From 'we should do AI' to actually doing it

00:00:02: Welcome to Designing AI Heroes, where AI and people align.

00:00:33: Welcome to another episode of the Designing Areas podcast.

00:00:37: Today I'm joined by Paul Smith, the CAO at MS International And he's also recognized as global CIO one hundred, World CIO two-hundred and UK CIO One Hundred.

00:00:51: So I think we are up for a very insightful conversation today.

00:00:54: Paul great to have you here To start could share where your organization stands with AI Today?

00:01:01: What is top of mind right now regarding AI?

00:01:06: Yeah so i guess the first place to start would be probably say We're still quite split I think in terms of those that are really super excited about very new technology and its potential benefits, where it can help versus some of those who have a bit more concerned about the potential risk.

00:01:22: And the harms of their technology... ...I think as an organisation we're cautious ethically aligned with our adoption of technologies specifically this type of technology but were extremely consultative right across the organisation in terms of getting a full view on what people care about, and where the opportunities may lay.

00:01:43: So we're very careful around the selection of our technologies as a model due to those concerns... ...and our external policies and positioning which would need to be mindful of what we think outwardly towards the world.

00:01:56: We need to eat out our own dog food internally.

00:01:58: based on that too We're still capacity building, like many organisations I guess the learning never stops.

00:02:05: The technology is moving so fast at that moment.

00:02:07: So capacity-building for us is a real three hundred and sixty degree thing.

00:02:12: It's communications it's education its sharing lesson learn.

00:02:17: Its about providing safe spaces to talk & ask questions.

00:02:20: but thats an ongoing thing.

00:02:23: What's running for us, I guess.

00:02:26: We have a use case-censored adoption strategy and that makes sure we can control risk and benefit realisation in very deliberate way.

00:02:36: And provided an approved capability to everyone right at the outset about two years ago.

00:02:46: That is already delivering value across people processes and business capabilities impact levels And there are lots of considerations in that around what we call as offensive, in terms of our ability to move our strategy forward or to enable our strategy versus what could be termed defensive which is about resilience and improving the organization's resilience.

00:03:06: So that can be sustainability.

00:03:08: it can information security whatever lens you apply.

00:03:13: We talked a bit about continuing education awareness...we're running cross-functional pilots with a team, feeding back into the backlog of beneficial use cases and that's identified along our core value chain.

00:03:27: And they're all queued for investment in scale up.

00:03:30: What are we thinking about?

00:03:32: I think as i said how to keep across everything.

00:03:34: there is so much it really is moving fast As non-profit funding our appetite to innovate responsibly through technology.

00:03:49: Our roadmap is agentic, but we're very conscious of geopolitics US type lock-in and digital sovereignty right now.

00:03:58: And associated with that I guess there were the realistic ethical and comparable alternatives And I guess last but not least the EUAI Act, so ensuring we are continuing to meet our responsibilities.

00:04:11: Great!

00:04:19: I like this because when you map first the opportunities and identify the communities of problems, and build use cases.

00:04:26: You also have a better adoption or you get results very fast.

00:04:31: so most organizations focus on technology and get exciting about technology.

00:04:37: but i like what you said.

00:04:39: And there's also my next question that I have how did we go from reshoots to AI?

00:04:45: to actually doing it, and what's yet the first practical step in your approach when you have a team that said okay we would like to use AI.

00:04:56: How do you start this AI approach?

00:04:59: Yeah I think that's a great question because i think many organizations are just faced with let's do AI without really knowing what they want to solve or the thing they're looking for.

00:05:09: so For us We shoot at high-level generative AI strategic intent documents than vision And that covered things like opportunities, risks and challenges we saw.

00:05:19: It provided a responsible AI framework, the high level approach and some potential strategic objectives for us in this space.

00:05:28: That helped to show some understanding as I say set of vision to show bit of control and start open conversation where hearts or minds are at and react accordingly.

00:05:41: We quickly provided, as I said a secure alternative to tools emerging in the market which in our eyes presented informational data leak avenues or were unethical to use.

00:05:52: So providing a safe, secure alternative was very important for us early on and we started it As i said with a bit of a three sixty view to adoption and learning And that initially started organically peer-to-peer so that was within our next turn also the organization but it was supported by very deliberate capacity building and communications, and education through virtual channels for engagement.

00:06:17: Through connect-and-learn sessions and through networking is sharing with others.

00:06:23: we issued an extremely highly consulted on policy for generative AI centered on ethics protection of human rights unsurprisingly prohibited use cases and the importance of a human in the loop.

00:06:41: Of course, that was augmented by all the existing policies we have around acceptable use of tech information security data governance because all these play into this technology.

00:06:51: We set out to revisit our frequently based on new things that we learned from landscapes at move.

00:06:57: And lastly I guess To start we established it cross functional team with very keen IT people Very keen business representatives That were focused on exploration of learning together and prototyping to validate assumptions, trying to push the boundaries off tech.

00:07:16: That was supported by a small consultancy budget and an external phone friend for when we got stuck.

00:07:21: Do

00:07:21: you have some space for experimentation?

00:07:28: Some teams come in together across one's explanation.

00:07:30: or do you sit down with your team really develop use cases mix button up top-down.

00:07:38: It's been both in truth, I mean when we started.

00:07:40: We had a high level idea of the things that we thought technology could do because what we've seen others with it and applied to our context.

00:07:49: but as we have learned together and explored and experimented together new opportunities ,new issues, new risks And as I've said, started to push the boundary of the tech itself.

00:08:00: So it's been great and its worked in a way that kind of small innovation piloting should work quick agile iterative learning you know?

00:08:08: It has been fantastic

00:08:12: Great!

00:08:13: When it comes yourself how on where do we use AI at your daily work A

00:08:19: lot more than what I expected.

00:08:21: to be honest with you Nadine, within the remit of our organisational policy.

00:08:26: I think i use it for personal augmentation right.

00:08:28: so whether that's creating summaries of long documents or personal research or initial drafts of stuff rewriting and rephrasing...I used as a sanding board.

00:08:39: actually by bounce ideas around off it explore and I use it to help educate myself, because sometimes i like things explained in different ways or i'm looking for an anecdote.

00:08:49: Or a different way to position something?

00:08:52: And I find that can get that super quickly!

00:08:54: And I could have a conversation and explore things but actually...I'll probably use it all the same ways as my personal life too.

00:09:01: My wife laughs at me because quite often she looks across at me every evening and sat having a conversation with Genitive AI on her phone talking about something really random trying to learn about something.

00:09:11: Well, it could be gardening.

00:09:12: It can be financial planning or anything?

00:09:15: Yeah great and back to your organization.

00:09:20: Can you share one example that made a real impact where you said oh wow this is something so helpful for us And we didn't expect much from them?

00:09:34: Yeah, but it's just not theory when you say okay we met the use cases.

00:09:38: We have some expectation about a result.

00:09:40: so... But really one example can be small and big son Just to-

00:09:46: I could probably talk a little bit more

00:09:48: about that.

00:09:48: Always people ask me about examples in use case or they are very keen on doing all of this!

00:09:54: I can probably give them couple within their kind of remit what i'm able to speak today.

00:10:00: We are an information-based organization, so our kind of production line is based around the Information Management Life Cycle because we do investigative journalism and research.

00:10:11: And that perfectly aligns with how information comes in or what you're doing with stuff... ...and where ever data exists there's no opportunity for tech to do something cool about it To deliver a leaping capability or speed or scale.

00:10:24: So there have been few identified actually ranging from support with improving the production of extremely large documents that can take over fifty people a number of months, to something as simple as a footnote agent which we do thousands off by the way in some of our documents but each have different formatting requirements based on the source.

00:10:50: Something that can automate that based on our policy saves us hundreds of hours per year.

00:10:56: and But wherever we are looking at data, look in information manipulating it presenting.

00:11:03: It there is opportunity for us that we've realized and on top of that you know all your kind of coordinated coordination style roles in the organization.

00:11:12: One area We didn't expect to see the benefit That we have Is as a very complex diverse multinational organization global organization even?

00:11:22: He's has an assistive aid or accessibility.

00:11:25: So we found that colleagues, maybe you don't have English as a first language or may be neurodiverse of particular education need.

00:11:34: We've had really big pull from our users for this and assistive aid to help them engage with the organisation Maybe speaking English in professional way Or more professionally.

00:11:47: And there's also been some real value around experimentation With things like policy agents to help our end users interact with what can quite organizational policies, which can be complex based on various scenarios and subject of a query containing hours for one pick.

00:12:06: But then being able to bring that the user twenty-four hours a day seven days per week in any language or natural language allowing real time discussion education iteration has been really big inclusion diversity when And again, it's probably added hundreds of hours of capacity back into the team.

00:12:27: Oh great yeah I see is also when you have a large complex document.

00:12:31: AI can be so helpful and also immediately help follow document routing or document routing to do right team?

00:12:41: Yeah that's also...I think technology not that complex but comes with that when you set it up, so people always say also complex technology but once come to document routing or understanding that's not.

00:12:57: That much complex when it comes from the technologies side absolutely

00:13:01: yeah.

00:13:01: But one year aggregating data information for many disparate online and offline sources some structure someone structured maybe cross referencing may be again cross-referencing.

00:13:11: that with things happened sixty years ago there is a real opportunity just to join the dots against things that you perhaps didn't see yourself.

00:13:19: So, that's where tools like this can give real speed and scale in order to learn the individual.

00:13:24: Great!

00:13:27: When you think about the AI journey of MS International or the journey today... ...or a journey past months with also years what was really surprising for your in a positive or negative way?

00:13:45: some main obstacles.

00:13:47: That also was very interesting, okay where the bottlenecks?

00:13:51: what we have to keep in mind when it comes with its data?

00:13:54: Was this a human side?

00:13:55: What is missing?

00:13:56: alignment approaches tools or what does really I think?

00:14:00: Okay there's both our main bottleneck uh In our journey.

00:14:05: Yeah great question.

00:14:06: Uh i-I think for me because Because of opportunity and risk around this sort technology.

00:14:12: It's ubiquitous, right?

00:14:14: You can touch any part of the organization at anytime.

00:14:17: I think there is a McKinsey Seven S' model which i often refer to Which is skills strategy structure shared values system staff and style.

00:14:27: it will touch on every aspect Of that organizational model.

00:14:31: And so for us specifically I think stakeholder alignment consultation.

00:14:35: as I said we're very diverse organization with different priorities Values beliefs perceptions, impacts risks opportunities and managing that aligning them in the recognition of people come from different starting positions.

00:14:50: In terms of their awareness.

00:14:51: there are fears they're hopes resources on.

00:14:55: I think educational when it's another is both ways.

00:14:58: but two IT and to our users because we have some very clever uses.

00:15:03: only learn a lot.

00:15:04: for my use this you know.

00:15:05: um i think on-the-ground impact I think finding new use cases and new issues, as we mentioned earlier.

00:15:13: Just experimenting realising values in ways that you didn't anticipate pushing the guardrails of platforms which for us legitimately need to do so.

00:15:23: We have to be able talk about torture About violence You know weapons attacks In the work they're doing And you can bump up to safety guard rails on some of these platforms.

00:15:33: I think the technology can be inconsistent.

00:15:37: It doesn't always do what it claims, or does not consistently.

00:15:44: Safe spaces to share and learn together being people-centric and user centric just providing people with ability to ask a question express a fear have conversation.

00:15:55: you know that has an argument thingy too.

00:15:57: but having in safe space where fear of retribution was important for information governance perspective The tech can bring information to the user.

00:16:07: They did not know they had access too.

00:16:10: Someone stored it in a wrong place or someone, and you could bring that to them.

00:16:13: So hence I use case-based approach really which helps to contain risk like that Rather than giving for example everyone under license And letting him have at It.

00:16:24: We've talked about speed of the Tech Hair.

00:16:26: the hell is anyone keeping up?

00:16:28: That's an issue for us too.

00:16:30: But we do have willing set of users You know, they want to get stuck in.

00:16:35: They wanna play with the technology and I talked a little bit earlier about that.

00:16:39: The ability of this kind of tech to pair accessibility which is really important for an organization like us that cares about equity you Know ensuring everyone can get gets where we need them to get too regardless of starting point.

00:16:52: Because because actually i don't ever recalling my career A technology That's ever had such a pull or appetite from user and I think since probably the advent of the internet, when everyone used to take turns on the Internet computer right.

00:17:06: And so for a technologist like me it's a dream to have that opportunity for transformation until make a difference.

00:17:13: but the challenge is doing that responsibly proving a benefit and doing this sustainably in really looking at what you could do versus what should be helped?

00:17:25: Yeah i think also AI not very different from other digital transformations.

00:17:33: So you always have the same identifying problems, aligning people staying up to technology.

00:17:42: It's in transformation itself but it is the same.

00:17:49: The same blame crown on any other digital transformation projects and organisations.

00:17:54: so not just integrating a tool by designing say, pitching technology organization also the people.

00:18:03: So you need every three aspects to design this journey here?

00:18:08: I

00:18:08: couldn't agree more.

00:18:08: it's really refreshing to hear someone talk about that because all i hear is let's just do it!

00:18:12: Let's go and spend money.

00:18:15: best practice for a reason.

00:18:17: change management exists for reasons.

00:18:19: benefits management exist for risk management planning.

00:18:23: so think you have find the right balance don't?

00:18:26: Yeah, I think it's the right balance.

00:18:27: I always like to say start small and design scale because you...I always encourage us starts to create momentum And people are very motivated when they do workshops or what can we do?

00:18:50: what are really the opportunities for us.

00:18:51: They're very motivated to learn technology, it's other than when we integrate.

00:18:58: so in the past that you have a new tool and there was not much motivation.

00:19:03: they had to learn on your tools or their technologies.

00:19:05: but everybody is curious about this technology.

00:19:09: So then start something small in a prototype.

00:19:12: You get results where fast And scale it.

00:19:16: But design to scale at a start.

00:19:19: So need, as I said in the right balance to keep this momentum with people and don't have these huge use cases and get something built and done some months.

00:19:31: you can start with something small then scale it really step-by-step.

00:19:36: so designing this journey for people yeah

00:19:40: hundred

00:19:41: percent great.

00:19:43: When you think about your approach way of working, we also say AI enforces its potential when you redesign workflows and redesign work.

00:19:54: Do you have an example for this?

00:19:57: Do it touch the workflow itself or is it integrating AI in existing workflows, existing tasks that you see?

00:20:08: We reinvented something with AI the different way, more innovative way.

00:20:16: Yeah I guess two parts in answer to that question...I guess what have i changed my approach?

00:20:21: and then kind of where do we see the impact level?

00:20:24: So I think it's in my own approach..i think- I listen first ,and learn .

00:20:30: And I see where can help.

00:20:32: before start leading don't just jump into feet and stop doing Because there's someone somewhere that has done this or knows something, and can help me already.

00:20:41: I think thats the first thing i learned particularly with technology And now links back to a challenge of just trying to keep abreast across it all by the moment Being agile in planning for disruption and having mechanisms deal with it.

00:20:54: You talked about small iterative steps Learned something early Be able stop change direction with that big commitment, you know?

00:21:02: I think that's really important.

00:21:03: but knowing your plans will change is an important place to start from.

00:21:09: Understanding what i don't and do not know.

00:21:14: so there's this constant learning process where you have to provide yourself space more than any other technology.

00:21:23: I've used my networks for these technologies you know, speaking to people like me.

00:21:29: Speaking to other organizations reading watching videos and more often than not.

00:21:34: this is kind of part my thinking processes.

00:21:36: now as I said because they use it help me brainstorm.

00:21:40: but i guess in terms of its impact on an organization...I've heard a lot of peers talk about it in three levels And the first one's on individuals' impact In that individual role They will find places To use their technology to help them Whether it's reading writing whatever.

00:21:57: The next level of impact is around the business process, and that can be automation.

00:22:02: It could be pair assisting a processing in some way but end-to-end process flow.

00:22:07: And then the third levels are on business capability things that absolutely create new capabilities That the organization could not do before or para particular capability there's reusable across the organization.

00:22:21: We talked about meeting coordination as well as one capability or drafting.

00:22:26: Maybe it's a way you can manage online data sources that you just couldn't do automatically before.

00:22:32: So there are different levels of impact, I think and clearly one can be linked to the other depending on the person who is affecting or their capability.

00:22:43: Would your recommendations start with Level One and then Level Up?

00:22:49: I think you can start anywhere, but it's just about being mindful of the higher level in terms towards organisational capability.

00:22:57: The

00:22:58: more we decide...

00:22:59: The more design and governance problem that needs around them.

00:23:03: I like this level approach, so then you can also say okay at which level is the use case?

00:23:09: or by myself.

00:23:11: So it's really great that we have some maturity

00:23:13: to understand the impact of your organization because clearly impacting one role was very different than impacting entire organisation.

00:23:20: but there may be some very key important roles in your organisation where if you transform an individual on their way they work.

00:23:33: Okay, when we think about leadership and people.

00:23:38: When you talk about AI do you see some differences in approaches?

00:23:42: In motivation?

00:23:43: In assessment?

00:23:47: or is this really working together in the organization to move AI forward all?

00:23:55: Do You See Some Difference?

00:23:58: I think what i see player in my organization is the same as what we might see play out In The Public and that's there are divides And different reactions, and differences of opinion.

00:24:09: As I've said those There Are Four Against Or Unsure About The Technology Because We See Those Debates Player Internally Trying To Line Our Internal Appetites As I Said A Strategic Opportunity to Make Jobs Easier to advance the fight for human rights with our external human rights perspectives, which care about the advances in technology and its impact on human rights or people.

00:24:37: And there are many other harms that we care deeply about.

00:24:40: so we need to harmonize them in what they do.

00:24:47: there are indigenous peoples, they're mission represented in biases.

00:24:50: There are those living close to new data centers and seeing firsthand the environmental impacts.

00:24:56: so I feel very privileged to have that perspective guide what i do And bring it into life make it real for me.

00:25:04: but theres also something around a digital divide For Me.

00:25:07: So Those That Have Connectivity Compute & Data Now Have Access To This Tech.

00:25:14: They Are Turbocharged.

00:25:17: Those that don't have connectivity, compute and data they risk becoming second-class citizens in getting behind.

00:25:24: But there's a thin layer in the middle.

00:25:26: So these are people That In some way shape or another sit in the Middle in terms of their access to resources.

00:25:34: but They now I've accessed this tool which gives them education health advice language support law advice medical advice at A level what they maybe didn't have or can afford before.

00:25:47: So it's impact on society in that way is fascinating for me and seeing how those three categories are sort of appearing, this really interesting.

00:25:57: Great

00:26:02: one last question a little bit.

00:26:02: Imagine a colleague from similar organization form another NGO calls you and said hey Paul I heard you did such amazing things with the eye at MST International, we also started or starting our journey and don't know where to start.

00:26:23: What are two of three concrete tips on takeaways that would give us?

00:26:32: It's very practical!

00:26:33: I'll give it a try i think.

00:26:37: so first build your governance foundations early.

00:26:41: establish your red lines The things which are absolutely unacceptable to you.

00:26:47: List the things that aren't acceptable in use of this technology, whether it's as an input too or a result of process and adding bed-responsible uses for example.

00:26:59: so lift existing policies to cover geometry of AI roles responsibilities approvals but put guardrails into place around what is not allowing your users No confidential or personal data in public AI tools.

00:27:15: Always have a human and the loop.

00:27:17: Have expectations around transparency, have expectations where model output monitoring And establish your principles The things that will guide you decision-making In the selectional use of tech.

00:27:29: Secondly I've mentioned it multiple times already.

00:27:31: But that capacity building That need to lift everyone In terms of awareness and understanding Space for dialogue Because it's important, you know.

00:27:43: It is just technology.

00:27:46: so things like prompt literacy and our own basics are important.

00:27:50: So people understand the capabilities or limits of tech because its still kind of trash in, trash out Its just a tool!

00:27:58: Its so reliant on effective and responsible utilisation.

00:28:03: So creates spaces for people to talk and ask questions, have a debate.

00:28:09: And we do that through always on channels or very focused working groups I think.

00:28:15: lastly as we've said several times just start small safe but high impact pilots.

00:28:22: don't try to build full AR roadmap instead design it an innovation pipeline follow the use case-centric roller to contain your risk relate that to your information and data flows, and prove the benefit with each one.

00:28:35: Because it's easier if you don't have small... And focus only on those use cases that align with your mission and value chain because there are a lot of things can be done.

00:28:46: not everything could be done by AI.

00:28:48: You still need other tech or to address people all to address process.

00:28:54: That would be

00:28:55: great.

00:28:56: Thank you so much.

00:28:57: So many great insights from your work and your AI journey at MSI International, I think uh... Many people can get something out of it for their own journey.

00:29:08: And thank you for this conversation Paul!

00:29:11: For the time you're very busy.

00:29:15: We appreciate that you took the time to talk about these conversations.

00:29:19: It's really important.

00:29:24: See you soon and have a nice

00:29:51: day.