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: Wed 2nd Sep 2026
What if AI could help you create more time for customers, strategy and growth – without losing the human touch?
In this practical session, Dave Gibson shares the real story of a company that used AI to rethink how it worked.
Rather than simply cutting costs, the goal was to free up time and reinvest it in the areas that matter most: customers, strategy and new business.
See what they changed, what worked, and how they made sure AI supported people rather than replaced the human judgement their customers value.
Find the time AI can give back
Know where to start with AI
Keep people at the heart of your customer experience
Dave advises founders and boards on putting AI to work inside small, medium and large companies, and invests in AI ventures.
He's the author of Use AI. It Is Awesome and The Amplifier and originator of the Human to AI to Human framework.
Watch more expert webinars
Access a growing collection of expert-led webinars covering marketing, sales, finance, growth and more – .
Lightly edited for clarity.
Monica: Hello, everyone, and welcome to today's Lunch and Learn. My name is Monica, 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 startups and small businesses. I am pleased to introduce Dave Gibson, who is a founder and mentor. In this session, Dave will share the real story of a company that used AI to rethink how it worked.
If you have any questions throughout the webinar, please post them in the chat, and we'll do our best to answer them at the end of the session. Today's webinar will be recorded, and we will send a follow-up email to you with the recording and further resources later today. Over to you, Dave.
Dave Gibson: Thank you. Good afternoon, everyone, and welcome back after a long weekend. I'm speaking to you from Google headquarters in London, where I'm attending a conference about building agentic AIs into companies, so this is a timely talk.
I'm going to put this in full screen. We're only restricted by my battery, because I'm on a NUC at Google with no plug socket – probably why it was the last one available.
This topic feels particularly relevant, because there are a lot of statistics out there from McKinsey and BCG about companies adopting AI but falling short, for one reason or another. Right now, I think we're in a huge exponential growth phase in the age of AI, and that means only one direction. We haven't seen anything like it since the industrial age, and it's ridiculously exciting.
A little bit about me: I started my UK experience at the London School of Economics, where I chose a professor who was really forward-looking. We published a lot in ubiquitous computing – looking at bridging humans and technology to produce better workplace efficiencies – starting in 2002, and continued publishing for some time after.
I then took over as head of end-user adoption at Cisco Technologies for Europe, the Middle East, Africa and Russia, worked at Adidas in transformation at their global headquarters in Germany, and trained at Accenture.
Now, what I'm really doing is helping companies and individuals adopt AI and get the benefits out of it, and the benefits are real. I've been working with it for a couple of years now, and it's been unbelievable – I feel empowered and amplified by it.
I've worked with a number of companies in this space, and CEOs and founders keep coming to me asking one thing: how do we move from human-to-human to agentic? One of the first quotes I showed a CEO I was working with was from Jensen Huang, who said you're not going to lose your job to AI, but you're going to lose your job to someone who uses AI. I love that quote.
The CEO I was working with said our job in 2026 is to be the company using AI, not the one being left behind. After eight weeks into the project, the diagnosis was that growing the old way – more clients, more hires, more hours – was the cause, not the cure.
So what was happening with the customer? I've replaced all the figures with euros to anonymise the customer, but the shape of the story is real.
Revenue was compounding at 77% a year; we were growing, and we were getting more clients, but net margin was stuck at around 8% per employee, which is not great. 54% of revenue was going to salaries, and revenue per person was about 80,000 per head.
Growth actually made things worse: the more clients we got, the more people we had to hire, and the more people we needed in front of clients. The important part was that we always provided a white-glove, customer-centric service, and we didn't want to reduce our net promoter score.
So the more clients we got, the more people we hired, and the more founder, CEO and senior executive time it took to onboard those new people and get them up to speed for clients.
Monthly revenue was swinging, so we set three objectives, which is how we start every single project: what do we want to achieve, at the highest level?
First, serve customers better and keep NPS high. Second, lift margin from 8% to 20-25%. Third, don't stay 100% human – incorporate agentic AI wherever we can.
The new operating model has three layers. Layer one is humans – never replace them; they're always going to be there. Two or three years ago there was a lot of talk about AI replacing humans, but over the last year, more and more of the people I speak to are actually hiring, because AI produces so much good content and information that humans get more time to focus one-to-one with clients.
There's so much good knowledge coming out of AI, prompted by humans, that we need more people to take those insights and actually execute on them – there are far more executable items coming out of AI right now. Humans have the judgment, the relationships and the accountability.
Layer two is AI agents – where can we apply agents to increase efficiency and reduce time spent by humans? Layer three is flexible capacity: repeatable execution in a lower-cost hub, plus a pre-vetted freelance pool, so we're not always hiring full-time as we sell more projects and grow.
There were 11 workflows we changed – I'll go through them quickly. We took process-heavy pieces of work and cut them from hours into minutes: competitive research on new clients went from 45 minutes down to five, moving from human to AI and back to human to make sure the research was solid.
Brief generation went from three to four hours down to under one hour, and copy variants went from two to three hours down to about 30 minutes. On quality and volume, internal quality checks went from 15 minutes down to two per piece, client feedback synthesis went from 30 minutes down to 10, and video and design production roughly halved. Right away, we were cutting time and increasing quality that we couldn't get without the agentic piece.
On the client rhythm, monthly reporting went from two hours down to about 25 minutes, and all of these add up to save a lot of time at the end of the month. Onboarding went from 30 senior hours down to 12, creator sourcing from five weeks down to seven days, and handovers – previously three months to get up to speed – became a 15-minute brief.
The founder wins back 13 to 21 hours every single week to spend on things that really add value.
I find AI really useful because it allows us to spend more time with humans, and the more time the founder could spend with people, the more client work they could win – they're the best business developer in the business, because they know it inside out and know how to win deals, or they wouldn't have founded the company in the first place. That was one of the key things: giving the founder back hours so they could spend more time with people.
In total, that added up to 400 to 500 hours recovered every single month – the equivalent of three full-time people – which we could set against the cost of new AI tools.
We did an interesting exercise analysing all the tools they were using, then worked out exactly what they needed in terms of AI. We actually cut the cost of their existing tools, and the cost of the new AI tools – plus the time saved – ended up cheaper than the legacy tools they'd been using.
On accountability – how humans and AI actually work together on a brief: a human designs the hypothesis and dictates it, the agent transcribes and assembles the brief, then a human reviews and edits it. A senior person checks brand voice, and a human delivers it to the client.
That's how we kept a human-to-human contact point throughout – always a human, with AI in support, but always a human.
The rule that keeps trust intact is that the agent never emails the client – the human is always accountable for what ships. If something fails, the named person owns it; it's never the AI's fault.
Customers know exactly where AI is used, and who stands behind the work.
Transparency was another key thing. On our first pilot with a customer, we said outright that we were going to try to make things more efficient by using AI, and they were actually very eager to learn how AI would take a hand in delivery.
The CEO and management team at this company were quite concerned about what would happen if the customer found out they were using AI, but the key turned out to be transparency.
Don't forget you can put rules into your AI – no em dashes, sound human, use our brand voice, mirror how the CEO sounds in their communications. You can upload a lot of the CEO's actual emails to mimic how they'd speak to clients, to get as close as possible to that authentic brand voice, knowing there's always a person on the other end.
The customer wins too – we tested three to five times more ideas. If you've ever worked with an external consultancy, agency or service provider, you'll know it's nice to get more ideas than fewer, because you start with a wider funnel.
Reporting got richer, not thinner, and starts got faster – days, not weeks – which meant getting to revenue quicker, plus freed-up senior hours reinvested in customers. On the honest numbers: the margin path went from 8% and accelerated toward 20 to 25%, and at 70% execution we're still seeing roughly 15%, largely self-funded from software spend already in the budget.
Targets are always in motion, not banked results, and honesty travels, as we discussed earlier.
So, my biggest recommendation is three moves to start. First, map your hours: where do senior people spend repeatable hours every week? Those common, recurring processes are your starting map.
Second, pick your line – decide what stays human (judgment, relationships, the final say) and what gets a first pass from AI. Third, run the win-win test.
Your customer also has to gain – the CEO's rule was: don't do anything that doesn't benefit the customer, and we stuck to that throughout. Efficiency without service is just churn, and that's something to watch out for with AI.
You can spend so much time in front of AI – it's like an endless knowledge bank of ideas – so you really have to be clear about the difference between efficiency and churn.
So how long did that take? Fifteen minutes – we rocked and rolled. I'll leave it there, and we can take questions.
Monica: Perfect. Thanks, Dave.
Dave: You're welcome.
Monica: We do have a few questions. The first one is: how do the benefits you've talked about translate to a sole trader?
Dave: I bounce back and forth between entrepreneurialism and intrapreneurialism, entrepreneurialism as a sole trader and then scale that, and intrapreneurialism in big companies. As an entrepreneur, I've never felt more liberated to do whatever I want.
If you go through ideation for a new business, or an existing business, as a sole trader, you'll have patents or copyrights that need to go through lawyers, NDAs that need to go through lawyers, and company setup that potentially needs a lawyer too. You'll also have accounting and government registrations, plus marketing and sales ideas to work through.
I used to spend time finding a good supplier and working with them to validate information, whether that was legal, marketing or HR – you name it. Now I can do 95-98% of that myself, as long as I understand how to prompt it properly.
That's a danger with less experienced people using AI too, because sometimes you don't know what to ask. As a sole trader, though, it's never been as affordable to run a business with AI as it is now. Does that help?
Monica: Michelle is asking: I spend a lot of time doing Instagram posts that lead to client consultations. How can AI help improve social media, and what are the costs?
Dave: Great question. If you take a marketing process – from ideation, to production, to launch, to measure, and then improve the cycle again – I think we're seeing AI bridge those steps.
There are multiple different AI tools that help you ideate. For the numerical or business side, I like to use Perplexity – it's strong on legal and financial, and I find it better than Claude for that kind of groundwork. That's your foundational piece, and then I work it against Claude, which is better for creatives.
I'm actually sitting at Google headquarters watching their Gemini presentations right now, and they're pretty confident about their suite of AI tools. You can also use Nano Banana, from Google, to do creatives – that's just fresh in my head right now – and then test those creatives.
For building a messaging tree, I tend to stick with Claude, although I don't think it's the strongest tool for creatives – there are better options out there for that.
For voice or video, there's ElevenLabs, which is fantastic – it can build any voice. There's also HeyGen, which can build you an avatar: you take a short video of yourself, and it builds an avatar that looks exactly like you, so you can type in scripts and it creates whatever video you want.
So if you want to test 10 different scripts or messages without getting in front of a camera 10 times, you can use a platform like HeyGen to test it.
The tricky part I'm finding right now is the CRM piece – connecting all your different channels: YouTube, X, LinkedIn, Facebook, and so on. I've been trying to find a good, holistic, end-to-end CRM tool in the AI space, and I can't find one, though I know people are building it right now.
That's probably one of the holy grails at the moment: being able to do market research, build your messaging tree, build your content, launch it across different channels, and then test it to find out what works best on which channel. I don't think there's one tool that does all of that really well, though there are central management tools like Make and n8n that act as hubs to bring everything together so you can measure it all in one place.
But I don't think there's a true end-to-end CRM holy grail yet.
Monica: Damon says that Monday.com is trying, but not quite there.
Dave: I agree – Monday.com is trying its best too, and it's probably the warmest one so far.
Monica: Annie is asking: is it best to sign up and pay for an AI agent and safer?
Dave: It's funny – I've tried pretty much all of the AI agents out there, and I'm happy to run through my experience with them.
ChatGPT was mine at the beginning, when I was just learning, but it ran out of steam when I needed to go deeper. Gemini has a good general capability model, but it's not exceptionally deep on anything – good at everything, but not very deep, and I'm happy to say that to the Google team right now.
I think Claude is quite good, but it feels like it limits me – the more I get better at using it, the more restrictive it feels.
Perplexity, as I said, is really good at legal and finance – exceptional, actually. Grok is good at graphics, although it's pretty loose and not very specific.
If you ask what my AI bills are every month – I think I'm paying 100 for Claude and 60 for Perplexity. I just tried Kimi, the new Chinese open-source model, and I really like it – it's really good.
I think that's why all of Silicon Valley is watching Kimi right now – it's the big Chinese model that's just come out.
DeepMind is actually pretty good too, but I have a whole list of licences I'm paying for right now, because we're still in the early stages of AI. It really depends where you are on your own journey – whether you're building workflows, working as a sole trader, or building a specific product or service.
It also helps to use tools against each other – I always use Perplexity against Claude; that's my favourite duo. So, the short answer is: it depends where you are and what you're working on – trial it, try it, try it.
Monica: There are quite a few questions along similar lines: if you're starting out, which tools should you use, especially as a sole trader, or in creative businesses or publishing?
Dave: I'll give this piece of advice. I was out this weekend with some friends who are also in the AI space, and I think you can ask anybody working in AI right now – we've never worked so much, because there's so much to learn.
My own learning curve reminds me of the first two or three weeks of every management consulting engagement I had at Accenture, but this is probably a higher learning curve than any of us have ever had.
But I think we'd all agree that the most underused capability in any AI agent is the settings function. When you go into a new AI tool and open its settings, it gives you the chance to fill in free text – I wish I could show you on screen.
Don't forget, these are called large language models, so speak to it like a human being. There are two things to focus on here.
First, in the settings, you can define exactly what you want the AI for – for example: I'm a sole trader launching a product or service, I want deep analysis on the market, and I want revenue-generating responses. You can tell it to act as a chief marketing officer, chief financial officer or chief legal officer at all times – you're telling it how to behave.
That's really important, so please don't just open up an AI model and start typing away without setting that up first.
The second thing, which my friends and I underused in the early days, is the projects tab: create a new project – say, a project for Enterprise Nation – and give it project context. For example: I want to get more engaged with Enterprise Nation, add value to the community, and here are my best topics for it.
You can treat it as a three- to six-month trial project to see how it goes.
So there are two pieces: settings, where you set out overall how you want the AI to operate – and, again, no em dashes; it drives me crazy when they give me em dashes – and then projects, where you get specific.
That would probably be my biggest piece of advice for how to start interacting with your AI agent, whatever that may be.
Monica: Thank you so much, Dave, for the excellent presentation.
Dave: Pleasure.
Monica: Thank you to everyone who joined us today. We are recording, so you'll see the recording on the Enterprise Nation website, and you'll receive a follow-up email with the link to the recording, further resources, and details of how to contact Dave.
Thank you very much, and enjoy the rest of the day.
Dave: Thank you. See you, everyone. Bye.
Monica: Bye.
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
