A recurring paid evening for regional decision-makers, and the talks that come out of running it.
Everything on this site was argued in front of an audience before it was published. age:AI is a recurring evening built for decision-makers who are being asked about AI and do not want a vendor pitch, and the talks below are the same arguments in a shorter form. If a claim on this site survives a room of sceptical operators, it is probably safe to print.
age:AI is an AI awareness evening for regional decision-makers, run about twice a year. People pay to attend, which changes the contract in the room: an audience that bought a ticket is owed an argument rather than a demonstration of enthusiasm. The seventh edition was held on 14 September 2026 during Waterloo Tech Week.
Each of these is a position this company actually holds, with a page on this site behind it. None of them is a product walkthrough. If you want the product, the demo does that better than a slide ever will.
The security owner carries the risk if they say yes. The operating owner carries the cost of saying no. Nothing moves until both can say yes to the same controlled capability, and most AI programmes stall in exactly that gap without anyone naming it.
Inference does not belong in a zero-error critical path, and a deterministic process should stay deterministic. How to tell the work a model genuinely improves from the work it merely makes less predictable, with examples of governed software that never calls a model at all.
Pinning inference to a jurisdiction is a real control and it costs real capability: of 104 models in the current generated catalogue, 13 run in-region in Canada and every current frontier model routes outside it. How to make that trade deliberately instead of discovering it later.
Your policies, applied in the system, not a PDF and not a black box. What changes when the description of your controls is generated from the same code that builds them, and a test fails the build when the two disagree.
Most private AI means the vendor holds your data privately. What actually changes when the platform runs in a cloud account you own, the compute invoice has your name on it, and the monthly fee buys a service rather than access.
Bespoke pipelines built at great expense were simplified or replaced by managed services within months, and the scarce value moved up the stack. What that means for how an organization staffs, buys and decides what not to build.
Every organization wants AI, and many have banned it. Not because it does not work, but because using it means sending their data to somebody else's computer.The opening of most of these talks
| Format | What it is | Best for | What the room leaves with |
|---|---|---|---|
| Keynote | One argument, conference length, built for a mixed audience | Association events, conferences, member evenings | A position they can repeat accurately to their own board |
| Executive briefing | A closed room with one organization's leadership, questions throughout | A leadership team that has to make a decision soon | A view of where their own exposure and their own opportunity actually sit |
| Panel | A seat on somebody else's programme, as the technical dissent | Programmes that need a voice arguing against the consensus | A disagreement worth having, conducted without hype |
| Workshop | A working session rather than a talk, with the material in front of the room | Teams drafting their own AI position or posture | A first draft of their own governance position, written by them |
Anything on this page can be shortened, combined or taken apart for your audience. What will not happen is a talk that turns into a product demonstration halfway through: if the room wants to see the platform, we will say so and show it deliberately, or point them at the demo.
Send a note through the contact form or book a call. There is no speaker agency and no booking desk, so the reply comes from Sam. Four things make the answer fast, and the fourth one is the one most requests leave out.
Speaking terms depend on the event, the audience and the work involved, and they are agreed in writing beforehand. Ask, and you will get a straight answer for your specific event, including whether we think another speaker would serve your programme better.
No, and an organizer is entitled to hold us to that. The talks are arguments about governance, ownership, cost and where inference does not belong, and they stand up whether or not anyone buys anything. org.tech is named because pretending the speaker has no commercial interest would be worse, not better.
If your programme wants an explicit product session, that is a different request and a legitimate one. Ask for it directly and it will be labelled as what it is.
Yes, and it is the part worth spending a call on. The arguments do not change; which of them leads, and how much technical depth the room can use, changes completely between an audience of security leaders and an audience of owner-operators. Tell us who is in the room and what they are deciding.
Usually yes, agreed in advance. One caveat worth knowing: live demonstration of a real deployment sometimes shows material that should not be recorded, so any demo segment is handled separately from the recording.
Yes. Association member sessions, client evenings and closed leadership briefings are among the most useful formats, because the room already shares a context and the conversation gets past first principles quickly. Advisors who want to host one for their own clients should read the partners page first, particularly the part about what we will not do.
It runs about twice a year and the next date is announced when it is set. The seventh edition was held on 14 September 2026 during Waterloo Tech Week. Ask through the contact form to hear about the next one.
If a talk raised a question about your own environment, the assessment is the bounded way to answer it.