The invitation arrived the same way it always does from Paul Greenberg and Brent Leary. Not as a transactional conference registration, not as a vendor-sponsored press pass. Just an invitation — the kind that carries the quiet assumption that you belong in the room.

I have been in technology long enough to know how unusual that is. Most industry events are organised around current relevance. You are invited because of where you work, what you cover, or what you can offer a sponsor. The moment any of those things changes, the invitation tends to stop arriving. IRL — In Real Life, now in its second year and held on 12 March 2026 at Atlanta Tech Village in Buckhead — does not operate that way. Paul and Brent do not audit your LinkedIn title before they reach out. Decades can pass. Roles can change. The invitation holds.

That is not a small thing. It is, in fact, the entire argument.

Atlanta does not try to impress you

People who have never been to Atlanta assume it is not a conference city. They are right, and that is precisely why IRL works there. Atlanta Tech Village is a working technology campus, not a convention centre dressed in sponsor banners. The corridors are functional. It's got the world's busiest airport and more hotels than needed. If you remove the ambient noise that most industry events rely on to fill silences. When there is no spectacle competing for attention, conversations have to carry the day. At IRL26, they did.

The room itself is the thesis

I looked around at various points during the day and noticed something that a conference organiser cannot engineer: competitors were having genuine conversations. Analysts from different firms were sharing actual observations rather than positioning statements. Analyst relations (AR) leaders — people whose professional function is, in part, to manage what analysts hear — were relaxed. Not because their guard was down, but because the culture of the room does not reward performance.

The CRM Playaz community, which Paul and Brent have built across years of consistent live broadcasting and intellectual generosity, has a texture to it. People know each other across time, not just across industries. There is a running joke segment called "Stump the Godfather" that has become a genuine institution because someone in the audience once started doing it organically, and Paul and Brent let it become part of the show. That is not a content strategy. It is a community forming in real time, being recognised, and being incorporated. The result is that when those same people sit in a room together in Atlanta, the relationship infrastructure is already built.

One detail stays with me. A speaker from a major Atlanta sports and entertainment organisation — discussing how they had scaled their live event business in the years following the pandemic — mentioned almost in passing that sports organisations share competitive artificial intelligence (AI) intelligence with each other every day. Strategy, contracts, ticket pricing: you do not share those. But which AI platforms are working, what adoption challenges have emerged, where the models are falling short — that you share freely, across rival organisations, because the problem is common and the industry benefits from solving it together.

That is not a technology insight. It is a relationship insight. And it landed differently in that room than it would have in a ballroom at a larger event, because the room was already primed to receive it.

What the sessions said when you read them together

IRL26 ran on a single main stage and a parallel community room broadcasting live shows simultaneously — a format Paul had described in advance but which you can only understand once you are inside it. The sessions were genuinely diverse in register. The morning carried productive scepticism. A panel on what some are calling the "SaaS apocalypse" — the idea that AI agents will render the enterprise software stack obsolete — reached a measured conclusion: markets consistently overestimate short-term disruption and underestimate long-term transformation. Nobody replaces a customer relationship management (CRM) system that has twenty-five years of encoded process, edge cases, and institutional knowledge in a single agentic sprint. Tribal knowledge, regulatory compliance, and trust are genuine moats, not nostalgia.

A parallel session on workforce transformation was less measured, and usefully so. The argument — stated plainly — was that enterprise leadership has consistently inverted the transformation sequence: technology first, people last. AI is repeating the pattern. Layoffs announced for quarterly optics rather than strategic clarity. AI budgets allocated while training budgets contract. The concern in that session was not about AI capability. It was about institutional courage.

The afternoon shifted register entirely. Vala Afshar from Salesforce offered a framework I found genuinely useful: physical AI systems — the Waymo autonomous vehicle being his illustration — are effectively sense-understand-decide-act architectures running in real time, and when enterprises implement agentic AI, they are adopting the same operating system. The practical grounding for this was Salesforce's own internal adoption, with around 83,000 daily active users of Agentforce, and a robotics demonstration in which an autonomous robot conducting fire safety inspections automatically creates and routes a Salesforce Field Service work order when it identifies a deficiency. The point was not the robot. The point was that the action layer and the system of record are now the same loop.

Vala also offered what I thought was the most quietly important reframe of the day: in a world where every engineer uses an agent as a working partner, individual productivity is no longer the unit of measure. All work is teamwork. The team now includes non-human members.

Esteban closed it

Esteban Kolsky opened his closing session by asking the audience to count to three on their fingers. Americans start with the thumb. Central Europeans start with the index finger. In Japan and Korea, the hand opens from a fist. Three different representations of the same mental model. He let that sit for a moment before making the point: this is precisely where we are with AI. Everyone is describing it differently — the next search, the end of software, an automation layer, an intelligence commodity — and most of them are pointing at the same underlying reality from a different cultural or professional starting position.

His central argument was that the correct historical analogy for this moment is not the commercial internet. It is the personal computer entering the enterprise in the 1980s. We digitised work then. We are democratising intelligence now. The difference in adoption velocity is striking: the commercial internet took five to six years to reach 33 per cent penetration. Large language model access went from near-zero to 66 per cent adoption in roughly one year, and to 96 per cent in approximately two and a half years. The cost collapse that made that possible is what turns a technology into infrastructure, and infrastructure is what changes how organisations are designed — not just what tools they use.

Esteban's three required actions for organisations were: treat AI as infrastructure, not as a feature addition; change processes rather than layering AI onto dated workflows; and build AI literacy everywhere, not only in engineering. The third is the one most organisations are avoiding, and he said so directly.

He also said something about his own work that took some honesty: the reading, synthesising, summarising, and aggregating that had defined his analytical practice for two decades is now being automated to a meaningful degree. His response was not despair. It was reorientation. The value of analytical work is not the output. It is the intelligence behind the output — the judgment, the relationships, the pattern recognition built across years of conversations in rooms like this one.

What IRL proves

Every AI platform at IRL26 — and there were several, presented thoughtfully — is ultimately trying to solve a version of the same problem: how do you make intelligence available at scale, embedded in experience, connected to action? Kaltura demonstrated an agentic avatar with what they described as six senses: face, voice, eyes, screen awareness, a connected knowledge base, and an action layer reaching into enterprise systems via application programming interfaces (APIs). The live demo broke partway through, switching languages unpredictably. The presenter kept going. The community appreciated the honesty more than a polished rehearsal would have earned.

None of that technology, however capable, builds the thing that IRL actually runs on. The platform Paul and Brent have created is not a streaming service or a conference brand. It is a sustained practice of remembering people — not their current titles, but the people themselves. The result is a room where a private equity operating partner, a former Amazon vice president, a residential smart-home integrator, an AI strategist from a major enterprise software company, and a collection of analysts from competing firms can sit through a day of genuinely contested ideas without the usual institutional defensiveness.

Relationship capital of that kind does not appear in any AI readiness assessment. It does not show up in a technology vendor's total addressable market calculation. But it is, in practice, what separates organisations that can move quickly in uncertain conditions from those that cannot. The trust infrastructure is already built, or it is not. You cannot sprint-build it when you need it.

Paul and Brent have been building it for years, one live show at a time, one invitation at a time. Atlanta just gave it a room for the day.