Subscribe to our newsletter to receive business tips, learn about new funding programmes, join upcoming events, take e-learning courses, and more.
Take the first step to successfully starting and growing your business.
Join for freePosted: Mon 15th Jun 2026
As small businesses integrate AI tools and agents into products and processes, many are moving too quickly to fully understand the risks involved.
In this practical session, Olugbenga Latinwo explores how you can build AI products more responsibly in the early stages – without slowing your innovation.
Drawing insights from building Culture Illustro at Craftdash Limited, his session covers practical AI guardrails, iterative experimentation and scalable governance-by-design – all tailored for start-ups and SMEs.
How to integrate practical AI guardrails early without slowing innovation
How iterative experimentation can help reduce the risks that AI products present
Key considerations for safely deploying AI tools and AI agents into products and processes
Olugbenga is a business growth strategist, an AI product builder and the co-founder of Craftdash Limited, where he leads the development of Culture Illustro, a cultural AI illustration platform.
He has led digital innovation and transformation projects across the UK and Africa, including serving as a lead implementing consultant for the UNDP-EU Nigeria Jubilee Fellows Programme (NJFP) Talent Management Services initiative in Nigeria.
Olugbenga recently took part in The Alan Turing Institute's Practitioners Hub as an Expert-in-Residence, contributing practical perspectives on responsible AI and governance-by-design within start-ups.
Watch more expert webinars
Access a growing collection of expert-led webinars covering marketing, sales, finance, growth and more – ready whenever you are.
Lightly edited for clarity.
Ryan: Good afternoon, everyone, and welcome to today's Lunch and Learn. My name is Ryan, and I'll be your host today.
For those of you attending a Lunch and Learn for the first time, Enterprise Nation is a vibrant community platform for start-ups and small businesses.
Today, I'm really pleased to introduce Olugbenga Latinwo, who is a business strategist. In this session, Olugbenga will explore how you can build AI products more responsibly in the early stages, without slowing down innovation.
If you have any questions throughout, as always, just pop them in the chat and we'll do our best to answer them at the end.
The webinar is recorded, and a recording will be sent out later today as well, so keep an eye out for that in your inbox.
On that note, I'll hand over to you, Olugbenga.
Olugbenga: Good afternoon, and good evening to everyone, wherever you are on the planet.
My name is Olugbenga, and I'm the co-founder at Craftdash Limited, where we build cultural AI tools.
Over the next few minutes, I'll be sharing practical lessons I've learned as a founder and product lead while building AI products. These are not just theoretical lessons. They're practical lessons I've learned along the way.
We're not just looking at compliance checklists. We're also looking at responsible AI from the practitioner's point of view.
I want to start with a question, and I'm posing this to everyone here. In 100 years, what will AI remember about us, about our businesses, about our culture and about our communities?
We're going to refer back to this question by the end of the session. Most importantly, who gets to decide what AI remembers?
These questions became very real for us at Craftdash while we were building Culture Illustro, and they taught us a lot of lessons along the way.
We are living through an extraordinary moment. Every conversation is around AI. AI has become very accessible and is being used for many different things. Everyone is talking about it.
But what we often fail to discuss, especially as founders, product leads, product managers or people wanting to build AI tools, is trust. Trust has become a difficult conversation to have.
Very few founders I've interacted with over the past few years have taken an interest in deliberately designing for trust. Often, the focus is just on the AI tool and its output.
That is one of the biggest lessons I and my team have learned.
For some background, Culture Illustro is a cultural AI tool that builds trust into the images we see online and makes sure underrepresented and unique cultures around the world have a voice within the AI space.
When we started building Culture Illustro, we concentrated more on better prompts. We wanted the outputs to look good. We wanted the generated images to look clean and impressive. For a while, it seemed as if we were succeeding, until we realised there were a lot of things we were doing wrong.
The picture here shows some of our cultural datasets. What did we discover? We discovered that an image can be technically correct and still be culturally wrong.
One example on the screen is from one of our sample datasets. A prompt was given to generate a particular cultural dance from the Yoruba ethnic group in Nigeria.
As part of the evaluations and tests we carried out to validate the problems we wanted to solve, we discovered a lot that was wrong with the images we see online, how they are generated and how trust is often missing from that process.
Suddenly, the challenge was not just about the quality we were seeing. The challenge became: was trust built into this from the design of the product? Was representation built into it? Was accountability built into it?
That was when we stopped looking at Culture Illustro as just an AI image generation tool. It went beyond the glamour of clean images.
I'm going to cover five lessons today.
The first lesson we learned came from our participation at the Alan Turing Institute, which was a six-month engagement where we worked on a lot of case studies and discovered what systems thinking is all about.
Before that, we were concentrating more on the training datasets, the AI model and the outputs. We had forgotten about the people, the communities, the users and the trust we were supposed to build in.
Systems thinking was one of the most valuable lessons I got from the institute.
I want to encourage as many people as possible who are building AI products to look at where they can get help from. It might be platforms like this, or other institutes running AI accelerator programmes. That support saved us a lot of headaches going forward.
Systems thinking helped us zoom out. Outputs affect people. People influence trust. Trust ultimately determines adoption.
Our breakthrough wasn't improving the model. Our breakthrough was understanding the system around the model. When we looked at the system, we discovered a lot of surprising things.
That takes me to the next lesson: the stakeholders we nearly missed.
If you look at this diagram, we discovered hidden stakeholders when we started mapping stakeholders across our AI lifecycle.
We needed to go back to the drawing board and ask: what are we missing? Which stakeholders are we missing? Then we started mapping them across the Culture Illustro AI lifecycle.
We discovered that, within our design, we had not properly concentrated on represented communities: the communities representing these cultures. They were one of the major sources of truth we could get.
Systems thinking revealed all of this and helped us start working on how to build trust into the product.
When we talk about governance and accountability, we also have to talk about provenance, which means being able to trace who is responsible for what. When something goes wrong, we can track it down.
The next question for us was: how do we involve the right stakeholders at the right stage and at the right moment?
As a product builder, or as someone building an AI tool, you need to understand that it is not just about the AI lifecycle. We needed to know what happens where, from cultural briefing through to deployment. Which stakeholders are present at each point? Who reviews? Who decides what?
Trust is easier to build early than it is to repair later. This is one thing I always tell founders I speak to: don't let trust be an afterthought. Build it into the design.
The same goes for inclusivity. We needed a way to experiment with what we had learned and build forward.
Then came lesson four. One of the most useful concepts we learned was the Amazon two-way door principle.
To be honest, we didn't realise that was what we were doing. We started working on things in smaller bits and making sure we failed while testing, instead of going through the one-way door where the damage has already been done.
For example, publishing culturally inaccurate content and putting it online is a one-way door. By then, the damage has already been done.
We are all familiar with the EU AI Act, which is coming into implementation in August, and we know the impact and importance of this for every founder and everyone building AI, whether you are building or integrating it.
The two-way door process has helped us make reversible decisions.
For instance, if we are working with a particular dataset and find out it is not culturally accurate enough, we go back to the drawing board and look at our confidence level. Is this plausible? Is it accurately in sync with the culture it represents?
It is much harder to go through the one-way door because, by then, the damage will have been done. Trust and governance need to be embedded into your product pipeline and at the heart of your design.
People often think guardrails slow down innovation. I disagree. I've had a lot of arguments about this, and one thing I can say from experience is that I used to think AI governance, trust, accountability and traceability were only for bigger companies or enterprises.
Later, I realised that even if you are adopting AI into your workflow as a business owner, or integrating different tools, you need to understand that guardrails are the lifeline of what you are doing.
They allow us to move quickly without losing direction.
In Culture Illustro, for example, we have cultural briefings, human reviews and feedback loops from communities and cultural experts. At every point, we know where we get feedback from.
Even when we release an output, we still get reviews and feedback, so we can improve what we have already experimented with.
As a founder, without guardrails, AI scales mistakes. Think about the big images we see online today that people complain about.
We cannot always trace the source. They may look beautiful, but many are culturally inaccurate. Not all of them, but many.
Often, we cannot trace the sources of these images. That is why we say, "It's an AI-generated image." Many images are trained on synthetic data, so we need to ask: where is this image coming from?
Good guardrails do not reduce innovation. This has become very important for us as we take action forward.
As we move into new ground, the world is changing very quickly, and many AI tools have helped us in one way or another.
We have also entered the agentic era, where it is not only about outputs. It is now about what acts on those outputs, and that is where AI agents come in.
AI can generate a good answer, but we should start asking: should AI be allowed to take action?
For example, would you allow your AI tool to execute a bank transaction? Would you allow an AI agent to answer customer queries? Are you confident that the AI tool will do exactly what it is meant to do?
That is why we talk about things like human in the loop. But AI agents can execute without humans in the loop, and that is why we need to talk more about trust and responsible AI. That is why governance matters even more in the agentic age.
This is where many people still misunderstand what we mean by safe AI.
I work within the safe AI space as well. So what does safe AI actually mean? It is not about slowing down innovation or development. Safe AI means trusted output.
In our case, we ask: are the images you see online trusted? It goes deeper than saying, "We are just using the image for this or that."
AI scales unverified images. When AI drifts, the output drifts as well. This content goes into classrooms. Students use it. We can easily scale what is not true over the next few years until it becomes accepted as truth in society.
That is the age and time we live in if nothing is done to build pipelines of trust, accountability, traceability and provenance.
So how do you, as a founder or business owner, remember all of these things?
I've tried to come up with what you can see on your screen: TRUST. That stands for task, risk, users, supervision and training. I use this with my team to make sure we are on track with what we are doing.
First, what problem are we solving?
Second, what risk is involved if we solve that problem? In the case of Culture Illustro, we are helping underrepresented cultures be accurately represented on digital platforms, and closing the gap between the use of synthetic data for creative purposes and cultural accuracy.
There are also questions around what could go wrong and who is affected. Who are the users?
If you refer back to the AI lifecycle and stakeholder mapping, that is why you need to understand who your stakeholders are at every point. When something goes wrong, can you track the users affected? Can you also track the stakeholders affected?
It also asks how the system will improve. That is where evaluation comes in, and where feedback loops are built into products. These feedback loops should usually involve the stakeholders who are relevant at that point.
With what we are building, there is not always a simple right or wrong dataset. We use confidence levels to determine how plausible, authentic and trustworthy an image is.
These five questions have helped us catch many issues before they unfold.
If I were going to be a start-up founder tomorrow, or an SME owner trying to adopt AI, there are a few things I would do.
· First, map stakeholders.
· Second, identify high-risk workflows.
· Third, run small pilots, which I've spoken about. Fail in smaller steps, using the two-way door approach, and be able to revert quickly if something is not working without burning too many resources. As a founder, you have a limited runway.
· Fourth, build feedback loops. I've mentioned this already.
· Fifth, implement guardrails. Guardrails are not there to limit innovation. They are there to help you go faster. Once your guardrails are integrated properly, you have a framework to work within and you can innovate more quickly.
Learn and improve. It also helps us scale what works. As I mentioned, once you have a distorted image, or an image that is not culturally correct, AI can scale that same wrong image over time until it becomes accepted in public use, content and education.
That leads me back to the first question I asked. In 100 years, AI will remember what we choose to teach it today, especially through the datasets coming in.
So let's design AI by intention, not by accident. Build trust before you scale.
Build it through your datasets, workflows, products and decisions. The future of AI is not being decided in 100 years. It is being decided now, by whether we build in accountability, trust, provenance and traceability, and whether we work responsibly on our products.
The focus should not only be on the model, but also on the systems built around the product.
I'll stop here. I've tried to do justice to the time I've been given. Ryan, I hope I haven't overrun.
Ryan: You were bang on, and that was really interesting and helpful, Olugbenga. Thank you for that. I really enjoyed it.
We have a couple of questions. Someone is asking: what is the EU AI Act, and what changes to regulations are coming in August?
Olugbenga: I would say the EU AI Act is a framework, and the part I particularly like is the risk categories. It categorises AI risk at different levels. We have low risk and high risk.
For example, some uses of IP cameras may not be permitted for identifying people. Part of why these things are put in place is to provide a framework, just as you have GDPR.
So you have the EU AI Act, and other parts of the world are also bringing in frameworks. The EU AI Act is taking a giant step forward in making sure we build AI more responsibly.
If you look at GDPR, it addresses data privacy and many other things. The EU AI Act is a large white paper, and I encourage every founder to go through it and study it properly.
The most important thing is to know where your risk lies. If you understand your risk, you will be able to build better and understand the consequences of what you are doing.
I'll give some context. I recently spoke with a founder here in London who had built a fantastic agentic AI platform to help make customer queries faster and support project teams.
The first question I asked was: what market is going to absorb this product? That question could not be answered.
Was it built for the UK market or the EU market? Are you aware of the EU AI Act coming up? If yes, how have you built governance and that framework into your product?
If that has not been done, the product may be dead on arrival. It means you need to go back to the drawing board and start again.
That is why I say guardrails are not there to slow you down. Guardrails are there to speed you up and put the right things in place from the beginning.
Ryan: It's so important to have those guardrails right from the beginning. That's really helpful. Thank you, Olugbenga.
Another question: a lot of SMEs are feeling pressure at the moment to adopt AI quickly. What would you say are the biggest mistakes you see companies making when they rush to adopt AI in their business?
Olugbenga: One of the biggest mistakes, and I've also taken this from some MIT research reports, is that founders, business owners and SMEs need to understand their workflow.
If you don't understand your workflow, you won't understand what you need.
When you're building a product or adopting AI, you need to understand the problem first. Define the problem you want to solve. Understand it, then look at the risk involved in adopting AI.
Before you even get there, define your problem and identify the gaps. It is like carrying out a gap analysis. This is where we are. We need X, Y and Z to cut down cost or enhance productivity.
Then ask: what are the needs of the people in that department? What are their skill sets?
You really need to drill down. Your feedback loops can start there. Your internal interviews can start there. Interview your employees. What do we need? What is their skill level? Then you can determine what works for you.
What I see many SMEs doing is founders adopting AI tools and saying, "I'm going for ChatGPT" or "I'm going for Copilot", but different generative AI models and large language models have different strengths.
You also have different needs. You need to match your needs with the solutions you want to adopt.
So I would say: understand your workflow. How does your company operate? What are your operations like? What do you do? How do you do it?
Ryan: Brilliant. Thank you, Olugbenga. That was really helpful and interesting. We've had people saying thank you in the chat and that it was really insightful.
We are just on time, so a big thank you to everyone for joining, and a big thank you to Olugbenga. That was a great session.
I've popped your LinkedIn and Enterprise Nation profile in the chat, so I'd encourage everyone on the call to connect afterwards if you have further questions or want to learn more.
Thank you, Olugbenga. That was a really great session.
Olugbenga: Thank you, Ryan. Thank you everyone for listening. It's been a pleasure.
You can connect with me on LinkedIn, and I also encourage you to use Enterprise Nation as a platform to connect as well. Please do sign up.
Thank you very much, and have a wonderful day.
Ryan: Thank you, Olugbenga. Thank you, everyone. See you all later.
Unlock more on-demand sessions designed to help you sharpen your skills, grow your business and stay one step ahead. Find more Lunch and Learn webinars
