The Agentic Future: Preparing Institutions & Organizations for Autonomous AI transcript

I'm excited to get started.

For this last session, we're going to take a look at a future where AI increasingly takes on more complex tasks alongside of us.

In The Agentic Future, Sebastian Weir is going to explore how agentic AI is beginning to reshape both institutions and organizations.

And also not only that, but what leaders need to think about when it comes to things like governance, people, processes, and culture.

Sebastian is IB, Ms.

UK data and AI transformation leader.

He has well over 15 years of experience delivering thoughtful, impactful change through innovative technologies.

His work across industries has given him a deep, deep understanding of what effective AI adoption looks like in practice.

So please everyone in the chat, give a warm welcome to Sebastian Weir to future learning live and, and we'll see you later.

Thanks, Vince and, and hi everyone.

Good morning, good evening, good afternoon from wherever you are.

Thank you for joining the session and and thank you Pearson for inviting us to to be a part of what's a great learning event.

I'm going to spend the next or so just talking through some of our experience as an organization and learning on how we are helping our own organization, other organizations prepare for what's becoming an increasingly autonomous application of AI.

We'll leave some time at the end for for some questions and discussions.

So please, if there's any questions that you have placed in the chat, we'll try and get to as many as we can by the end of the session.

So to begin with, I just want to talk about the external impact that AI is having.

So AI is is no longer something that is just happening to our organisations that that we work in.

Then it is something that is happening internally too.

And it's a really difficult conflict that's changing the way in which organizations show up in the market, but also the expectation that we have as individuals within the workplace on on the work that we have to do.

It's the changing the way that we think about making decisions and how we make sure that we are still accountable.

And I'll come back to that concept around how do we make sure we remain accountable for the decisions that we make every day?

It informs how we shape the judgement that we make on how we work and how we perform and how we use the tools that are becoming more and more pervasive in our day-to-day lives.

And then it's also changing the way in which we think about creativity, how we use these tools to assist what we're doing and really kind of objectively assess the application of the technology.

And changing what we do is critical to thinking about how we can use and become work alongside these tools in an effective way.

We spend a lot of time thinking about how does this Dr.

productivity or changing in the way in which we perform our jobs.

But crucially, the role that we play as individuals is that creative aspect that we bring.

Whilst lots of these technologies and tools are generative, they are generative of the basis of information that is already ingrained in a in a model.

The creativity comes from the uniqueness that we bring and how do we bring that and apply it to the technology to change the way we want to perform.

And this is a critical way for organisations to think about acting and operating differently as we move through what just thinking about fundamentally what's changed in the last 12 months or so, we've crossed over from what's been increasingly generative AI.

That is something that's providing us with answers, right?

We've talked previously about how that changes what we call the cognitive load.

How does that change from me being able to get the answers to the questions that I want to develop?

There's a response to prompts and produces information very, very quickly from a variety of different informations that drives the decisions that we make.

The skills that we've needed to be able to use and deploy to be able to get the right answers at the right time have changed.

Three or four years ago we were talking about the advent of prompt engineering as a core and critical skill and, and a job market was evolving quickly such that prompt engineering was a talent that we were seeking very quickly.

That's evolved to becoming the type of skill that everybody needs in their day-to-day work.

It's, it's basic, like using the Internet.

It becomes the lifeblood of how you go about getting value from these models and how we therefore assess the information that's coming back to be really effective in what's coming through and and decide the role that we are going to play with the technology.

The shift that's taking place is that we are now talking about agentsic AI, and that is AI that has the potential to act on our behalf and provides agency to what it does.

That then allows us to think about autonomy, acting autonomously towards an outcome, towards a goal, defining a plan, allowing the agent to the AI to then navigate a set of systems based on the data it has to execute steps and drive an outcome.

Now, certainly that shift is quite important in that it's assisting us to get an answer to performing an outcome for us.

And how do we make sure that outcome is the one that we would have derived or would have arrived at had we been given that information in the 1st place?

And is it using the right information to make sure that the outcome it gets to is one that we trust?

And we'll then pick the steps that it goes through and how organizations need to think about really understanding the transparency of these steps to make sure that at the outcome that you get to is one that you can trust.

Because at the end of the day, an organization's application of this technology puts its brand and its reputation on the line on the outcome and relationship that it drives.

And if you cannot account for the decisions that the systems are making, then how truly can you take the responsibility and the risk that the application of AI is placing on your organization?

When you think about the impact that this has, it now starts to really shift the operating model as an organization, and this is happening incredibly quickly.

The problem isn't no longer about technology advancement, it is about our ability to change as individuals, and our ability has changed with organizations to adapt around how technology works.

This is no longer just setting tasks, it's about making decisions, identifying patterns, sequences, and operating differently.

As soon as you start to apply that technology to a traditional organization, you, you amplify silos, You precipitate bias and the lack of transparency that may exist in data.

And ultimately you start to amplify the cost that's being invested in this technology without any real meaningful change or return on how you're applying it.

Number of statistics here around how we're thinking and seeing executives starting to look at breaking down boundaries of an organization to operate differently.

We talk about channels of engagement and how they are becoming increasingly unified because the technology again, can understand across any type of medium, any type of mode of engagement, what I'm asking for, any language.

We've got people around the world listening to this right now, and the models can be translating information in real time to make sure that this is hugely accessible and open to the way in which people can engage.

But equally, we recognize that for organizations to really assume that the potential fundamental changes is needed to be able to think about how you deliver change differently, how you deliver services differently with the power that this represents.

And a lot of talk at the moment is around how do you reinvent, how do you redefine what you're going to do rather than applying A gentic AI to a set of processes or tasks that are ineffective or inefficient.

We've been through that cycle with advent of RPA and robotic process automation.

And if you automate a bad process, you still get a bad outcome, you just get there quicker.

Actually, the advent of the Gentic AI allows us to be much more thoughtful around what are the activities we want the system perform and how do we use that insight to make sure that the human element is amplified and differentiates the service that we're providing as a business we've done once too far.

So this is where we're thinking about the movement to what we call an interact interconnected enterprise.

And it's really important that at the very top of this is people, right?

We are not trying to displace an organization that is becoming agentic and therefore the role of people as being pushed it down in an organization's priority.

It is absolutely how do we amplify and improve the role of people in that organization?

How do you use work flows and data to drive better outcomes for the individuals, better outcomes for their end users?

How do you use the people around it to amplify what it is that's represented such a successful brand over the over a number of years?

And how do you then simplify and complement that with the agent technology from meaningful impact?

That means that a material shift in how you think about measuring what you are trying to get to is an outcome.

It's no longer about average handling time of a call centre or is no longer about impact and and navigations to a website.

It's about why are customers navigating to it, what are they trying to achieve and how can you better anticipate and support them in that journey.

With the data and understanding that you have, you then start to then shift the role that we see that people and AI will come together for.

This is about Co working and not displacement of work.

day-to-day activities start to shift.

There is increasing roles that AI is playing to assert productivity, right?

This, this meeting is being recorded, it's being transcribed.

It's a fantastic opportunity to use it to better understand and summarize what's going on.

It doesn't change the need to listen and necessarily understand and interpret what it is that we're talking about, but there's efficiencies to be gained by applying the technology effectively maintaining control and autonomy.

So how then do you make sure that you're elevating the decision making process and not becoming increasingly complacent with the role that AI plays and how it is interpreting what is going on and recommending outcomes, recommending or even taking decisions?

Ultimately, the accountability still remains with the leadership of every organization that's there.

In many regulated organizations around the world, boards are both personally and professionally liable for the outcomes of their consumers, their reputation, and the reputation of their business is at risk of getting that wrong.

So understanding the application of the technology, what you're willing to give agency over and how to control it, and then how do you make sure that you're doing so the in the interests of both the organization, the people that make that up and the end consumers is how you then dictate control.

You're thinking about the collaboration and we'll come on to this topic in a moment, but ultimately, how do you create and relief reduce the friction that exists between the application of AI and the use of these tools and the roles that you or I will play in the use of it?

There is an inevitable friction and frustration sometimes and an apprehension in how AI is going to be used, but actually navigating how to make sure that there's a complementary in the way in which you're driving outcomes and driving the changing behaviours, creating capacity for your teams to think and operate differently.

It's going to be really important to make sure that AI is seen as an additive capability and not a displacement capability to the workforce and certainly where we've seen it effectively applied.

It's about elevating the capability that you or I have in the day-to-day work that we do, not trying to displace it by applying it to mundane repeatable tasks.

That's the things we don't want to be doing.

We want to be elevating and providing our judgement on the outcome that's then driving to excellent outcomes.

It creates the capacity to think differently around problem statements, thinks differently around how we are getting to that outcome and applying that judgement consistently to make sure that you have all of the information available before you do it.

We work closely with a number of financial institutions around the world where consumer duty and customer duty, customer outcomes are, are incredibly important.

They have real meaning on the end user.

And making sure that you're able to consistently achieve that successful outcome repeatedly and at the scale that these organizations need to operate at is incredibly important to make sure that they maintain that responsibility and trust.

And then ultimately this comes down to to changing in roles and changing as leaders.

We talked in and the parallel conversations already today about the importance of AI literacy, but making sure that leaders of organisations understand the role that AI is going to play on their organisations.

The sympathetic way in which it can be applied and effectively Dr.

change positively to the organization is important.

But also understanding what it is that AI is not going to do and ensure that it's an, an effective tool in an organization's arsenal to, to effect change, not something that is used to be able to just demonstrate or publicize advancement.

We then come to thinking about how our work changes.

So we focus on outcomes.

We want to spend less time executing steps.

We want to spend more time setting the, the goals, the process and understanding how do we make sure we're applying our judgement because that is what differentiates us and what's going forward.

And actually, you think about how we're using it.

Actually, when you're using things like transcription technologies, it reduces time of taking minutes for a meeting.

But quite often reflecting on that meeting is the important aspect that determines why you met in the 1st place.

How are you going forward and therefore actually spending the time to review what's been taken.

It's not about the exercise of writing the minutes I often find, It's actually reflecting on what I sat in that meeting for, why I was there and what I want to do differently.

We met for a reason.

How do we use that to drive the actions and purpose forward to move forward?

Huge efficiency gain in in the application of the technology.

But making sure my judgement is there and how to be able to move forward is an important aspect of how we start to drive change in the working behaviours and working patterns of everyone.

So we proceed, that promise of productivity come through and I want to spend a little bit of time and thinking to the controls.

So a lot of these things are underpinned on how we have underpinned our AI development as an organization since the since the advent of the term, we've always advocated for transparent and observable technology.

How do you make sure that you're using the data to identify objectively what is going on?

And is that something that you would do as an individual's?

Is it something that you can abide by and it's something that you trust as an outcome?

How do you use that to objectively make a decision?

Make sure that also we're not becoming increasingly lazy and how we're allowing the AI to operate because it was right the last time doesn't mean it will be right the next time.

And making sure that observability informs the decisions that you're going to make.

The measurements of quality and evaluation and the execution of tasks that are taking place often inform the guardrails that are putting in place.

These aren't processes that you need to assess.

These are functional things that underpin how an agent performs, the tolerances that allow you to move forward, and then also how you make sure that you have policy and procedures to be able to intercept when things go awry.

They will quite often go off track.

How do you make sure that you recognize the symptoms of that taking place and interact at the right point in time allows you to make sure that you're empowered in how you're controlling the system rather than just consuming the output that it's being generated?

That really starts to then shift the role in which we are thinking about.

For a number of years we've been talking about it's really important to have AI in the loop, right?

And that's been the way in which we perceived control to be the right way of doing it.

But inherently, when you think about that, in the loop implies that we are now stopping or pausing or providing friction to what would otherwise or otherwise be a seamless execution.

We're now shifting to this concept of on the loop and actually elevating the role of of the human, the individual.

And our role there is to review what is taking place in these agent, in the agentic systems.

Our job is reviewing and ensuring that they continue to operate as we've dictated, as we've outlined and as we want.

We're applying the judgement that we would have been on the left hand side as at certain stages, but applying those constantly to make sure that we are reviewing the boundaries, the thresholds and the processes that are in place.

And then ultimately not just waiting till the end to make a final decision, reviewing every executional step of these processes and plans that are taking place.

And our role is elevated in that to make sure that we are the supervisor or the accountable individual making decisions for the processes that AI is starting to complement.

You then move into the, the, the changing dynamic of an organization.

We have elevated the role of the individual.

We have increasing amounts of agent supervisors.

They are not just reviewing what agents are doing.

They are building the technology so they have an intimate understanding of how they're working, the tools they're exposed to, the data they're understanding.

And that is a completely different skill added to what they're already providing, where they are also necessarily a manager to individuals or humans.

And that blend of skill is going to be increasingly powerful, increasingly differentiated for those organizations advancing where agentic systems are becoming second nature to the actual under bedding processing of an organization.

Finally, as the work changes, leadership activities need to change too.

We need to be thinking about how do we change measures of an organization, the behaviours of an organization to be able to act differently.

There's a lot of huge opportunity, but it can't be a blunt instrument to be able to drive change.

There has to be an identification of what do we want to be doing differently as organizations.

How does that translate into measurable KPIs that we want to see to make sure that we make, maintain and drive the right type of customer outcome, consumer outcome that comes?

And how does that then translate into the changing room behaviour change that we want to see in an organization?

For the last few years, we've seen lots of organizations buy licenses to lots of technology and then can complain that very little changes has happened.

The translation of investing in AIAI at its core can be is simply a cost until it's adopted sympathetically and driven through change.

The real leadership, that is how you change the way in which the business performs driving value out of its investment.

That trust then comes through and translates into the brand and reputation of the business that you're representing, and you start to get the opportunity to really think differently about what work you want to be doing and what outcomes Dr.

success for that business.

But until the leadership of that organization really can grapple with what AI can and can't do, how it works, and how the organization needs to adapt effectively to do so, there will be constant friction between investment of AI and meaningful change ROI that we've seen a lot of challenges over getting sustainable return on the application and investment being put into the technology.

I'll finish before we open up for some questions on some of the critical paths and things that we're starting to see.

Five steps to help manage this complexity.

Thinking about how do you look at ownership of AI, much like we've talked about ownership of data, ownership of applications, ownership of things, ownership of risk, accountability on who owns the application of AI, who helps you understand the risk and the the role that it's going to play.

Understanding how it is driving value because as I mentioned, in isolation, it is a cost.

It has to deliver an outcome and that outcome can be translated into meaningful value.

It doesn't always just have to represent productivity and cost.

It is an opportunity to differentiate the business and a meaningful growth opportunity to drive incremental value.

For lots of businesses, governance remains critical.

So embedding controls and auditability across that to ensure that there is clear accountability on the decisions that are going in place and how we are effectively delivers change.

Everyone on on the chat will be very aware of the security and resiliency considerations that go into the application of AI.

It's impossible to to look at the news without recognizing the the threat that sits in how AI can is evolving.

But equally, that is an opportunity to think about how do you protect your organizations?

How do you use this technology responsibly in a way that ensures that you are anticipating the both the the advent of security concerns, but also protecting the organization and the responsible application of technology and then ultimately moving to this concept of being AI first understanding what that really means.

That isn't that you put AI before everything else, but it's understanding how you think about the world whilst putting AI and technology at the forefront of delivering change and understanding what it takes to independent architecture to deliver that change.

With that note, I'm going to hand back to Vince, who I think is going to facilitate a little bit of this question and answers.

I hope that's been an interesting presentation and I welcome any questions in the chat.

Thank.

Thank you so much Sebastian.

You know, that was a great talk and I see the chats been buzzing with so many different questions coming in.

I've been trying to keep up.

We've been doing our best to keep up.

So I've managed to pull a few.

Before I get on to the questions, I just wanted to say, you know, you had one slide, I think it was slide six or something where you asked 5 questions about AI becoming a Co worker.

And I thought that was really such a powerful way of framing what comes next that I wrote down what they were myself.

You know, how might the day-to-day change?

How can we maintain control?

How might humans and AI collaborate and how can we ensure excellent outcomes?

And finally, how how do our roles as leaders change and evolve?

And what I really liked about that slide and those questions that you posed is that, you know, it, it moves the conversation.

We've been so hyper focused on like model evolution and, and you know, what's its scoring on this assessment and benchmark and things like that.

But it, it moves the conversation beyond just the technology and the benchmarks.

It, it forces us to think about how work is really done, how decision making happens, how accountability manifests, how trust and leadership need to evolve as AI becomes ever more deeply embedded in our organization.

So that, that really made me stop and think, I appreciated that.

And I think for education leaders in particular, you know, that there's an important parallel that's ongoing.

It's, you know, if this is the world where students are graduating into, we need to really pause and, and take some time to think about how we prepare them, you know, not just how to use AI, but to work alongside of it thoughtfully and responsibly and, and effectively.

So I really appreciate that conversation.

I was able to pull a few questions to from a wrap for people around the world.

So the first question came in from OB based in the UK.

OB says AI is clearly here to stay.

As organizations begin adopting more autonomous forms of AI, how can they manage it transparently?

And importantly, what are the consequences of failing to govern and regulate it ethically?

Right?

Great question.

Thank you, OB.

And I'll pick up the first point around transparency, right?

The, the great thing about AI agency is there's a huge amount of data in the decision making that goes in.

We may not necessarily agree with the outcome, but they, the transparency means that we can understand to a degree what process they've gone through to derive the decision and step they go on to take.

And we can intercept that.

What's quite often difficult is when you or I make a decision, we're inherently subjective and it's very difficult to get that level of objectivity on how we go about making our decisions.

There's lots of things that inform it, but actually when it comes to AI, there's an opportunity to use data to be much more deliberate and objective around how we make decisions.

But that comes to the transparency and the processes that underpin it and the data that's in place.

Now, there are ways of doing that and it's making challenging how you how you go about applying the technology effectively and flipping on to the second part of the question.

When you get that wrong, the brand and reputation of the organization is at risk.

Because as soon as you're unable to account for the decisions that AI is going on to make, then actually the the reputational damage that that can do is significant.

Because one day it might be a small decision, but that could be amplified incredibly quickly for the way in which AI is so pervasive around organizations.

And there is a huge amount of risk associated with brand trust and loyalty from a consumer that will not return if you get this wrong.

Yeah, that that's a really important point about, you know, I'm, I'm sure we've seen a few examples, but there will be others that happen.

So we definitely don't want to get this wrong and lose the, the trust and human connection that we built up with our with our customers and and students, for example, of us an education organization.

Second question has come in from Doctor Ross, who's based in the Netherlands.

They ask as universities adopt more autonomous AI systems, what evidence should they require to show that AI supported learning is strengthening rather than displacing a student judgement?

And second part of that question and where should the accountability ultimately reside within that, within that university ecosystem?

Great.

Yeah, it's a it's a really interesting space, right?

And one thing that I've seen with a lot of the students that we work with is an organization and, and a number of on one of the apprentices actually I was sat down with a week or so ago.

We've seen the advent of what they termed as the traffic light system in, in that process.

So what they're able to do by doing so is a red light, the red, green amber system in, in the assignments that they get given, right.

So they've got very clear boundaries when they set the expectation of those students to say if it's a red light assignment, you're not to use AI, right?

This is about independent port, this is about independent activity.

And we want to see that come through.

If it's an amber assignment, there are practical tools that we're going to give you, but we're going to be very prescriptive about the use of the technology for very given, very deliberate reasons.

And then there are assignments that agree just free reign, you use what you think is right to get to the outcome.

And actually that flexibility of being able to allow students to experiment in a Safeway, but understand very clearly where the boundaries are in the application of the technology is really important.

And, and Vince, to your point around readying people for the work environment, I think that's really important to make sure that we're really deliberate in how we put accountability onto the individual for the tools that they're going to use, the way in which they're consuming it and the outcomes that they're putting the assessments that they're providing.

11 interesting conflict.

The, the kind of reflection on that though, is lots of education organizations were incredibly quick to, to restrict the use of AI when it first came out.

They cut the barriers up and, and that kind of put a kind of subverted use of AI.

People went underground to try and use it.

And not all organizations came the other side of that problem and said, OK, we're now accepting the use of AI in these responsible ways.

And when you go through that, that behavioural change is incredibly difficult to unpick.

As soon as you've told a whole cohort of people they cannot use these things.

They are unclear on where the boundaries start to become released and when they can start using it responsibly.

And that for education and organisations is particularly important around how do you encourage the responsible use?

Because if you don't do that, you inadvertently encourage the irresponsible use, and that can be really dangerous.

Yeah, that's great.

I I love the stop light.

Not even analogy, just the ACT is a practical implementation of that.

There's you can, you know, professors, educators here can look into that.

It's something I use in my course because I think early on in the AI debate, two years ago, maybe even three years ago, we got kind of caught in this push and pull of whether to ban AI or whether to allow AI.

And I think it's much more we've learned, it's much more nuanced than that.

It's much more of a continuum.

And I know in my courses that I teach, it's not about banning AI or allowing AI.

It's, it's I have 5 assignments in my introduction to kinesiology course.

And I go through each and I think about, you know, in which assignments do I want to allow AI to be used openly?

What assignments there is not about?

And what is more of that where clear boundaries and clear use cases of AI are permitted that more of that amber analogy.

So I encourage everyone to look into that.

Pearson has done some work on if you Google Pearson's future of assessment, there's a paper that we've done on that that includes that stoplight model.

Third question came in from Mary based in the US.

How do we move from from AI that can behave like a sycophant catering to what it thinks we need to hear and I hey, I love being told it's I'm I'm right.

What a great idea.

And Oh my God, that was such a great suggestion for dinner.

But we we also have to be realistic and we want the truth, right?

We want to get to what's most important.

So how can those agents move from being, it's kind of that sycophantic behaviour to agents that can be really trusted to deliver honest, unbiased information gathering and and support our decision making in the best way possible.

Yeah, it's really interesting, right?

As you've asked those questions, I've gone, it's a great question, right?

I've almost imitated what we've seen in the cyclantic way in which AI models operate, right?

So you can see why these these systems operate the way they do.

But in order to start to move away from that, what what we're saying is it becomes a skill problem.

And it's actually critical thinking, right?

Because it's incredibly gratifying to see that validation when you get the response because of the way in which the system provides it.

But actually that critical thinking and challenging what's coming back to make sure that you are remain, you remain accountable, you remain the person making the final decision.

And this is augmenting the way in which you get to that decision making process.

And ensuring that actually you are critical about that system is something in the short term is really important because I'm sure we've all done it.

You go and challenge an answer from energetic system and it will quite quickly flip to a different kind of perspective.

And if you challenge it again, it can flip back.

And actually doing that exercise is a really important way to understand whether or not it's using the right data to get to the right outcome or it's just telling you what you want to hear and in in Google or in your search engine in your day-to-day lives, that's fine.

When you're making material decisions as an organization, as an educator, it's critical to get that right.

And critical thinking increasingly is the the one skill that we see is important.

And it's an incredibly difficult thing to advocate and maintain and and kind of develop as, as as an individual.

Yeah, totally.

I want to squeeze in one more question before I give everyone a couple minutes back to, to get a coffee or, or use the bathroom before we end our closing keynote.

This question came in in the chat from Erica and I thought it was really insightful.

Erica asked as leaders, how can we help our teams shift from just using AI to shortcut right to to an answer shift from using it just a shortcut to an answer for more shift towards more like action oriented goals, right.

So I think using it probably is asking more of like embedded into the decision making process and collaborating with it rather than just like shortcutting it for the the final solution.

I think the most important aspect to that as a leader is to demonstrate how you are doing it right.

Really embodying that type of way in which you're expecting your teams to work is critical because this is a change where actually it's very difficult sometimes if you haven't done that, to understand how it's going to work for for your teams.

And really amplifying the way in which you're expecting, how you are using it as an individual, as a leader and how it's changing the way you are working is the most impactful way to be able to drive change into your teams.

Because this is about we are all understanding and learning how our world is going to change with this technology.

And it's, it's everyone's learning journey and making sure that as an honest leader, you're demonstrating how you're doing that is really important.

Then your teams will feel like they can also experiment on they can go ahead and do that change was they've seen that from leadership and that honesty is a really important point of transparency between teams and and the teams you lead.

Yeah, that's, I'm going to echo what Felipe said in the Felipe Costa said great answer.

And I think that was a really great answer in terms of modelling the behaviour that you want to see implemented and cascaded down to your organization.

It is the best way to do it.

So many people are are incredibly appreciative of of your time, Sebastian, and so many people want to see the slides and we do have the slides available for download on the in the Pearson Zoom lobby.

So thank you for making those available as well.

So please join me in thanking Sebastian for his time and his perspective.

Thank you all again.

Bye.

Thank you everyone.