After years of lanyards, carpets, and expensive coffee, I've realised the real ROI from events isn't on the show floor—it's in what happens months later, when you've stopped trying so hard.

By Ben Anchor — Wednesday, 2 September 2026 · Listen to the podcast episode

I've just spent a week in Nepal at a wellness-meets-ambition gathering with about 100 people, and in nine days I'll be standing in Amsterdam with 45,000 others at IBC. Same planet, wildly different planets.

One's got sunrise yoga and summit circles. The other's got branded lanyards and €8 espressos. Both are called 'events'. Both promise connections. And yet I'd wager most people book the wrong one, show up with the wrong expectations, and leave disappointed.

This bonus episode—recorded at Nomad Basecamp in Pokhara with investor Sheikh Saifudeen and e-commerce founder Biplab Karki—pulled apart that tension. We talked about reach versus peers versus procurement, about spiritual reprogramming and agentic commerce, and about why the hustle to close a deal on the stand is usually the least valuable thing you can do.

The Three-Way Split: Reach, Peers, Procurement

I floated a framework: talks are about reach (getting your message out), conferences are about peers (building your network), and trade shows are about procurement (doing deals). It's a tidy model. It's also probably wrong.

Sheikh's take was that people come to events—whether it's a spiritual gathering in the Himalayas or a procurement fair in Europe—to improve their lives. Maybe it's a single idea. Maybe it's a relationship. Maybe it's just permission to think differently. The form of the event matters less than whether you walk away with something that shifts your trajectory.

Biplab, who's been to startup expos across Asia, agrees it's mixed. An exhibition can yield clients or collaborators or just a better sense of what's happening in an adjacent industry. He went to a goods expo in Thailand not to buy anything, but to understand the economy and lifestyle. That's a peer move at a procurement event.

I think they're both right, and my neat categories are less useful than I'd hoped. But the underlying question stands: why are you there, and are you set up to receive what you actually need?

The Myth of the On-Site Close

Ten years ago, I'd rock up to a stand with a target: X meetings, Y leads, Z closes. The pressure was real, especially if you'd spent five figures on the footprint. Sheikh's been on both sides—exhibiting for his own e-commerce business and working procurement for others—and he remembers that expectation to recuperate funds. But he's also clear: the most valuable thing was always the contact list. Emails. Ways to stay in touch. Because deals rarely close on the floor.

He reckons it takes five to nine touches before trust is built and a deal gets done. You meet someone, they get an email, maybe a follow-up, maybe they click through to your site. A few months later, when the need aligns, then they buy. The stand is just touch one.

Biplab frames it as ROI over time. You invest now—booth, travel, team—and you get direct ROI in three, six, twelve months. But you also get indirect ROI immediately: branding, content, social proof, understanding. If you're only measuring the former, you're missing half the value.

I've lived this. I've had conversations at IBC that led to nothing for two years, then suddenly the person's at a new company, remembers me, and we're in business. If I'd been fixated on closing them in Amsterdam, I'd have written them off as a waste of time. Instead, I stayed in touch, stayed useful, and let the relationship mature.

The insight: stop trying to force the close. Start collecting the threads.

AI as Co-Founder, Scheduler, Spiritual Advisor

We pivoted to AI, because it's what the main podcast is about and because both guests had strong views. Biplab's building an agentic commerce platform—think AI agents running your e-commerce ops, talking to your ERP, optimising your supply chain. He reckons you could halve your headcount and double your efficiency. Bold claim, but the trend's real. Google's betting on it. So are a dozen well-funded startups.

Sheikh uses AI differently. He's plugging his thoughts, goals, and internal blocks into a model and treating it like a personal advisor. Not just for business strategy, but for subconscious reprogramming. He's asking: where am I limiting myself? What beliefs are holding me back? What do I need to shift internally before I take the next leap? It's consciousness work, but with a chatbot.

I loved that framing. AI as the best on-demand co-founder, consultant, and therapist rolled into one. It sees what you can't see—or at least, it mirrors back patterns you've been too close to notice.

For me, the next 12 months are about taking that idea into broadcast scheduling. I'm building Ancast Nowcasting: a custom model that doesn't just learn from a broadcaster's historical data, but ingests live public signals—search trends, social media, cultural moments—to give schedulers a superpower. Not an LLM wrapper. A genuinely new intelligence layer. I'm at IBC to talk about it, then South Africa, then the Middle East. This time next year, I want pilot studies done and proof of concepts in production.

What I'm Taking Away

Three things stuck with me from this conversation.

First, events are format-agnostic when it comes to value. A 100-person basecamp and a 45,000-person trade show can both deliver life-changing connections—or neither. It's not about the carpet or the catering. It's about what you're open to receiving.

Second, ROI is a lagging indicator. If you're measuring success by what happens at the event, you're looking at the wrong timeframe. The real return shows up months later, when the trust compounds and the timing aligns.

Third, AI is only as good as the questions you ask it. Whether you're using it to optimise a supply chain, reprogram your subconscious, or build a nowcasting engine for schedulers, the magic isn't in the model—it's in how clearly you've defined the problem.

I'll be in Amsterdam from the 11th. If you're there and want to talk broadcast AI, Nowcasting, or just compare notes on trade show coffee, book a slot. I won't try to close you. I'll just collect the thread.

And maybe, in six months, we'll both be glad we did.

---

Fairground AI's generated channel launched to brutal reviews—and that's beside the point. The real shift is infinite supply hitting fixed demand, and it points straight at a question most of us are ignoring: when the machine does the work, who's left holding the pen?

By Ben Anchor — Wednesday, 19 August 2026 · Listen to the podcast episode

On 13 August this year, the first fully generated television channel went live on Roku. Within 48 hours, the internet had christened it 'AI slop.' I'll be honest—my first reaction matched theirs. My second reaction, about a day later, was that arguing over the quality misses the entire story.

The channel is called Fairground AI Creator TV. Every frame is generated. No actors, no crew, no shoot. It looks like a channel, it schedules like a channel, it sells ads like a channel. And yes, if you sit and watch it, the failure modes are exactly what you'd predict: faces that drift between shots, physics that's nearly right (and therefore worse than obviously wrong), and a total absence of the thing that makes narrative work—a scene built because somebody wanted it to mean something.

But here's why I don't want to spend this article on that, and it's the same reason I wouldn't have spent one in 1999 explaining that video on the internet looked terrible. Of course it did. The quality of the first version has almost no predictive value. What has predictive value is the cost curve underneath it.

The Economics Are the Story

Free ad-supported television is brutally tight. You make money on advertising against watch hours. The rates are low, so the entire business is a race between the cost of filling 24 hours a day and the revenue those hours can generate. Which is why the free streaming ecosystem is largely built on library content—formats nobody's licensing at a premium any more, back catalogue, the stuff whose marginal cost of re-airing is close to nothing.

Now watch what generated content does to that equation. It doesn't lower the cost of content. It removes the floor entirely. No acquisition, no rights window, no residuals, no revert. If your channel costs almost nothing to fill and it earns even a small amount, it's profitable on day one—and it's profitable at a scale of one channel or 1,000 channels.

That's the bit that matters, not the quality. The fact that the supply side has become effectively infinite while the demand side—which is human attention in the evening—has not moved at all.

Infinite supply against fixed demand is not a technology story. It's a pricing story, and it points in one direction. If the amount of watchable material goes up by orders of magnitude and the number of hours in an evening does not, then the value of any individual hour of content falls. Not the value of the best hour—the value of the average hour. And most of our industry is not in the business of making the best hour. Most of our industry is in the business of reliably making the average one, on schedule, on budget, to a standard. That's where the employment is.

So the honest read is not that generated content will replace prestige drama. It's that it competes directly with the enormous middle of the market: the filler, the acquired formats, the daytime, the things that exist because the schedule has to be filled. And that middle is where a very large number of people in this industry actually work.

The Company That Runs Itself

This connects to a claim I heard at an AI summit recently: that within roughly 18 months, we'll see a genuinely autonomous company. Not a company that uses AI—a company where the agents do the work end to end. They find the customer, they make the thing, they invoice, they handle the support ticket, they file the accounts. No human sits in the loop on any given day. The summit framing is quite specific about the stages: you go from co-pilots (where most organisations actually are) to agents doing bounded tasks, to autonomous workflows that run without a person initiating them. The end state they describe is an organisation chart with nobody on it.

My instinct is that this is the same species of claim as the fully automated newsroom, which has consistently failed to arrive. But the evidence is genuinely split. Task-specific agents went from under 5% adoption last year to a forecast of around 40% by the end of this year. Something close to three-quarters of businesses say they intend to deploy agents inside that window.

Against that: over 40% of agentic AI projects are forecast to be cancelled before the end of next year on cost and on unclear business value. So the picture is not a straight line up. It's a very large number of organisations starting, and a very large number of them stopping.

But there's a harder objection than cost. A company is not just a set of tasks. A company is a legal thing. Who signs?

The Wall

This is the wall, and it's the most useful thing in this entire conversation. A model can execute the work of a business. It cannot be the business. Legal personhood requires a person or a registered entity. A bank account requires a legal owner. A contract requires a party that can be sued. So every one of these so-called zero-human companies, when you look underneath, is running inside somebody's limited company with a director, and that director is on the hook.

The machine executes. The human signs.

Which means the 18-month claim, in the form it's usually made, is not wrong so much as it is misdescribed. What's arriving is not a company with no humans. It's a company with one human, whose entire remaining job is accountability.

And that's our world exactly. A broadcaster is not a set of tasks either—it's a licence. Somebody holds the licence; that somebody answers for what went out. It doesn't matter whether a person, a playout automation system, or a generative model put the frames on the air. It doesn't transfer to a vendor because you bought a tool, and it doesn't transfer to a model because the model made the choice.

So take that AI-only channel and ask the question that nobody in the coverage asked: if it broadcasts something defamatory, who is liable? Not the model. Not the creator who typed the prompt, probably. The entity that published it—which is a company with directors, with a registered address, sitting behind $4 million of venture funding.

The channel has no people in it. The business absolutely has people on the hook.

What to Actually Do

Somebody running a media operation hears all this on a Monday morning. What do they actually do?

Three things, and none of them are glamorous.

First, be extremely honest about which part of your output is the middle of the market. If a meaningful share of what you make exists to fill a schedule rather than to be sought out, that's the part exposed to infinite supply, and you should be planning for it now rather than being surprised by it in two years.

Second, stop evaluating this on quality and start evaluating it on cost per hour, because quality is the variable that moves fastest and cost is the one that decides.

Third, and this is the one I'd actually act on this week: write down where accountability sits in every automated process you already run. Not where the technology sits—where the signature sits. Because the organisations that will handle the next three years well are not the ones with the best models. They are the ones that can answer, instantly and in writing, the question of who is responsible when it goes wrong.

That's the thing the machine cannot take from you, and it's also the thing that quietly turns out to be the job.

My Honest Prediction

I think we'll absolutely see a business that operates with no staff and generates real revenue, and I think it will happen sooner than 18 months, because in a narrow enough niche it has arguably happened already. What we won't see is a company without a person who answers for it, because that's not a technology problem—it's a law problem, and the law is not moving at the speed of the models.

So the channel with nobody in it still has somebody behind it. It always did. And the question worth asking, about any of this, in our industry or anyone else's, is not what the machine can now do. It's who's standing there when it gets it wrong.

---