How I Became an Apocaloptimist About AI (and What Your Agency Should Do Next)

The QR Code and Audience Matchmaking Initiative

Before getting into the substance of the talk, Mike draws attention to a QR code that appears on screen multiple times throughout. Scanning it leads to a survey where attendees can enter their mailing address and opt into two things: A physical gift sent by Mike personally. He is known for sending unexpected, creative physical objects to people he meets professionally (more on this later). He specifies it will not be anything that attracts customs charges or duty.

A professional matchmaking introduction, using open-source algorithmic matching code originally developed at Princeton for romantic matching on Match Day. The same code has since been adapted for matching founders with co-founders and other professional relationships. Mike uses it here to connect attendees in mentor or mentee pairings based on their survey responses.

He encourages anyone interested to include a full mailing address — not just a city or country — and to opt into the matching feature if they want to be introduced to someone at the event.

The "Not Married to the CMS" Philosophy

Mike attributes one of his most frequently used phrases to Jessica Dunbar, formerly an evangelist for Concrete5, a PHP-based CMS. The phrase is: "I'm not married to my CMS, I'm in an open source relationship."

He broadens this beyond CMS choice to encompass any tool decision an agency makes — including whether and how to use AI. This connects directly to a philosophy he and Andrew explored on their podcast Tools Are Tools, which also gave rise to Mike's second well-known saying: "Tools are tools. Just don't be one."

The point is that the open source and agency community has long suffered from tribalism around tooling. Debates about which page builder is best, which backup plugin to use, which host to choose, which CMS is superior — these debates consume enormous energy that could be spent on doing good work. Mike's position is that none of these choices are inherently superior. What matters is whether the tool helps the agency serve its clients effectively.

He gives several examples to anchor this:

• You can be a highly successful agency using only Shopify.
• You can be a highly successful agency using AI as your primary toolset.
• You can be a highly successful agency building everything by hand without any CMS.

He draws an analogy to Nathan Ingram, who reportedly once gave a talk where he said "I hate WordPress, it's going to ruin my business" — back when WordPress was first becoming dominant. Mike notes you could substitute "AI" for "WordPress" and get a response that feels entirely current.

- You can be a highly successful agency building everything by hand without any CMS.

Spaceship Earth as a Frame for Communication and Technology

Mike references a specific attraction at Disney's Epcot theme park in Florida: Spaceship Earth, the large geodesic sphere that is the park's icon. The attraction is a slow-moving ride through the history of human communication, from cave paintings through movable type, the printing press, the internet, and up to AI.

Mike uses this as a conceptual frame for the role of web agencies. His argument is that web developers and agency owners are, at their core, communications professionals. They help a client communicate a message to an audience in order to fulfil a need. The medium changes over time — 20 years ago the primary medium might have been print catalogues; today it might be mobile apps or interactive web experiences — but the fundamental purpose remains constant.

The lesson from Spaceship Earth is that every major shift in communication technology has been met with anxiety and disruption, and every one of those shifts has also ultimately extended what humans can express and accomplish. AI is simply the current chapter of that story.

Two Camps on AI: The Apocaloptimist Frame

Mike introduces the documentary that gave him his session title: The AI Doc, or How I Became an Apocalypse Optimist. The film explores the deep polarisation around AI and attempts to hold both extremes in view at once.

Camp One: Utopian enthusiasm. AI will solve everything. It will cure disease, achieve world peace, end hunger, and liberate humanity from drudgery so that everyone can paint for a living. If you are not using AI aggressively, you will be left behind.

Camp Two: Existential dread. AI is destroying humanity. It is taking jobs, degrading the planet, eroding creativity, and poses dangers worse than nuclear weapons.

The documentary filmmaker attempts to reconcile these two positions and ultimately does not arrive at a clean answer. What he does conclude, Mike says, is that "people are people, and this is just the new thing." It will have an impact. It does have a future. But the all-or-nothing framing is counterproductive.

Mike applies this same lens to the agency world. He observes that agency owners are having the exact same polarised conversation about AI that their industry has already had about WordPress, page builders, hosting choices, and every other significant tool shift. His prescription is to step back, calm down, and ask a more useful question: what can this tool actually do for us, and where does it make sense to use it?

Human Connection as the Primary Differentiator

The most sustained argument in the talk is that human connection is what differentiates agencies, not their technology stack. This is positioned as the most important thing for an agency owner to focus on above everything else, including AI.

Mike is blunt about the client relationship: clients do not care what theme you use. They do not care what CMS you use. When a client says they need WordPress, they do not actually need WordPress — they need their goals achieved. He illustrates this with a personal example from an agency he worked at that specialised in banks and credit unions. The agency had hundreds of clients in that vertical. Not a single one was put on WordPress; every one was deployed on Joomla. How many clients came in requesting Joomla? Zero. How many came in asking for WordPress? Many. But they hired the agency as experts, and the agency made the right technical choice for the client's actual needs rather than simply fulfilling the stated request.

Mike distinguishes two types of agencies:

• Order takers — the McDonald's model, where a client asks for a Big Mac and you hand them a Big Mac.
• Expert advisors — the equivalent of a private chef or a high-end caterer who sits with the client, understands their dietary needs, their audience, their constraints, and then makes informed recommendations.

He references "Vito" (likely another summit speaker or well-known figure in the community) who uses a spectrum from private chef to fast food to describe the different agency positioning options.

The conclusion is that human connection is what clients will remember, what they will tell stories about, and what will drive referrals and loyalty in ways that no AI tool can replicate.

The Perception Paradox

Mike introduces what he calls the "perception paradox" through a thought experiment. He asks the audience to think of their favourite web host and then their least favourite. He notes there is a strong chance some people's favourite host is someone else's least favourite.

He then anchors this with a non-hosting example: Comcast, the large US internet and cable provider (also the owner of NBC Universal, which is parodied in the TV show 30 Rock under the name "Cable Town"). Comcast is widely regarded by American consumers as one of the most disliked companies in the country. And yet Comcast consistently wins JD Power customer service awards.

The explanation is that 98% satisfaction can coexist with a perception of universal hatred.

The 2% of customers who had terrible experiences — who waited all day for a technician who never came, or were misled by a salesperson — tell those stories over and over again.

Those human stories travel. The satisfaction data does not.

The implication for agencies: you need to generate positive human stories that your clients and partners tell on your behalf. Not case studies or statistics, but genuine experiences that people repeat naturally. Mike points to Disney as a company that engineers these moments deliberately — for example, when a child's stuffed animal is found lost at a Disney park, the company does not simply mail it back. They photograph the toy going on "adventures"

around the park and send those photos along with the toy, so the child understands why the toy ran away. That is the kind of story a family tells for years.

Mike's position is that AI can help you get there, but it cannot create those moments. That requires human intention and creativity.

He quotes Seth Godin: "Networking that matters is helping people achieve their goals."

Streamlining Success with AI: A Practical Framework

After establishing human connection as the foundation, Mike turns to AI with a pragmatic and sceptical lens. His framework has several components:

Map Your Workflows First Before spending any money on AI, map out where you are actually going to use it. Mike does not advocate for a particular tool — Copilot, ChatGPT, Claude, or anything else — because the tool choice is secondary to the workflow design. What matters is having a system.

Do Not Spend Money Just to Spend Money Mike recounts being at a conference where someone bragged that if you were not "wasting"

at least $20,000 a month on AI tokens, you were doing it wrong. He is clear that this kind of posturing is meaningless. Mistakes and learning costs will happen, but spending without a purpose is not a virtue.

Test and Validate Before Deploying

Identify where AI saves time and where it does not. Run tests. Validate the results. Do not assume AI is better simply because it is AI. Many people spend more time reprompting and correcting AI outputs than they would have spent doing the task themselves.

Track Time Honestly Mike acknowledges that most people hate tracking time, but if you are running a billable-hours model, you need to know where you are actually saving time. Without this data, your sense of AI's value is purely anecdotal.

Start with Non-Billable Work For agency owners whose own time is billable, the first place to look at AI is non-billable internal tasks: bookkeeping, internal communications, administrative processes. This is the lowest-risk environment in which to learn and the easiest place to measure concrete savings without putting client work at risk.

He gives a memorable example: an agency owner who gathered all of the internal emails he had ever sent to his team, trained a model on them, and created a virtual version of himself.

His team now consults this AI before coming to him directly, which frees him to focus on winning larger accounts.

Never Set AI Loose on Client-Facing Work Without Disclosure Mike is emphatic on this point. Any use of AI in client-facing deliverables must be disclosed clearly, with expectations set upfront. He notes that if an agency takes the opposite approach and produces all content as human-written, they should make that equally clear.

He mentions that plugins now exist for Google Docs and Microsoft Word that can verify human authorship through keystroke audit trails, which could be a useful proof point for agencies that want to market fully human-produced content.

AI Security Risks

Mike briefly but pointedly flags that AI tools carry real security risks that are not always discussed candidly in agency communities. He cites: Lovable, a vibe-coding / AI development tool, which exposed user credentials.

Meta, which has faced scrutiny over training data exposure.

His recommendation is to run AI workloads in secure, enterprise-grade environments. He mentions that Hosting.com offers what he describes as an enterprise-grade AI infrastructure product that includes Cloudflare Enterprise top-to-bottom, a server-level AI security suite (described as "Monarchs"), and a vibe-coding tool within the same environment. There is a free tier available. His broader point, independent of any specific product, is to be deliberate and cautious about what data you expose to AI systems and to trust only environments with clear security commitments.

The Case for Accredited AI Education

Mike delivers what he acknowledges may be a controversial position in a summit context where free content and influencer-led learning are common: he believes strongly in paid, accredited AI education from universities.

His argument against influencer-led learning is not that influencers are dishonest, but that their incentives are often misaligned. Many influencer AI courses are affiliate-driven.

Accredited university courses, by contrast, have no skin in the product game (with the partial exception of Microsoft, which has its own AI product). You can trust the curriculum and engage in genuine discussion about what works and what does not.

Institutions and courses he recommends: University of Wisconsin School of Business — AI Essentials Certificate, which covers prompting, thinking, and communicating effectively with AI.

MIT Professional Education — a No-Code and Agentic AI course distinct from the one Mike took (which was heavier on mathematics and machine learning fundamentals).

Microsoft and MasterClass — the "Lead with AI" certificate and an AI Fundamentals course.

These courses cost hundreds to thousands of dollars. Mike argues this cost is an advantage, not an obstacle, for two reasons. First, a financial investment creates a real commitment to apply the learning — free content is too easy to dismiss or forget. Second, the credentialled environment creates genuine accountability and a community of learners who can share honest assessments of what works.

He closes this section by restating that taking his MIT course was one of the best professional decisions he has made.

You can trust the curriculum and engage in genuine discussion about what works and what does not.

Human Connection in Practice: Physical and Memorable Gestures

Mike returns to human connection with a set of concrete, personal examples from his own career. These are not hypothetical suggestions; they are things he has actually done.

80 pounds of pasta sent to WordPress influencers and partners across the community.

Custom stuffed animals made in the likeness of client logos, sent as gifts.

Personalised mugs with recipients' names, sent at the start of the COVID-19 lockdown period when people were isolated.

Personal telegrams — Mike confirms that telegrams still exist and are expensive. He states that he has never once failed to receive a reply when he sent a telegram, and that recipients talk about having received a telegram for years afterward.

Viewmaster slide reels — instead of sending a pitch deck as a PDF, he has had the content printed onto a Viewmaster reel and mailed it physically. The format is unexpected, tactile, and impossible to forget.

The throughline across all of these is that people remember experiences and tell stories about them. The goal is not just to be remembered, but to give someone a story worth telling to others.

Mike quotes two foundational principles from business literature: Zig Ziglar: "You can have everything in life you want if you just help other people get what they want."

Dale Carnegie: "You can close more business in two months by becoming interested in other people than you can in two years by trying to get people interested in you."

He attributes his entire career — including the acquisition of Andrew's company — to building genuine human connections consistently over time. His guidance is to add value constantly, introduce people to each other, help without expectation of return, and do this not once but as a sustained practice. The business results, he says, follow naturally.

Closing Remarks and Post-Session Exchange

Mike closes the session by apologising for the pace ("speed run") necessitated by the schedule, encouraging attendees to connect with him online, and reiterating the QR code for the gift and matchmaking survey.

Andrew, returning after the session, offers warm commentary. He confirms that Mike worked on him for three years ("three coffee offers") before the acquisition of his company in December 2019, and reflects on the extraordinary timing — the deal closed on 31 December 2019, and the world "closed down" shortly after with the COVID-19 pandemic. He also mentions that another Atarim colleague (Ali) was left craving pasta after hearing Mike's anecdote.

A community in-joke surfaces around the name "Krakoa.com," which Andrew explains was born at an earlier WebAgency Summit (Summit 6), referring to another attendee nicknamed "Krakoa." This contributes to the sense of long-running community relationships that underpin the summit itself.

About the speaker Mike Demopoulos Co-founder, Apiotics

Mike Demopoulos is co-founder of Apiotics, where he works on AI strategy for agencies, taking a clear-eyed view of where the technology genuinely helps and where the hype outruns reality.