bump into Lewis Hamilton

the Formula One drive and he goes no no

no you've you just don't understand what

the Futures look like on the right hand

side there's a picture there of the

steering wheel of a Formula One car

somewhere on there is a button called

the launch control how you actually get

it off the uh from from the the pit from

the pits or from the uh from The

Starting Line none of us could even get

the car off the track without stalling

and I think that's really where we are

with GPT we're at the very very early

stages and I think the first danger is

that if people think about the potential

of GPT Through The Eyes or or through

their experience of what chat GPT can

can do the other analogy is we're still

at the flip FL flip phone stage we're

nowhere near where an iPhone now is so

we are really at the very very early

stages it's not quite the wild west but

actually it's very unstructured I think

people are still finding out what use

cases work what don't work and we need

to try and cut through some of that and

work work through to a go what what as

leaders should we be worried about and

how can we support our teams in terms of

what they're currently doing with

GPT so I think when when we think about

chat GPT at the moment we tend to think

it as Google on steroids we'll ask it a

question it comes back with the answers

and the better the question the better

we ask the question the more complete

the question we ask the better the

answers if you just said to your intern

who's bright and excited book me a

restaurant they come back and they book

your a restaurant the uh they booked you

into the Italian restaurant for next

Tuesday they go no no no no no not not I

don't and and I really wanted a

restaurant for next Thursday and we seem

to have that conversation back and forth

with GPT if you said to your intern I'm

entertaining A clients they like Indian

food and they like Chinese food uh we're

driving and I'd like a table in a

private room for four that's that's a

good prompts suddenly the internet can

come back and go right if I know that

you're driving so I need we need parking

space I know when I and I think we need

to get to a point where we're asking

better prompts so that's the first thing

uh prompts need to be longer be need to

be lot more explicit to get some of the

benefits out of it and that is a new

skill call it prompt engineering but

there's actually more to prompt

engineering than that and I'll explain

that in a moment but the other thing

that GPT is very good at it's very good

at solving puzzles if you give it all

the puzzle pieces now a slight challenge

is that the limitation on the amount of

information you can give it uh but you

could give it I know a contract and then

say Okay I want to terminate this

contract against these four bullet

points and it will write a reasonably

good termination letter if you give it

all so rather than trying to get it to

use its knowledge of the world and

knowledge of History instead think about

it actually is using it its ability to

reason and then come back with text and

it's really really good at doing that so

think more about how you give it the

puzzle pieces than ask it to solve the

puzzle and that's actually where I think

we're getting into bit more of prompt

engineering or certainly where you're

seeing applications embedding AI into

them where they've already thought about

I'm asking GPT to do something in the

context of my application it knows where

to get the puzzle pieces it's now simply

using GPT as the back end to solve those

puzzle pieces and I mean an easy example

is is our world elements. cloud if you

map out a business process using the uh

Salesforce UPN standard Universal

process mapping

notation AI rgpt application will write

really really good user stories user

stor with acceptance criteria and also

test scripts also because we know what's

in your org because we pulled all the

metadata AI can then look at the

acceptance criteria and then work out

which metadata you could reuse or which

metadata needs to be you so again

solving what would take eight hours

manually could be done in five minutes

um but before we start getting worried

about losing jobs or actually the Coe

lead is going great well we don't need

anyone in the team we need we need to

think about what's actually happening

here first of all we need to delegate we

can't

abdicate we can't ask GPT to do anything

we can't do ourselves because we then

can't validate the answers I think

everyone's heard of the term

hallucination AI is very good it

actually it's rather like the male 22y

old intern uh and I I deliberately use

the word male because they are never in

doubt but not necessarily right it gives

very convincing answers which may not

necessarily be correct so you need to

make sure you a delegate the work scope

it correctly but secondly you need to

validate whether the answers has come

back with are correct and I think we're

still in the early stages of

understanding the use cases where it

works really really well and there's

less hallucination we're going to see

new skills not new roles so we won't see

the role of a prompt engineer the same

ways we don't have in our organizations

the job of Google search engineer but I

do think everyone needs Google search

skills the same way as we all need will

need prompt engineering skills um but I

think what AI is doing is it's

highlighting that some of the things we

know we ought to be doing and don't do

very well it's highlighting the

importance of those so message here is

first of all I mean phobo at the top

there fear of being obsolete I think

that's a very valid fear for all of us

we need to stay current we need to

understand the implications of new

technologies coming in and what it means

for ourselves but also for our teams but

don't think about hiring new ski uh new

roles think about actually how the

skills we need to generate inside our

organization I think this is about

Career Development which is where we

started this and then think about how

you delegate you can't just go gbt told

gave me the answer we need to think

about how we delegate the the question

to make sure we get the right in a way

that we get decent answers

back so first of all let me just think

very briefly about the sorts of skills

we need to to to create the first is

prompt engineering which is something we

haven't had before so the idea of

writing a prompt and again back to my

analogy of the Porsche versus a Formula

1 car a prompt isn't just like a Google

search with a little bit more there's

actually quite a lot more involved in

this first of all we need to a prompt

can actually be quite detailed it could

be almost like an email

template inside our own organization

we're using templates for prompts where

we're inserting say three or four bullet

points in different places and that

prompt is then writing a website for us

our internal website for all our new

features and fun functionality so

there's a website aimed at our customer

success marketing and sales team so they

get early sight of functionality that

website is built by GPT using some

templated prompts which are product

management team of have optimized and

revised over time so that's the first

thing a prompt isn't just what you type

in you could actually have a templates

which you reuse so that's the first

thing I think the second thing which is

interesting is that when you test that

prompt the result you're going to get

could be different when it's done today

versus a week's time versus A month's

time because that prompt is hitting a

large language model which could be

optimized being trained could be being

refined so unlike where I know we write

a flow and it executes the same every

time with a prompt it could change over

time so we need to start monitoring it

for

drift we also need to make sure that

when our teams are using those the

result of the prompt that we understand

how much they have to modify it so we

understand how good that prompt was if

it generates say an email that we send

out to a customer and then the service

agent or um the um support agent has to

make loads of changes to it maybe the

prompt needs to be

refined is that prompt even being used

uh or is it being used with no changes

whatsoever which would be quite

concerning so again there was some

there's some monitoring of prompts that

we probably would never do in terms of

code or um say a flow so declarative

code and then the other thing we need to

worry about dependencies so if that

prompt is is using say metadata from our

Salesforce system or a third party

system like data Cloud if that metadata

gets changed the prompt will still work

so we need to understand when we're

actually making changes to metadata does

a prompt use it almost the same way as

does an email template use it except

there's a bit more at stake now if that

prompt is using a large language model

to make some changes or decisions based

on the data coming out of that metadata

so think of prompts as certainly code

but actually um almost a riskier set of

code than maybe um a flow or Apex so

that's a new skill the other the other

four items on there are things that we

should or ordinarily be doing but maybe

is it probably doesn't have the level of

importance as it as we probably should

allocate to it uh but AI punishes

mediocrity if you've got poor data it

will give you poor results really

quickly if you haven't done very good

business analysis and you so the way AI

is reading those business process Maps

it will it won't give you very good user

stories and what we've discovered over

the last couple of months of using El

elements GPT is we looked at the user

story go oh that's not very good that

was GPT and then we went back and looked

at the process map that wasn't a very

good we actually didn't set it up very

well how often are we actually giving

poor user stories to our development

teams because we haven't thought about

the business analysis so it's suddenly

becoming making business analysis way

more important because now ai is reading

that documentation we're creating it's

not just a user and quite often when I'm

presenting when I I talk about um AI

reading either process documentation or

even metadata descriptions is it going

to be

confused or disappointed no longer have

we got a a a an individual or person

there think understanding the Nuance

understanding the assumptions

understanding those weird acronyms that

are specific to your company when AI is

now reading your business analysis

documentation it's reading it literally

literally and the the results are based

on us following some well-known

standards so UPN for process mapping um

say uh the user story has a set format

erds entity relationship diagrams data

flow diagrams all have some standard

industry formats we need to be following

those and as an organization we created

something called um MDD metadata

description definitions so if you're

going to document your metadata what

let's work out the best way of

documenting so a I can read it because

what we've discovered is with good

descriptions it can come up with some

really good recommendations with no

descriptions it's okay but it's not

nearly as good as when it's got good

descriptions so the last the last bullet

point on their documentation that

suddenly become more important on the

quality of the documentation

architecture again always important but

suddenly the architecture of our systems

is is not just our internal Salesforce

system and maybe the the uh the other

internal systems that we it connects to

but if we're now touching a foundational

Model A large language model suddenly

that that architecture got way more

complicated Salesforce is putting in

place the Einstein trust layer so that

we've got some confidence that our data

isn't being sent to a large language

model but we're now relying on yet

another moving part inside the whole of

our ecos inside of inside our it stack

plus of course we're using a large

language model which is potentially

outside or we have our own large

language model which we bring inside so

again there are a lot more moving parts

and understanding the architecture of

what we're building and and if we make

changes to any of those and the

interdependencies becomes really

important these these these things we're

talking about we should be doing anyway

it's just AI is forcing us to do them

we're realizing we haven't done a

particularly good standard um yeah we've

got all these fields but actually have

we got decent data governance on the

critical fields that are maybe in our

dashboards that our executives are

looking at have we even done that

probably not AI is going to be looking

at those same same Fields so I I also

because I've been around Chang projects

for 20 years I'm always looking for that

Catalyst that reason for executives to

go yeah we should fund this yes we will

support you we will give you the

sponsorship and I feel AI is that that

point where we can go if we're going to

benefit from AI

we've got to get our act together in

certain areas and this is the

opportunity to do to get the buy from

senior Executives to do the things we

knew we wanted to do but we never got

the time to do oh documentation we

haven't got time for that now we have oh

don't worry about architecture we'll fix

that later you can't yeah um well don't

worry about there's that classic cartoon

which is you start coding and I'll go

I'll go and ask the users what they need

big issue is that people don't spend

enough time understanding what they

really should be building before they

build it and hopefully will shorten the

time it takes to help understand what

should be built and then we can use this

as the excuse to get projects funded

correctly so let's turn our turn our

thoughts to the center of excellence and

what this means um you've heard probably

enough a man being but let me just go

and I've was looked at these these these

13 pillars of a center of excellence so

before everyone goes well hang on we

haven't got 13 people we can't have 13

pillars again these are a set of skills

that you need inside your organization

around a COA and we let me just run

through them very quickly so Vision so

who's driving the Strategic Vision the

direction for Salesforce from both the

business and the IT perspective and

that's different from leadership

leadership is the steering committee the

key sponsors who again are then

validating and setting that direction

governance is about now the business

case the investment uh risk management

is about the overall control of the

Strategic Direction which is different

from change control which is a lot more

tactical that's management of changes to

all all aspects of the program whether

that's code changes or changes to

training material or changes to or org

charts uh methodology okay this is your

implementation methodology and that's

how I mean covers like people process

technology so think about it as the

business analysis piece devops adoption

monitoring so that how do you drive

changes you don't Implement once but how

do you drive changes around your that

cycle uh standards is I know we just

talked about documentation standards and

business analysis but it's metadata

naming it's coding standards testing

standards uh standards for training um

uh standards for change management so

again setting some standards that make

it easy so people aren't writing

documents and recreating Things From

First principles uh metadata management

okay uh every Salesforce now is run on

metadata if you don't have a good handle

on that metadata management is as

important as as managing the code um

architecture clearly we're now building

complex uh systems which are

interdependent uh and we need to make

sure we think about the technical

architecture but also how it relates to

the Integrated Systems and then security

which leads me to security um security

needs to be architected in it's it's way

harder to try and actually do it after

the fact and we need to think about

security and performance and

architecture hand in hand um change

management okay this is the people

change management as opposed to the code

change management which I think comes

under change control so this is about

how do you heart change the hearts and

Minds how do you think about

organizational structure how do you make

sure the training is there to to get to

put the skills in place not just

Salesforce skills but I know the domain

skills the business skills and and

obviously know think about AI skills

uh to a pmo so this is your project

program management office or project

management office which is managing the

overall program making sure that we're

delivering against the targets um earn

that earn value um tooling okay what are

the platforms for tools that we use not

everything that you need comes in the

box when you can when you when you buy

Salesforce you need business analysis

tools data quality uh devops um backup

store so there's a set of tooling that

you need to run Salesforce um and they

either need to be built or you need to

be buying them from third parties um and

the last one on the list is innovation

so how do you Foster

Innovation that and spot it that's in

certain areas of your organization and

org and how do you then build that up

operationalize it and then make reapply

it across other orgs if you've got a

multiple org implementation or just make

sure that you're in get getting the

collaboration to get that Innovation and

make it

operational that's a huge list I've gone

through and many of you going fun we're

we're a we're a relatively small or this

feels like Overkill so what we've done

is we've put together what we think are

the most important items if you're a

small and by small it doesn't

necessarily mean not very many users you

could be a small wealth management

operation a VC a private Equity Firm not

very many years users but actually the

complexity means you essentially fall

into the large bucket so it's sort of

small low complex versus large High

complex and we've looked at this and try

to go okay which things do you have to

have in

place if you're a um if if you're at the

lower level of complexity so some

leadership metadata management

architecture security and so on um

obviously when you get to the top end

and you're You' expect to have all of

those all of those aspects in place

which ones I think are impacted by AI

directly and I the reason I I want to do

this is some of those things where if

you're small a smaller maybe that's

suddenly because of AI they they're now

they're now fitting into the mandatory

box if we are starting to use AI to

maybe write emails inside a uh inside a

page layout for sales May suddenly that

governance has become way more important

than actually just the whole idea of

actually we're not doing anything very

complicated because we've only got five

10 users or 50 users or 100 users fine

we got some email templates but suddenly

with AI Change Control governance the

methodology about how we're going to

control those prompts um metadata

management all of those things have

suddenly become way more important

because a it's forcing us to put in

those disciplines in place but secondly

we've actually we we've added a whole

set of extra risks into our org because

we've got prompts hitting a a large

language model which may which is

outside our control if you let it get

out of control it can kill reput

reputation so quickly so I think one of

the things on the bottom there

Innovation is typically oh only when

we're a large organization can we think

about Innovation I think that Innovation

piece is really important I mean we're

not a huge organization St Cloud but we

had an all hands call last you yesterday

Thursday and the topic was Ai and what

we wanted to do was not tell our

employees how AI works or educate them

but instead it was an open forum and

said okay let's go around each of the

different business units and talk about

how you're using Ai and what's working

what's not working so we actually used

it as a collaborative forum and I think

the the role of a Coe in terms of a AI

first of all is acting as that

Innovation Hub how do you get different

teams to collaborate

because even if you say our organization

is not using AI you know that

individuals are you know that people are

playing in the margins they they're if

they're not thinking about it they're

about to think about it there are people

outside looking using testing playing

and the center of excellence is the

perfect way to get those people together

and find out who are the people who are

passionate and out there doing things

versus those people are sitting on the

sidelines going maybe it won't work for

me so I think in amongst all the hype we

need to try and find what the sweet

spots are because AI won't solve every

problem for every person and I think I

the way I'm seeing it is we need to find

the intersection between is there a

really strong return on investment can I

get a 50x a 20x Improvement so with us

we're seeing 100x Improvement building

those user stories but the second one is

can it actually be

implemented if you've got to have

gigabytes of data data that's all

perfect then that's probably not

achievable with our again back to our

story if you can write a process

map boxes and lines in the UPN standard

you can immediately get that benefit so

can we find a sweet spot where it can be

easily adopted and then the I think the

other piece is and we we tucked on it

slightly earlier which is are the

results consistent enough to be to be

usable there's no point going oh 60% of

the time it gets the right answer if you

have to check every single line of every

then probably not worth using and

another a sweet spot there is code it's

very good because it's hoovered up lots

of code it's quite good at writing code

it's actually quite good at looking at

legal documents because it's got of

legal documents that it's read and

therefore it's quite good at

interpreting them so there are certain

places you can look to where it's very

good I mean writing lyrics for songs

because there's no right answer for a

song again not relevant to us but for

there are certain places where there is

that sweet spot and that's where that's

where I think open a and chat gbt is

taken off but as as C leaders we need to

start thinking about our Innovation Hub

how do we start to find those sweet

spots