Aside from being a most excellent adventure, the Marke family’s summer holiday to Thailand unexpectedly changed how I use tech. A bit late to the party maybe, but our Siamese foray has converted me into an enthusiastic Gen AI adopter – and, come September, I suspect I won’t be alone.
I’m of an age where, alas, I’m a bit stuck in my ways. It takes a lot to change something I’ve been doing for years. I like tech, and love a gadget, but also love a good old fashioned paper book, and my mobile is left in a corner at home and acts more like a landline. Probably for this reason, my early adoption of Microsoft Copilot at work, while equipping me with some useful new tools, didn’t change the way I work in any meaningful way (If you followed my Diary of an Augmented Exec series last year, you’ll know I’d already been putting these tools through their paces, but it still hadn’t become second nature.) And I’ve been stubbornly ignoring LLM (Large Language Model, a type of AI trained to understand and generate human-like text) apps. However, Thailand changed that.
Summer holidays are when many of us search the internet most voraciously – to book flights and accommodation, navigate, translate, find the best restaurants or the top ten unmissable things to do for bored teenagers. And you will have had to have been severely jet-lagged to not notice that this year’s browsing is very different from last year’s, largely because Google and Microsoft have embedded AI overviews into search. Search for things to do wherever you happen to find yourself and, rather than the usual list of paid and organic links in your search engine results page, an AI overview may – like a beautifully summarised Lonely Planet – provide a comprehensive summary of everything you are likely to need to know.
The Rise of Zero-Click Search
Apparently, in Google’s case ~15% of all searches now yield an AI overview. This rises to more than 30% in Bing. Certain types of search are most likely to trigger Google AI Overviews or Bing Copilot summaries – ‘Long-tail queries’ ( ‘Tell me things to do in Thailand’), ‘Informational queries’ (‘What fruits can I buy in Thailand?’) and ‘Fact-based queries’ (What is the Population of Bangkok?) – because they can be answered directly from authoritative sources, don’t require complex transactions (like buying something) and they can be summarised in a few lines or steps.
The thing about AI overviews is in many cases they avoid you having to move on and click a link because the AI figures out why you are searching and presents the answer to this in its response. And due to significant advancements in LLM technology, these overviews are becoming extraordinarily, uncannily good. Better, you might say, than something a human expert might return for you.
As of July, an extraordinary 75% of searches on mobile devices are now zero-click (it’s much higher on mobile than desktop btw because on mobile we are terribly impatient). Meaning the AI has given the user what she is looking for. And this trend has sent marketing agencies into overdrive, trying to figure out how to adjust approaches to paid search and SEO to influence the LLMs, given that punters are becoming less and less likely to click through and land on your website.
But here’s the thing. Because these AI summaries are so good in my browser, I now want AI summaries for all my searches. So, I’ve now downloaded ChatGPT, Perplexity, Claude, Grok and Gemini apps onto my mobile and am using these, instead of Safari and Google. Because of this search threat, Google’s core search bar has very recently incorporated an option to go into ’AI mode’ to ensure an AI query. Bing does the same with its Copilot tab.
AI in Everyday Life
These tools are so powerful, so useful, so productive for day-to-day life. On returning from Thailand, I’ve been arranging a new mortgage. I tried three routes. I got onto the broker who organised my original mortgage and asked them to find me the best deal. I searched Google to shop around for myself – and it returned the usual array of comparison engines. 20 minutes later having entered all my personal information into Comparethemarket.com, I was super frustrated to then not be given any answers but rather referred to a broker to book a call.
I then went on the latest ChatGPT app, asked it to ‘find me the best mortgage rates for a fixed rate 2, 3 and 5 year term, for a loan of £x with an LTV of y%’ and, voilà, in around 10 seconds it gave me all the answers I was looking for, a comparison of all the best rates in a lovely summary table and a whole host of helpful follow up likely next steps for me including links to apply for offers, a list of brokers, things you may have forgotten about the process, advice on insurance – everything. I then used this information to go back to my broker, challenge their offer, which they then improved, and I went ahead with them. The business models of comparison sites and brokers will be seriously challenged based on my experience.
It’s hard to believe, but the version of ChatGPT that broke all records when it went viral (1million downloads in a week) was only launched end Nov 22 – the app was crude, and text only but still appeared to us like magic. I first got my 365 Copilot beta shortly after this in Spring 2023. Both experiences appeared magical at the time, highlighted huge potential but were flawed and of limited practical value.
Since then, those tens of billions of $$ pumped into the industry in both research and infrastructure, driven by intense competition between the major players, has driven the unbelievable improvements I was seeing surfaced through my Thai searches. Better architectures have reduced the price of running LLMs by 9/10ths and collapsed release cycles from 18 down to 3 months driving the pace of change. Training data used to coach LLMs has moved from soupy large internet models to much more domain-specific, high quality data sources, meaning results are more accurate and trustworthy. Huge (and very controversial) factories of humans, usually based in developing economies, run tens of thousands of human feedback interactions every day to align LLM outputs more closely to human expectation hugely improving our experience. LLMs have gone multimodal (multimodal AI is like a super assistant that can read your email, interpret a photo, and book your flight—all in one go), making them astonishingly human-like in their capabilities. And this innovation loop is getting faster and faster.
In the next 6 months, we won’t just see LLMs gaining new tricks. We’ll see existing capabilities sharpen, scale, and become genuinely usable at speed. Multimodal is already here, but it’s about to become the default. Persistent memory (where AI remembers past interactions to provide more personalised responses) exists today, but soon it will be smarter, more seamless, and more personal. These models are getting better at acting like humans. They can plan multi-step tasks, run those tasks with tools like code, APIs, and search, and increasingly know when they’ve got something wrong. They won’t just apologise when they mess up (as they do now). They’ll proactively self-correct and improve their output in real time. Going back to my mortgage application, how long before my LLM is able to manage the whole workflow of the mortgage application for me through to completion? We’re getting to the point where you’ll genuinely have to stop and ask: is this a person, or a model pretending to be one? That’s not years away. We’re talking months.
With the launch of GPT-5, we’re now seeing early signs of ‘software on demand,’ where you describe what you want in natural language and the AI builds it. “This idea of software on demand will be a defining part of the GPT-5 era,” said OpenAI founder Sam Altman. This so called ‘Vibe coding’ will accelerate the production of code, enabling non coders to code and professional coders to speed up.
And business leaders know it, too. According to a July 2025 Gartner survey, 62% of CEOs and senior executives identified AI as the defining force in the future of competition over the next decade.
What is the relevance of this very visible, consumer-led change to resellers and MSPs? If you believe the LLM hype, then you must also believe the near-term threat and opportunity for business of LLMs replacing humans. I did a GPT search to see what it reckoned were the jobs most exposed to replacement by AI in the short term. It characterised these roles as any involving repetitive, text-heavy or predictable knowledge work – content creation (copywriting, content marketing), customer service (call centre, live chat and helpdesk staff), routine legal work (para / junior legal staff), market research (analysts and researchers), basic journalism, junior level coding, translation and tutoring.
It also pointed to likely near-term issues where the structures of businesses as they work today collapse due to entry level work and workers disappearing, with the obvious knock on to mid-level staff and execs. All very real changes that will impact en masse by 2030.
These technology-driven changes, quite simply, mean that for resellers and MSPs to remain relevant to their business customers, in the same way that they cannot avoid getting skilled in security, cannot avoid figuring out their place in the AI eco-system.
Tyler’s Story: Innovation from the Beach
Let me take you back to Thailand. The last week of our holiday we spent on an idyllic, palm fringed white sandy beach on Koh Phangan. We stayed in beach bungalows owned by a lovely guy called Tyler, who left his LA home and UX designer / coder job 20 years ago to marry his Thai sweetheart and set up his resort. Tyler is an unusual small business owner in that he is also something of a tech expert.
We chatted at length about how he was innovating guest service using Agentic AI (AI that can act independently to complete tasks without human prompting) – and he proudly showed me the beta of his new agentic hotel concierge. Under his own steam, he had used Grok, Llama, Gemini and ChatGPT to build (code from scratch) an agentic solution – getting one LLM to build v1.0 then passing it to the next LLM to build v2.0 etc – and then took data from his 20 years in business – from his booking system, WhatsApp for Business, APIs into booking.com and AirBNB – structured and secured the data, and pointed it at the agentic solution – voilà, a button on his guest mobile app where you could chat or text questions in any language and get an instant response – and it was super impressive.
He runs a business entirely staffed by his family, so wasn’t looking to replace anyone with a robot – rather his motivation was being tech curious and seeking to improve guest experience. We then talked at length about his ideas to package up his solution and sell it to fellow guest house owners across the island, and he quickly identified that his problem would be who would do all that back-end work to pull the datasets together from each hotel customer and configure and enable the agent solution in each customer installation. Sounded to me like a perfect job for a Koh Phangan MSP.
Where MSPs Can Win – More Efficient, More Secure, Faster Growth
So, we come to where I started this article. Bright, tech savvy business owners are already implementing solutions into their businesses that can transform and differentiate many aspects of SMBs in every vertical market. For every Tyler in a guest house, there will be a Rachael in professional services, a Bob in construction, a Gillian in care homes, a John in retail, solving these problems with citizen coding.
But to scale they need MSPs. MSPs on the lookout for businesses they can partner with to propagate pre-existing IPR, or inventing their own when they work with customers, to then scale out and repeat. These solutions could be based on specific application expertise in leveraging AI capability within SaaS apps, or more integration style like my Tyler example – but all will involve data and security – because these are the essential underpinnings for any serious, quality LLM enabled business solution.
The opportunities are there now. The LLMs are capable now. The APIs are there now. The SaaS apps are available now. The security frameworks are well known. It needs MSPs prepared to be bold, prepared to invest in innovating with willing test customers. And there is margin and huge differentiation for first movers – wait 2 years and everything will have changed.
And there are significant opportunities for MSPs to leverage AI to drive efficiency within their own delivery models, automating the lead-to-cash-to-care processes, unlocking efficiencies and to provide more intelligent, effective cyber security solutions to protect themselves and their customers. Solutions like AI-powered ticket triage, predictive maintenance and security alerts, and smart onboarding workflows are all very achievable.
At Giacom, we’re not just encouraging partners to explore these opportunities, we’re on the journey ourselves. Across the business, we’re identifying high-impact use cases, putting the right guardrails in place, and building the skills to make sure adoption sticks. For developing our platform, this means using AI to speed up coding, pulling insights from large datasets, and taking repetitive work off our senior engineers – giving them space to focus on prototyping and pushing the pace of innovation. It’s the same co-innovation mindset we’re calling for in the channel, just applied to our own dev, engineering, and product teams.
Over the autumn, I want to bring to life the opportunity for Giacom partners to make money out of Gen AI, selling into SMB and by becoming more efficient and more secure. I will interview leaders from our strategic vendors to get their views on their best solution ideas, I’ll interview MSP leaders to uncover SMB wins and strategies, automation programmes and cyber wins, and I’ll interview SMB owners, like Tyler – who are actively innovating the opportunity themselves.
My aim – to shine a light on where our channel needs to move to stay one step ahead in this most exciting revolution – one that is moving so fast that I can hardly stop for a breath. But one that if we nail, will keep us relevant and drive growth for us all and productivity for our SMB customers.
If you’re an MSP ready to explore AI opportunities, there is low hanging fruit inside your business and in your customer base. Get on the automation, agentic journey inside your delivery model and identify customers willing to co-innovate your go to market proposition. The tools are ready. The market is waiting. Let’s lead the change.
This article was produced in association with Giacom and is classified as partner content. What is partner content? See more here.
Nathan Marke
Nathan is Chief Strategy Officer of Giacom
